quitchy@yahoo.co.uk
"Quitch" <quitch@fickle.demon.co.uk>
new email address: quitch@x-stream.co.uk
Reply-To: "Quitch" <quitch@fickle.demon.co.uk>
To: "David Sellick" <switeck@hotmail.com>
Subject: Re: AI Guide
Date: Sat, 31 Jul 1999 23:36:51 +0100
From quitch@fickle.demon.co.uk Sat Jul 31 15:36:53 1999 

Thanks for your reply. 
As you seem to be busy, I'll keep this brief. 
TAMEC is no longer closing, it has decided to stay open and I think will be merging with another site soon? 
Anyway, thank you for your time, and your AI. 
I'm off to make a little Voodoo Doll of the man (or woman) who designed the CaveDog AI and said "Wouldn't it be a fantastic idea to have it build construction units, but not just a few, hundreds, that way demonstrating the flaw where it blocks itself in, fails to manage it's resources, doesn't build things that it needs, and basically every other flaw we cunningly designed, implemented, and slipped past QA" 
Hurrah cheer the board, and the TA AI 3.1 was born. 
It certainly pads the ego :) It's good for anyone who's average or below, but good players just smash the flaws that the AI has built into it, so I've written into my readme file, things people should not do to make it harder (not use Dragons teeth for a start). 
I'll certainly have a look at your fix. 
Thanks for your time. 
Paul Cocker aka Quitch
http://members.xoom.com/XXX_Clan/
http://www.fickle.demon.co.uk/Index.html
--------
>From: "Quitch" <quitch@fickle.demon.co.uk>
>Reply-To: "Quitch" <quitch@fickle.demon.co.uk>
>To: <switeck@hotmail.com>
>Subject: AI Guide
>Date: Sat, 31 Jul 1999 12:10:37 +0100
>
>I would like to thank you for producing such an excellent and through guide to AI.  I have been working on my own AI recently, the "Queller" AI.  The idea is that it should better all current TA non-cheating AI's, and it's beaten everyone but yours (that I know of).  It hasn't played yours yet.  It's also caught out my testers, and destroyed the odd Boneyards player or two :)

Thanks, I only have time for quick answers to some of your questions. Imagine the shock of people who played only the Cavedog ai when they first played mine... For a while I was doing blind tests the same way you are and it certainly pads the ego a bit to make an ai that can at least occassionally beat good players.

>Anyway, back to your guide.  Well laid out, well written, incredibly useful.  I find it quite interesting that the v2.0b1 provided a better AI than 3.01.  It is only recently that I have gotten back into TA, and so I cannot remember that AI, only the terrible construction frenzy of 3.01, which could be beaten by 5 fleas I imagine (I do it with 5 flashes :)).
>
>May I ask you a question?  You say that the computer rarely builds geothermal, and it is a pain to do so.  I have had that problem (makes you  want to scream doesn't it :)), with shipyards.  Is there anyway to persuade it that it should build one?  Currently I am testing my default profile on "The Cold Place", but the AI will only build land units, and very rarely shipyards.  My current theory is that it will build land stuff as long as it has room, and only then move to the sea.  I was hoping that you can prove me wrong and suggest a way to get it interested in ships.  For this I would be most grateful.

Shipyards on a big land/water map like Metal Isles or 2 continents is almost an aberration -- like you said, the ai wants to fill up the LAND first. The trick I use to partially get around that is to have it build lots of C-airs, which seem more willing to wander further afield to build high-weight buildings like shipyards and geothermals. BTW, The Cold Place isn't really appropriate as a "default" type ai, it's more of a wind/land/some water type ai - with Arm ai going more land perhaps and Core's going slightly more water than Arm's. My Default ai is designed to NEVER build ships, which causes people to complain about hundred isles ai response -- Cavedog made THAT map a default map.

>You have no idea how hard it was to track down your AI (Mostly Harmless).  Everyone went on (on the TA forum), about how it was the best non-cheating TA AI, but no one knew where I could get it, it's taken me several weeks to find it.  I must admit to admiring how you have almost completely prevented nano-stall for your AI.  I didn't watch a long game, but it handled it's resources very well, quite impressive.

www.tamec.org is about the only remaining site that has it, and it's closing its doors.

>Anyway, thanks for your time.  I don't even know if you still play TA (moved onto TAK?  Or just given up?).

Giving up, TA:K is not my cup of tea -- and Cavedog made many of the same mistakes in TA:K as in TA -- as well as adding a whole list more (units won't even attack while on the move). BTW, my TA Bugfix v1.1 is out at TAMEC too - it is my (and quite a few others) attempt to fix many bugs in TA. Please download it and skim the readme with it.

David
--------
From: "Quitch"
Subject: Bugfix 1.1
Date: Sun, 1 Aug 1999 10:22:11 +0100

I am having a problem applying your patch. I have CC and version 3.1 (which for some reason misaligns the main menu buttons in TA when I use CC). 
I have moved all the files into their appropriate spots, but get the following error message every time to I try to run your fix: 
The file 'OLEAUT32.DLL' is out of date. This program requires a newer version. 
I am running Windows 95 OEM. Is there anyway I can correct this problem? 
Thanking you in advance. 
----------
Subject: Re: Bugfix 1.1
Date: Sun, 1 Aug 1999 15:53:55 +0100

I'll search the web :) 
I believe your non-easy instructions said I needed to open a ccx file, but every time I try that with my HPI Viewer, it crashes. 
It's not a friendly program :) 
By the way, after reading your readme, it is now my mission to get this patch working, proper working Arm Warriors! Increased chance of Geothermal!!!!!! It's like a miniature heaven :) 

>Subject: Bugfix 1.1
>Date: Sun, 1 Aug 1999 10:22:11 +0100
>
>I am having a problem applying your patch. I have CC and version 3.1 
>(which for some reason misaligns the main menu buttons in TA when I use 
>CC).
>
>I have moved all the files into their appropriate spots, but get the 
>following error message every time to I try to run your fix:
>
>The file 'OLEAUT32.DLL' is out of date. This program requires a newer 
>version.
>
>I am running Windows 95 OEM. Is there anyway I can correct this problem?
>
>Thanking you in advance.

To start this off, I say "ARGH!"

Anyway, Creeperman designed the installer -- which unfortunately seems to work fine on Win 98 and probably Win 95 OEM C - but chokes on earlier versions of Win 95. I can suggest getting newer versions of 'OLEAUT32.DLL' via Microsoft's website, or just search the web for it.

Barring that, just do a manual install of the patch using HPI viewer and HPI packer.

David
--------
Subject: Re: Bugfix 1.1
Date: Mon, 2 Aug 1999 22:37:46 +0100

Thank you for all your help, I now have your fix working, and have just carried out all your packing instructions to help TA run faster. 
Damn Cavedog, I almost finish the default profile my Queller AI (non-CC version), and they go and release another unit!!!! 
Sigh. 
Paul Cocker aka Quitch
--------
Subject: Re: Bugfix 1.1
Date: Mon, 2 Aug 1999 23:46:44 +0100

Argh! It's me again :) Just a beam of light in everyone's day :) 
I have been following your instructions at the end of your bugfix patch to help TA to run faster (the packing and unpacking stuff). I tried packing 3 of the maps from the Cavedog site and the latest unit release into a file called rev31.gp3 just like it said. However when I went into the game I couldn't select the maps or the unit. 
Do I need some subdirectories? Do I need to pack them into the current rev31.gp3 rather than moving it so as to create a new one? 
Since discovering that the AI Commander is no longer confused (as far as I could see from my limited tests) by multiple targets, I am very happy :) Is this something to do with the tolerance values? 

Thanking you in advance. 
Paul Cocker aka Quitch
--------
Subject: Re: Bugfix 1.1
Date: Tue, 3 Aug 1999 21:10:32 +0100
From quitch@fickle.demon.co.uk Tue Aug 03 13:24:22 1999 

Yes TA:K was less in the way of "progress" and more a giant sideways step. Rather than calling it better, I always describe it as "different". 
I'm not a big fan of the swords and stuff, but if I ever see anyone mention customisable AI, I'll be buying it :) The default AI is *slightly* better than 3.1 TA, it gets some units out before construction, and it actually handles it's Commander/Monarch well. Stunning :) 
About the Commander. Believe me, it's changed aright. I'm very happy about it :) Before, you sent your 3 flash tanks at the AI Commander, and he'd flip from one to the next without actually doing anything, now they start attacking him, and he casually flicks around and D-Guns the little devils :) Unfortunately he still has that damn 15 second pause when shot bug. Oh, and don't get me started on what you call the "Coffee Break Bug", I call it a "Pain In The Ass" :). May the TA AI/Joke department die painfully. 
Good thing you don't write a list, the e-mail would be too big for a humble modem to download. 
Sigh, what exactly *did* they patch, considering almost every major bug still remains. Makes you worry about the original release if we're now in version 3 and there are still more bugs than can be named around. 
I guess QA is a myth to them. 
Paul Cocker aka Quitch
--------
>Subject: Re: Bugfix 1.1
>Date: Mon, 2 Aug 1999 23:46:44 +0100
>
>Argh! It's me again :) Just a beam of light in everyone's day :)
>
>I have been following your instructions at the end of your bugfix patch to 
>help TA to run faster 9the packing and unpacking stuff). I tried packing 3 
>of the maps from the Cavedog site and the latest unit release into a file 
>called rev31.gp3 just like it said. However when I went into the game I 
>couldn't select the maps or the unit.
>
>Do I need some subdirectories? Do I need to pack them into the current 
>rev31.gp3 rather than moving it so as to create a new one?

Study the structure of existing map UFO's -- your resulting map ufo's *MUST MUST MUST* be identical or it won't work. And it gets ugly if there's duplicates of the maps seen by TA as well.

>Since discovering that the AI Commander is no longer confused (as far as I 
>could see from my limited tests) by multiple targets, I am very happy :) 
>Is this something to do with the tolerance values?

Quite possible it's only your perception of such as an improvement. Or possibly a result of freed-up processor cycles devoted more to AI. Either way, I dunno.

>Thanking you in advance.

There's a LOT of things that Cavedog could've done to FIX to, or let "slip out" so 3rd party groups could fix TA. That didn't happen, so a lot of TA's inner workings are as of yet not understood -- and probably never will be.

TA:K goes one step further and grandfathers many bugs from TA straight into TA:K. I'd give a list, but then I'd want to go hunt Cavedog employees down...

David
--------
Subject: Re: Bugfix 1.1
Date: Tue, 3 Aug 1999 21:01:48 +0100
From quitch@fickle.demon.co.uk Tue Aug 3 13:24:29 1999 

I wish I could remember or download 2.0b1 again. I think I'll look for it on an old CD or something. 
Well since I've a lot of free time on my hands, I added CC units over the last 3 days. God, I thought typing in codes was boring...........I was right :) 
CC doesn't seem that complex, there's the added pain of introducing extra plants, without crippling the AI's resources. But generally, the Hovercraft are just weaker vehicles, which are faster, have longer range and slightly more powerful guns. Never mind, I'll get it. 
Do you think they'd been having a heavy drinking session when they made the 3.1 AI, and it was all just a big joke? After all, in just 10min of fiddling.....I'd improved it. It truly was terrible. 
Anyway, I've got to test my AI against yours tomorrow :) If it wins, I'll have a nicely padded ego :) Anyway, the default profile for TA is virtually finished, and I've set my testers on the CC version now. 
I'm sure I'll e-mail you again, brace yourself :) 

>Subject: Re: Bugfix 1.1
>Date: Mon, 2 Aug 1999 22:37:46 +0100
>
>Thank you for all your help, I now have your fix working, and have just 
>carried out all your packing instructions to help TA run faster.
>
>Damn Cavedog, I almost finish the default profile my Queller AI (non-CC 
>version), and they go and release another unit!!!!
>
>Sigh.

It took me about 6 months to add the TA:CC units to my ai. And even 
afterwards I feel (and still feel) that the ai engine was BETTER in v2.0b1 
than in v3.0 and v3.1...

David
--------
Subject: Re: Bugfix 1.1
Date: Wed, 4 Aug 1999 14:34:12 +0100
From quitch@fickle.demon.co.uk Wed Aug 04 10:05:00 1999 

Did you remove the Commanders laser? That would account for the new, improved Commander. 
Anyhow, I have just switched my AI over to your Nukes rather than the AI_Nuker file (after reading your readme), and I need to know, do these AI units replace the normal ones for the AI, or will it attempt to build *both* of them, unless I turn off the original units. 
Basically, do I need to only weight the new AI units, or do I also have to use weights to turn off the Nukes that the AI cannot use? 
Also, as both your Nuke file and the AI_Nuker file use the same codes, which takes preference? 
Finally, in your readme (I got the file with your Counter Strike AI), it gives the CORE Neutron the same cost to fire as the Nukes, I take it this is a typo? It isn't any different from the EMP (except in who it freezes) is it? 
--------
Subject: Re: Bugfix 1.1
Date: Wed, 4 Aug 1999 18:00:12 +0100
From quitch@fickle.demon.co.uk Wed Aug 04 10:05:21 1999 

I just noticed that you said AI "engine". Do you just mean the weights, or some of the programming too, such as the Commanders behaviour and such? 
By the way, in a previous e-mail, I said it had taken me just a couple of days to add CC units, by that I only meant the codes, I haven't gone and created an entire AI in only 3 days! :) 
--------
Subject: Re: Bugfix 1.1
Date: Wed, 4 Aug 1999 18:17:02 +0100
From quitch@fickle.demon.co.uk Wed Aug 04 10:43:08 1999 

However, in the single game of TAK I played (yes only 1 so far :)), I found the AI monarch handled combat better (even if at the start of the game he ignored the Mana point next to him and tried to build some unit production thing, that screwed him). In combat he would use his powerful spell close up (Taros I think he was from), then he would run away hitting my troops from afar. When his mana recharged he charged in, killing more people. I tell you, after getting used to beating an AI Commander with 3 flash tanks, losing about 25 close combat troops came as a bit of a shock. 
Don't worry, I already have your AI guide, and in fact I have everything by you I could find on TAMEC. The AI guide is what started this conversation in the first place remember :) 
I'm prepping my AI to mess with Mostly Harmless, I'm just trying to get it near consistently to build a factory early, the problem is when I do get it to, it cripples itself by building another almost straight away. And using the current weights it plays very well, but not quite well enough to stop a quick raid. 
TA fixed things in their new versions? Wow, I thought they just created new ones and added some nice shortcut keys (that obviously takes them a couple of months to do :)). 
And hey, what's this, TAK where they're doing it all over again! 

>Subject: Re: Bugfix 1.1
>Date: Tue, 3 Aug 1999 21:10:32 +0100
>
>Yes TA:K was less in the way of "progress" and more a giant sideways step. 
>Rather than calling it better, I always describe it as "different".
>
>I'm not a big fan of the swords and stuff, but if I ever see anyone mention 
>customisable AI, I'll be buying it :) The default AI is *slightly* better 
>than 3.1 TA, it gets some units out before construction, and it actually 
>handles it's Commander/Monarch well. Stunning :)

That is because in TA:K, the monarch/ai in general has LESS to worry/think about than in TA. With fewer build possibilities, and simplified build trees, there's LESS to screw up -- thus it does better. My ai mods work on exactly that same principle.

>About the Commander. Believe me, it's changed aright. I'm very happy 
>about it :) Before, you sent your 3 flash tanks at the AI Commander, and 
>he'd flip from one to the next without actually doing anything, now they 
>start attacking him, and he casually flicks around and D-Guns the little 
>devils :) Unfortunately he still has that damn 15 second pause when shot 
>bug. Oh, and don't get me started on what you call the "Coffee Break Bug", 
>I call it a "Pain In The Ass" :). May the TA AI/Joke department die 
>painfully.

That list, at least in a partial one, exists in my TA AI Guide FAQ. You can still download it at:
http://www.tamec.org/Utils/AI/ai_guide.zip

www.tamec.org has my ai's under resources, ai's as well...
I thought you knew this?

>Good thing you don't write a list, the e-mail would be too big for a humble 
>modem to download.
>
>Sigh, what exactly *did* they patch, considering almost every major bug 
>still remains. Makes you worry about the original release if we're now in 
>version 3 and there are still more bugs than can be named around.
>
>I guess QA is a myth to them.

QA consists of you tell us what's wrong in vX.X and we tell you a week later what we already have fixed in vX.X+1 and make you feel like we fixed it because we listened to you instead of using our own in-house debuggers.

David
--------
Subject: Re: Bugfix 1.1
Date: Wed, 4 Aug 1999 18:31:22 +0100

Well I was lucky, there was that wonderful list of CC units and their codes :) 
Even if that hadn't have been there I would have used the HPI viewer, mind you, the old versions did allow you to access the CC files. 
My AI currently only has a standard default profile, it is currently designed to deal with every map without actually specialising at any of them. I'd use an AI friendly map, Painted Desert is open, so no factory blocking (my AI cannot handle having any of it's factories blocked, but I'm asking BSR for his modified factories with a couple of changes). Greenhaven isn't bad, Yerrot Mountains is also good, as is Moon Quartet. Maps like Ashap Plateau (my favourite), are terrible as the metal is concentrated in certain areas, and these areas are not near the start positions. Probably the best AI map would be totally flat with metal concentrated around it's starting spaces. This means it's defences can cover it's metal, while it can make use of it's all seeing power to use missile troops with no hills to block them.. 
Actually my first few AI changes were just so I could create an AI that could hold up to my rushes on the 4th minute. However as I started to enjoy it more, I decided to carry on with the AI and allow it to handle an entire game. Version 3 was probably the biggest leap, it could handle the early game, but also play a long game too. Version 5 was a huge step backwards, but I'm trying to block out the memories :) 
I personally consider the BAI to be a rather weak AI. Rather than using good weightings and such, it relies heavily on having the cheats to allow it to stomp all over it's opponent. I personally believe the troop mixes could be vastly improved, and if it didn't cheat would probably be steam rolled. I haven't looked at the Counter Strike weights. 
Your AI Tweaking Guide? Do you mean the AI guide on TAEMC? If so, remember, that's what started this conversation in the first place :) I had already done most of what the guide said, though it did contain some points which I had never considered. Then I watched Mostly Harmless in action, and with that made some big changes to my AI. Before, I had considered a ratio of twice as much metal being used as being produced, reasonable, then Mostly Harmless showed this need not be the case, and so I have produced a better resource system for my AI. 
MH also showed me (through it's weightings), that resources could be weighted higher than plants, and although this may seem odd, the idea had never struck me before. 

>Subject: Re: Bugfix 1.1
>Date: Tue, 3 Aug 1999 21:01:48 +0100
>
>I wish I could remember or download 2.0b1 again. I think I'll look for it 
>on an old CD or something.

It's at Cavedog's website (under their FTP site), I think you can just modify the download addy from the v3.1 patch (ta1x-31c.exe) to the v2.0b1 (ta1x-2b1.exe) but I could have the name slightly wrong.

>Well since I've a lot of free time on my hands, I added CC units over the 
>last 3 days. God, I thought typing in codes was boring...........I was 
>right :)

At least you KNOW the codes! I on the other hand had to compile a fucking LIST as I went!

>CC doesn't seem that complex, there's the added pain of introducing extra 
>plants, without crippling the AI's resources. But generally, the 
>Hovercraft are just weaker vehicles, which are faster, have longer range 
>and slightly more powerful guns. Never mind, I'll get it.

CC does a lot of really cool things that are more like nuances than simplistic methods. The ai is no good with "finer details" and chokes on anything complex, thus for most TA:CC units the limit is zero.

>Do you think they'd been having a heavy drinking session when they made the 
>3.1 AI, and it was all just a big joke? After all, in just 10min of 
>fiddling.....I'd improved it. It truly was terrible.

Your first major improvement over the TA:CC ai will probably be the biggest improvement, each progressive improvement will probably be less and less better.

>Anyway, I've got to test my AI against yours tomorrow :) If it wins, I'll 
>have a nicely padded ego :) Anyway, the default profile for TA is 
>virtually finished, and I've set my testers on the CC version now.

Testing my ai out of context isn't very good though. My ai's are all for "standardized" maps -- testing my default on Pincoushin for instance against your specially tweaked ai specifically for that map is definitely going to put mine at a disadvantage.

>I'm sure I'll e-mail you again, brace yourself :)

If you haven't read my Ai Tweaking Guide, you better brace yourself! I wrote it before BAI or Counterstrike ai came out -- which are the 2 most powerful ai's I know of released to date. Still, without direct knowledge of hacked ai's, my conclusion for what makes a good ai should floor you...

David
--------
>I just noticed that you said AI "engine". Do you just mean the weights, or some of the programming too, such as the Commanders behaviour and such?
>By the way, in a previous e-mail, I said it had taken me just a couple of days to add CC units, by that I only meant the codes, I haven't gone and created an entire AI in only 3 days! :)

When I say AI "engine" I mean the actual programming jump IF-THEN codes the computer uses to decide the ai's actions -- everything from what it decides to build, where it decides to build it, how it attacks, how it moves its units, etc. All an ai profile can change is what it decides to build -- and an ai profile STILL has to fight built-in prejudices (such as ones against geothermal plants and moho metal extractors on nonmetal maps).

Your ai profile must be in an organized manner or it'll be hell for anyone but yourself to edit -- and it may even prove hard for you. The reason why I organized mine the way I did was I determined that the game didn't care if a unit was original,in the v3.1 patch, in TA:CC, or some no-named 3rd party addon -- if it was there, it would try to build it unless told explicitly (with a limit unitabbriv 0 line) not to. A rare exception to this is the Arm 3rd party unit Dreadnought hovercraft which is built via a gui rather than a tdf build menu -- the ai would NOT build it originally.

>However, in the single game of TAK I played (yes only 1 so far :)), I found the AI monarch handled combat better (even if at the start of the game he ignored the Mana point next to him and tried to build some unit production thing, that screwed him). In combat he would use his powerful spell close up (Taros I think he was from), then he would run away hitting my troops from afar. When his mana recharged he charged in, killing more people. I tell you, after getting used to beating an AI Commander with 3 flash tanks, losing about 25 close combat troops came as a bit of a shock.

That's because TA:K's combat and resource structure is vastly simplified. (Units couldn't even fire on the move before the first patch!) If you simplify the ai, it appears to get smarter because there's fewer dumb mistakes it can make -- but as you pointed out, it still made some incredibly dumb ones: ignoring early available resources and over-taxing its poor resources. The ai's monarch seems smarter also because Cavedog figured out how to turn its stimulus responses into sensible IF-Then type behaviors -- something which seems a bit lacking in TA because the ai's comm is already trying to do many other overriding things.

>Don't worry, I already have your AI guide, and in fact I have everything by you I could find on TAMEC. The AI guide is what started this conversation in the first place remember :)
>I'm prepping my AI to mess with Mostly Harmless, I'm just trying to get it near consistently to build a factory early, the problem is when I do get it to, it cripples itself by building another almost straight away. And using the current weights it plays very well, but not quite well enough to stop a quick raid.

If you consider what my ai makes a "quick raid", you'll find any rusher-style players more than capable of trashing your ai. That is what prompted the developement of Counterstrike ai -- I'd gotten adept at taking apart a nonhacked ai, even to the point of using its own weapons and resources against it.

>TA fixed things in their new versions? Wow, I thought they just created new ones and added some nice shortcut keys (that obviously takes them a couple of months to do :)).

I remember noticing some fixes in the newer versions that DIDN'T make it in the readme's. Cavedog doesn't WANT the extent of its bugs in TA to become public knowledge -- it would sabotage too much credibility. I'd just prefer they fixed the bugs, quietly or no. Since they didn't (or hadn't last I looked), I put up quite a fuss for a while - and for now have just given up.

>And hey, what's this, TAK where they're doing it all over again!

Except it's a different set of people now - many of the old hats that worked on the original TA are gone, and the newbies are just copying the older work -- not realizing the problems they will have by not understanding what they're doing.

>Well I was lucky, there was that wonderful list of CC units and their codes :) 
>Even if that hadn't have been there I would have used the HPI viewer, mind you, the old versions did allow you to access the CC files. 

The NEWEST version of HPI viewer at TAMEC (or Joe D's site) should allow almost flawless access to CCX files. Are you telling me it doesn't?!?!

>My AI currently only has a standard default profile, it is currently designed to deal with every map without actually specialising at any of them. I'd use an AI friendly map, Painted Desert is open, so no factory blocking (my AI cannot handle having any of it's factories blocked, but I'm asking BSR for his modified factories with a couple of changes).

BSR is lazy - as am I - understand what he did then do it yourself. It's a yardmap trick, and it results in factories that are "hollow" as far as shots are concerned too! It sucked bad enough that BAI factories built faster, that they were harder to hit (and thus kill) was a real crippler. Building an ai AROUND hacked units results in an ai that simply won't work at all when these hacked units aren't available -- case in point BAI. Avoid it if you can, because it'll draw you some flack if you don't.

>Greenhaven isn't bad, Yerrot Mountains is also good, as is Moon Quartet. Maps like Ashap Plateau (my favourite), are terrible as the metal is concentrated in certain areas, and these areas are not near the start positions. Probably the best AI map would be totally flat with metal concentrated around it's starting spaces. This means it's defences can cover it's metal, while it can make use of it's all seeing power to use missile troops with no hills to block them.

Probably true, but incredibly boring (at least to me) as far as gameplay is concerned.

>Actually my first few AI changes were just so I could create an AI that could hold up to my rushes on the 4th minute.

I don't think any of my ai's can do that except for the most cheesy earlier version of CounterStrike AI. I have to carefully define what rushing IS before I can tell you how effective my ai is AGAINST rushing attacks. Rushing is getting an early strong attack in on the opponent to KILL him before he can properly counter it. Rushing at the 4th minute can mean anything though -- I ussually define an ai rush "launch" as the moment the ai commits all of its armed units to their first attack, in other words the moment they LEAVE its base, as opposed to when that attack first arrives at the enemy base -- which is extremely variable because of random distances, map qualities, and what the opponent is doing in the meantime. On a nonmetal land map in a 1k/1k game, any nonhacked ai that can launch its FIRST attack from its base before the 10 minute mark is a good start on a good ai. The BEST I've seen on a nonmetal map 1k/1k start is an ai that launched its first attack at just under 4 minutes. For comparision, I have seen CounterStrike ai launch its first attack in under 2 minutes -- which is earlier than BAI even on hard and hits harder. It's only when BAI builds up for a while that it can withstand the onslaught of CounterStrike ai.


>However as I started to enjoy it more, I decided to carry on with the AI and allow it to handle an entire game. Version 3 was probably the biggest leap, it could handle the early game, but also play a long game too. Version 5 was a huge step backwards, but I'm trying to block out the memories :)

I had about a 6 month period where nothing I did for the ai made it any better than my earlier versions and tended to make it much worse. I feel for generalized maps, my ai is about as good as I can get it till the late-game (after 45 minutes, when the fusions start appearing.)

>I personally consider the BAI to be a rather weak AI. Rather than using good weightings and such, it relies heavily on having the cheats to allow it to stomp all over it's opponent. I personally believe the troop mixes could be vastly improved, and if it didn't cheat would probably be steam rolled. I haven't looked at the Counter Strike weights.

Funny thing is, BSR said it was tweaked DOWN to a level where it could barely be beaten. The released BAI was version 3, versions 1 and 2 were much much harder. I will not deny that such could be true -- his ai-only units could EASILY have produced more resources and units at a faster pace. My friend Meph made a hacked c-air that could build almost anything you could want at a rate 100 times faster than a commander can build, is invulnerable to almost all forms of attack, is faster than a hawk, makes more resources than even its incredible buildrate uses, and is buildable by all the level 1 factories. Oh, did I mention that it cloaks? :p He used it to quickly test ai's and new units he was designing -- otherwise, it took too long to test modified krogoths. Now he can build one in about a minute from the moment the game starts. Although later versions were NOT buildable by the ai, it was meant as a near-ultimate player cheat, I rigged up an ai that could build it to see what would happen. The ai hit 500 unit max before the 5 minute mark, and there were so many of these hack c-airs that they made targetting anything on the ground DIFFICULT -- I'd forgotten to limit them to 1 or 2. (Have you seen the movie Andromeda Strain? -- that's basically what happened. :p )

"I personally consider the BAI to be a rather weak AI." <-- One of the few things you can write that makes me very angry, but probably not for the reason you imagine. Shortly after BAI's release, BSR asked for open feedback on BAI to be posted on Cavedog's message board. So I wrote that it was my opinion that BAI was rather lame and stated my reasons. I called it that to see if he could take harsh criticism (but I didn't tell him this) and because I felt the same as you about BAI -- just more strongly. BAI takes a no holds barred approach to hacked ai-only units, not only are they ludicriously more powerful that anything the player gets - in the case of the construction units and factories - they don't even work if you capture them for yourself! This is because of a lack of player-accessable build menus. BAI wastes around 80% of all the resources it makes (this is mainly because it lacks weights and limits on some of its units, particularly its hacked resources) -- I didn't see the point of giving the ai extreme amounts of resources if it couldn't find a way to use them. BAI is also very poor at dealing with even a 500 unit max limit and often ends up hitting the limit before it gets a diversified base and attack force. My final conclusion was BAI could certainly kill all but the very best players (also hinting that you had to use game cheats or bugs in the ai to beat it) but ONLY because it haphazardly cheats its way to victory.

BSR's response came as an almost complete shock to me, despite the fact I was critical of his BAI. He accused me of slandering him for no reason, also that I was stealing his ideas, and that I was totally lacking any creativity of my own. I also caught a few other negative posts because of my poorly-stated opinion of BAI -- esentially I was both laughed and insulted off the Cavedog message board. Not long after that, I never posted there again.

The ideas I was supposedly stealing from him was how BAI used hacked units to be a super-powerful ai -- via hacked ai-only units, particularly resource-providers and increased buildtime c-units and factories. Since I hadn't released an ai that almost exclusively used hacked units to provide resources (but said that I knew how) -- his only evidence was my own posts where after hearing a bit about a pre-release BAI I said that I had made a hacked Arm vehicle plant only the ai could build that provided its own resources, and had a 10x buildspeed over the original. If an ai was designed with that particular unit in mind, it could be a VERY tough ai to beat as my unsuspecting friends playing on a LAN versis some ai's found out! (That plant later became part of Counterstrike ai, albet with a much-reduced buildrate (2x regular) and with less resource production than it needed to run at full speed.) At the time I mentioned the hacked ai-only factory, I did not know for certain how BAI was designed -- just that it was (supposedly) nearly unbeatable.

CounterStrike ai is my counter to the accusations that I was copying BSR's BAI. Developement of it started about the same time as BAI, but almost everything similar to parts of BAI I made from earlier sources (mentioned later;) and are done for different reasons. BAI sustains its base production with massive resource-producing hacked units that it builds more and more of throught the game. Counterstrike builds 2 special resource storage buildings that provide a VERY good starting income, (+60 metal/sec and +1200 energy/sec -- the fluxors!) and builds normal fusions/hacked cloaking fusions/normal moho metal makers to sustain its later production. Often, the fluxors end up providing less than half of its income by the 40 minute mark -- meaning that killing the fluxors will only slow it down (and may not even do that if because of the overbuild bug it decides to build 10 more using its many adv. c-units!) It uses no hacked c-units, only 2 hacked factories (of which neither is neccessary, but the hacked krogoth gantry is cool...), only 2 hacked resources per side which have very strict limits. There's a limit of 2 on fluxors and 0-4 cloaking fusions -- a fluxor is roughly equal to 2 ai-only solars and 1.5 ai-only metal extractors with the BAI on hard -- and BAI doesn't stop with 4 solars and 3 metals either! The cloaking fusions make twice the energy of regular cloaking fusions, produce +10 metal, and cloak for only -1 energy all for the same cost as the originals. (But they must pay the cloaking cost even while being built -- giving the appearance of adv. c-units building nothing :)

MadRat, Creeperman, and Speed are the 3 guys behind the REAL cool ideas in Counterstrike. MadRat came up with the hacked ai-only bombers that use much-increased LOS and better badtargetting lines to find targets from outside missile tower range. (Up till that point, we all thought the ai always played as though it had perm LOS!) I took that ability one step further and applied it to all armed aircraft normally found in the game. Speed created the first known (to me!) ai-only unit, the ai nuker -- a nuke silo that the ai can build that works. Creeperman came up with the ai-only jamming units that use a 0-damage weapon so they chase the attack force around. (Now the ai's attack force doesn't always show up on radar...) He also helped me make the ai-only nuke silos and bombers even better. I came up with the ai-only EMP silo and Neutron silo -- a simple offshoot of the ai-only nuke silos. If nothing else, the ai-only EMP silo has an annoying habit of draining player's antinukes in multiplayer vs ai games. Lastly, I came up with ai-only cloaking fusions because the ai NEVER cloaked the original cloaking fusions but did landmines -- however I found out that unless I changed the cloakcost, the ai would have to pay -450 energy/sec (-550 for core's!) even while the cloaking fusion was being built which could take ~10 minutes of gametime thus using up about 250,000 energy to keep it cloaked before it was finished. Last but not least, I made a cloaking ai-only Arm Shooter (that cloaks for -1 energy while still and being built and -200 energy while moving) that is easily among the most annoying things the ai can do to a player -- big blue lasers shooting from out of nowhere. :p

Strangely missing is an ai-only antinuke silo. This is because neither Speed, BSR, nor I could get one that would work -- it's a limitation of the code that defines how antinukes work.

Hacked units to add abilities the ai should have but doesn't (cloaking fusions, bombers that actually drop bombs occassionally, nuke silos that nuke, jammers that don't act like buildings, etc.) is good but results in an ai that requires ALL players to have all the hacked units and ai to work. An ai that uses Ai profiles only can work even if other players DON'T have it -- which fortunately for me allowed me to test ai vs ai games via the internet and on LAN's.

>Your AI Tweaking Guide? Do you mean the AI guide on TAEMC? If so, remember, that's what started this conversation in the first place :) I had already done most of what the guide said, though it did contain some points which I had never considered. Then I watched Mostly Harmless in action, and with that made some big changes to my AI. Before, I had considered a ratio of twice as much metal being used as being produced, reasonable, then Mostly Harmless showed this need not be the case, and so I have produced a better resource system for my AI.

Ideally, the ai's resource drain should start out much greater than its production and by the end of the game it should have almost the reverse -- its resource drain should be less, but not MUCH less than its production. It's ok if the ai wastes much of its energy production later in the game, but it should probably be more or less equal on metal useage/drain -- or any player worth their salt will be able to outproduce the ai so badly it isn't even funny.

>MH also showed me (through it's weightings), that resources could be weighted higher than plants, and although this may seem odd, the idea had never struck me before.

Even in a 5-minute game it makes sense for the ai to have at least SLIGHTLY more solars and metal extractors than factories. The ai profile numbers don't really correspond to ideal interpretations, because the ai game engine even "interrupts" and overrides the ai profiles at times (case in point is how many moho metal makers it will build even when weighted extremely low and limited to 1.) The numbers in my ai profiles aren't always "sensible values" though, they are often what I had to plug in to get the desired results out. I finally settled on a weight range of 0.05 to 1 where the values could be roughly computed as percentage build chance -- with 1 = 100% (build any chance you get) and 0.05 = 5% (rarely build this.) If you use BIG weights and little weights like 9 and 0.1, even a small weight like 2 would get built more than the ai profile says it should. In Mostly Harmless, I tried to force the weights down to as tiny a value as possible - to minimize the chances of late-game items being built early and to maximize the must-build-now items.

Even if you only tested the ai on one map, say Painted Desert for instance, you could still make over 10 different ai's to deal with VASTLY different conditions on that map:
An ai geared for 1k/1k starting resources is vastly different than one geared for 10k/10k. The 10k/10k one should attempt to build a lot of early factories and c-units to quickly convert the metal and energy into something USEFUL, the 1k/1k one should build more early resources (higher weights on resources, lower limits on factories) so it has enough metal and energy to make something USEFUL.
An ai geared to rush early and one that's geared to survive an early rush is also vastly different -- which is also VASTLY different than one that tries to get lots of resources quickly. An early rush ai (Rusher) makes minimal defenses and spends most of its resources making attack units. The defensive ai (Porcupine) would concentrate more on defenses - neglecting both attack units and resources some as a result of it. And the resource-grabbing ai (Octopus) would have practically no defenses (because that was time "better" spent making more resources) and fewer attack units than a Rusher ai -- at least early on.
An ai that expects a unit limit of 200 (as defined in the missions or pre-v2.0b1 patches) or 250 will almost always have very low limits on everything it can build - and generally has to eliminate many units the ai can use only when it already has a big base/attack force already built (like the Arm Annihilator or the Arm Shooter.) On the other hand, an ai designed for a 500 unit limit should attempt to build bases that look like cities and armies that look like ants on the move. :)
The Arm ai and Core ai should be thought of as 2 separate ai's (they are, btw!) with Arm using lots of level 1 attack units and few base defenses, and Core using lots of base defenses and trying (hoping) to build level 2 units (and krogoths!) to counter Arm's strategy.
It could be that you want a generalist ai that builds a little of everything (best for high starting resources on a metal-rich map with few constrictions) or a specialist ai that builds far fewer types of units but builds a shitload of them (the extreme of that would be a flash-rushing ai.) It's almost as a rule specialist ai's designed for a single map and under known starting conditions will beat generalist ai's, all other things being held equal.
Last but not least, there are 3 different skill settings for which an ai can be designed for - easy, medium, and hard. Cavedog's ai's treats each as a totally different ai - and as such on easy the ai might play more defensively and do even poorer on resources, on medium it might rush but lack base defenses, on hard it could both rush and build more resources and build more base defenses. Maybe it goes for more aircraft on hard too? (That's called an Eagle strategy -- strong air, weaker ground. It's probably the hardest player strategy to defeat in the late-game.)

You have to KNOW what you want your ai to do, and you have to recognize what kind of opponents it's likely to be up against. What's more, you have to do some little things to make it better at fighting its counter even though it might weaken it against what it's BEST at defeating.

Take my Mostly Harmless Default.txt Arm ai, here's what I expected:

(from the map/starting settings)
1k/1k starting resources, (it tries to build lots of early resources and few defenses assuming it cannot afford them)
250 or greater unit max, (it has limits too low for 500 unit max, but too high for anything less than 150 unitmax)
mostly open all-land maps, (it ends up spreading out a lot)
starting locations that are either defended on some sides, or in a corner (thus the lower number of defenses should prove enough for a single attack front)
and ample metal spots. (it builds LOTS of metal extractors)

(from the other player)
The ai "expects" only one opponent - because its attack force always heads towards its nearest opponent, and its base defenses (if there ARE any!) aren't adequate against attacks from more than one side.
Also, attacking the ai from multiple sides with each attack having specific goals in mind (one to blunt its assault force, another to kill all its merls from behind, some to destroy wandering c-units and metal extractors, and a major attack to kill fusions and factories, etc) will generally make the ai look like a joke.
The ai doesn't handle a player who raids resources instead of trying to rush into the center of the ai's base and kill its commander. (Ussually, the greatest concentration of base buildings and defenses is near the ai's commander.)
That the opponent's goal gives the ai at least 5-10 minutes of time in the beginning of the game where it doesn't have to contend with getting its pathetic base destroyed. 
That the opponent doesn't rig up an "impenetrable" wall (actually quite hard to do on true LOS, unless you are also destroying the ai's base) and just use the metal from dead ai units to run more and more of their base, or worse yet build an undefended dt wall to hedge out the ai's ground forces entirely. (Even if the ai cannot get to you by ground units, it is too dumb to quit building them.)
That the opponent doesn't build a 100+ aircraft swarm and swamp the AA defenses of the ai and kill or cripple it quickly. (This is almost impossible for players to counter, especially over the internet with its lag, so the ai has even MORE trouble.)
The ai cannot handle a concentrated attack on a specific target, such as a fusion -- it will NOT, as players might, concentrate in a manner to stop the attack. Heck, it might act like it's not being attacked at all. It tries to attack the NEAREST target to the leading edge of its attack force, regardless of what or where it is -- that is ALL it knows as far as mobile defense is concerned.

(from itself)
It tries to expand and cover all nearby metal spots with metal extractors - all while building at least half as many solars as metal extractors.
It builds few factories, and instead relies on hordes of c-units to give it a rapid base expansion.
It uses missile towers as its primary form of base defense, both from light air attacks and ground assaults.
Its attack force is designed mainly as an anti-base weapon, but should do ok against a similar attack force in the field.
The early bertha gets the worm, thus it tries to get level 3 units earlier than may be wise.
If left alone for an extremely long amount of time, (like 4 hours) it might build enough aircraft factories to field considerable numbers of aircraft and replace them in 10 minute intervals -- thus resulting in a large swarm of aircraft hitting your base about every 10-20 minutes.
If you attack its base but fail to overrun it, any c-units remaining in the area (particularly the ai commander, ai decoy commanders, and ai c-airs) will probably reclaim the metal and use it against you.
Its attack force is meant to be (but may not succeed in being) something the player cannot ignore, no matter HOW strong their wall/s are.
It cannot be counted on to do anything as early as possible, but should manage to "get by" better than the average TA player found on boneyards.

Because the ai spends so much time having its c-units just wandering around instead of building (I don't know about you, but I build groups of solars almost as quickly as buildtimes allow by moving little), someone who moves their c-units less per build order, uses a shorter route to metal spots for c-units, builds concentrated defenses only in a very few key areas, gets a fusion/moho metal maker combo going sooner (or geothermal/moho metal extractor combo on metal maps or light carriers/moho metal maker combo on water/land maps), makes a massive attack force from the resource advantage over the ai can defeat the ai with little more than CTRL+W (select all mobile attack units) and 1 left mouse button click on the minimap telling their units to move them to the far side of ai's base (with +SHOOTALL on to kill resources as well).
Resources are more important than kill ratios - you might have 10 losses per every enemy you kill (if you choose your targets wisely!) but because you constantly increased production at 1.5 times your opponent from the very start till the finish you can afford such losses. Losing 20+ aircraft but destroying 1 fusion or using 10 flash tanks to kill 1 commander (you don't get credit for units killed by his explosion) is almost always a good deal.

So, as you might expect, I get the occasional email message stating that although my ai is ok or even good at hitting players with an early attack, it seems totally lacking in the late-game -- and much of its base can be destroyed with attacks of 10-bombers at a time. CounterStrike on the other hand ussualy gets emails about cheating too much, sometimes even in ways that it either doesn't cheat (such as being able to build aircraft faster out of its air plants than players) or cheating in ways that all ai's can do -- such as acting like it has perm LOS. I have learned to find such emails not as criticism so much as a source of humor, because I know many limitations to TA that prevent me from making an ai like I WANT to make. BTW, give a try on the modified krogoth campaign and the extra campaign I made for TA in my v1.1 bugfix -- I'm sure you'll find the ai at the least VERY annoying. :) Campaign missions allow you to tell the ai EXACTLY how to use its units, but the ai will only build original-game v1.0 units unless explicitly ordered.

DON'T FORGET THE OVERBUILD BUG! My ai profiles have what appear to be too-low limits because I *KNOW* the ai will more often than not build far more than the limits I used. Had I put in the ACTUAL numbers I wanted, the ai would just produce that many and more!

Lots of construction aircraft tend to allow the ai to build metal extractors quickly, but the ai cannot build them as fast as it can c-vehicles. I use c-air's for the ai's "explosive expansion" phase starting roughly 10 minutes in the game and lasting up to 30 minutes. The ai ussually hits its metal extractor limit before the 40 minute mark, if you're haven't been killing its c-units and metal extractors. And 40 metal extractors making +1.0/sec metal each adds up to a LOT of metal produced for very little metal and energy investment (less than the cost of 3 moho metal extractors or 1 fusion plant.)

If you've ever played Magic the Gathering the card game, designing an ai for TA is very similar to how to design a GOOD magic deck. Almost anything the ai can build is benificial, but some things are more benificial than others. And one thing's for sure - an ai cannot be made to do everything and do it quickly unless it cheats like a dog. (BAI and Counterstrike would be like a magic deck that gets twice as much mana or more per land and draws 2 or more cards per turn.) A small deck of 1-3 mana-cost green creatures can easily be made, but will have poor long-game surviveability. (IE: an Arm Flash-rushing ai that can handle a max unit limit even as low as 100 because flash-rushing is about all it does.) A blue "denial" deck with lots of powerful flying creatures has a very slow start - it is also very hard to play because of the difficulty in knowing what spells to counter. (IE: a Core Vamp-swarming base-from-hell ai -- the ai is poor at concentrating its air attacks into one big punch, or even getting lots of aircraft to begin with.) EVERYTHING has a counter, but some things are cost prohibitive to counter such as an Arm Pelicans swarm in the open ocean (mostly because of the Pelican bug.)

This ended up being longer than I intended.
David
--------
Subject: Re: Huge TA ai message
Date: Thu, 5 Aug 1999 22:09:24 +0100

Woah!  A long message, I like them, it means detail (or long drivelling :)).

> >Subject: Re: Bugfix 1.1
> >
> >I just noticed that you said AI "engine". Do you just mean the weights,
or
> >some of the programming too, such as the Commanders behaviour and such?
> >By the way, in a previous e-mail, I said it had taken me just a couple of
> >days to add CC units, by that I only meant the codes, I haven't gone and
> >created an entire AI in only 3 days! :)
>
> When I say AI "engine" I mean the actual programming jump IF-THEN codes
the
> computer uses to decide the ai's actions -- everything from what it
decides
> to build, where it decides to build it, how it attacks, how it moves its
> units, etc. All an ai profile can change is what it decides to build --
and
> an ai profile STILL has to fight built-in prejudices (such as ones against
> geothermal plants and moho metal extractors on nonmetal maps).
>
You gotta hate them, I wish it would be smarter than the current "have less
than 100 metal/energy, don't build, even if you are generating more than you
would lose by building something"

> Your ai profile must be in an organized manner or it'll be hell for anyone
> but yourself to edit -- and it may even prove hard for you. The reason why
I
> organized mine the way I did was I determined that the game didn't care if
a
> unit was original,in the v3.1 patch, in TA:CC, or some no-named 3rd party
> addon -- if it was there, it would try to build it unless told explicitly
> (with a limit unitabbriv 0 line) not to. A rare exception to this is the
Arm
> 3rd party unit Dreadnought hovercraft which is built via a gui rather than
 a
> tdf build menu -- the ai would NOT build it originally.
>
It *was* organised nicely; with labels and all, but in the end I wasn't sure if comments would slow down the AI, and as it goes from top to bottom, I assumed it would, so I removed them.  Anyway, I have speant so long staring at the profiles, just trying to get that plant up a second quicker (know the feeling? :)), I can find stuff just by looking at the scroll bar :)  It isn't desgined to be edited by others, and that suits me fine.

> >However, in the single game of TAK I played (yes only 1 so far :)), I
found
> >the AI monarch handled combat better (even if at the start of the game he
> >ignored the Mana point next to him and tried to build some unit
production
> >thing, that screwed him). In combat he would use his powerful spell close
> >up (Taros I think he was from), then he would run away hitting my troops
> >from afar. When his mana recharged he charged in, killing more people. I
> >tell you, after getting used to beating an AI Commander with 3 flash
tanks,
> >losing about 25 close combat troops came as a bit of a shock.
>
> That's because TA:K's combat and resource structure is vastly simplified.
> (Units couldn't even fire on the move before the first patch!) If you
> simplify the ai, it appears to get smarter because there's fewer dumb
> mistakes it can make -- but as you pointed out, it still made some
> incredibly dumb ones: ignoring early available resources and over-taxing
its
> poor resources. The ai's monarch seems smarter also because Cavedog
figured
> out how to turn its stimulus responses into sensible IF-Then type
behaviors
> -- something which seems a bit lacking in TA because the ai's comm is
> already trying to do many other overriding things.
>
Some point threat system would have been nice.  I'm being shot by 1 samson, that's worth 3 points, but we need resources, that's 2 points, however we are also under attack, so we need resources, so add an extra 2 to the second score, and that makes 4. Build a panel!  OK, it's quite hard, but the AI's thick, almost anything would have been an improvement.  I think QA and the programmers fell out and didn't talk to one another for the duration of development.

> >Don't worry, I already have your AI guide, and in fact I have everything
by
> >you I could find on TAMEC. The AI guide is what started this conversation
> >in the first place remember :)
> >I'm prepping my AI to mess with Mostly Harmless, I'm just trying to get
it
> >near consistently to build a factory early, the problem is when I do get
it
> >to, it cripples itself by building another almost straight away. And
using
> >the current weights it plays very well, but not quite well enough to stop
a
> >quick raid.
>
> If you consider what my ai makes a "quick raid", you'll find any
> rusher-style players more than capable of trashing your ai. That is what
> prompted the developement of Counterstrike ai -- I'd gotten adept at
taking
> apart a nonhacked ai, even to the point of using its own weapons and
> resources against it.

Sorry, should have defiened that.  A "Quick raid" in my book, is a raid that is made at about the first possible moment, generally the attack force would be ready at about the 4 minute mark.  That's why my AI currently handles the larger maps better, a raid cannot reach it until it has it's own force, but on small maps, it isn't ready at 4 minutes, it currently takes 5 to hold against a raid.

> >TA fixed things in their new versions? Wow, I thought they just created new
> >ones and added some nice shortcut keys (that obviously takes them a couple
> >of months to do :)).
>
> I remember noticing some fixes in the newer versions that DIDN'T make it in
> the readme's. Cavedog doesn't WANT the extent of its bugs in TA to become
> public knowledge -- it would sabotage too much credibility. I'd just prefer
> they fixed the bugs, quietly or no. Since they didn't (or hadn't last I
> looked), I put up quite a fuss for a while - and for now have just given up.

I think they're trying to buy people off with a new unit, except people don't like that either :)

> >And hey, what's this, TAK where they're doing it all over again!
>
> Except it's a different set of people now - many of the old hats that worked
> on the original TA are gone, and the newbies are just copying the older work
> -- not realizing the problems they will have by not understanding what
> they're doing.

Sigh.

> >Well I was lucky, there was that wonderful list of CC units and their codes
> >:) Even if that hadn't have been there I would have used the HPI viewer,
> >mind you, the old versions did allow you to access the CC files.
>
> The NEWEST version of HPI viewer at TAMEC (or Joe D's site) should allow
> almost flawless access to CCX files. Are you telling me it doesn't?!?!

Ah, no.  I had v1.2, I didn't realise that this was out dated, I assumed the site I got it from had an up to date copy.  That word did, should have been didn't.  Whoops!  The latest version, 1.9 works fine with ccx.

> >My AI currently only has a standard default profile, it is currently
> >designed to deal with every map without actually specialising at any of
> >them. I'd use an AI friendly map, Painted Desert is open, so no factory
> >blocking (my AI cannot handle having any of it's factories blocked, but I'm
> >asking BSR for his modified factories with a couple of changes).
>
> BSR is lazy - as am I - understand what he did then do it yourself. It's a
> yardmap trick, and it results in factories that are "hollow" as far as shots
> are concerned too! It sucked bad enough that BAI factories built faster,
> that they were harder to hit (and thus kill) was a real crippler. Building
> an ai AROUND hacked units results in an ai that simply won't work at all
> when these hacked units aren't available -- case in point BAI. Avoid it if
> you can, because it'll draw you some flack if you don't.

However I am doing BSR a favour, and so this is in return.  This AI is avoiding anything that cheats, I hate cheating AI's (no offence to Counter Strike).  I prefer to try and beat the flaws.  Sure it attacks in a straight line, at one point.  Ouch!  So I just try to make sure it has a force that nothing in that straight line is going to beat.  Also, a few fast units because just by sending a flea succring around near some goliaths, you can distract them for the whole game, which is a God almighty pain.

The problem is, my AI is designed to work with a set number of plants, it cannot handle one of those plants being positioned badly, but if I increase the limit, it cripples it's resources at level 2.

> >Greenhaven isn't bad, Yerrot Mountains is also good, as is Moon Quartet.
> >Maps like Ashap Plateau (my favourite), are terrible as the metal is
> >concentrated in certain areas, and these areas are not near the start
> >positions. Probably the best AI map would be totally flat with metal
> >concentrated around it's starting spaces. This means it's defences can
> >cover it's metal, while it can make use of it's all seeing power to use
> >missile troops with no hills to block them.
>
> Probably true, but incredibly boring (at least to me) as far as gameplay is concerned.

Very true.  However, as you said in your guide, an interesting AI and good AI are not the same thing.

> >Actually my first few AI changes were just so I could create an AI that
> >could hold up to my rushes on the 4th minute.
>
> I don't think any of my ai's can do that except for the most cheesy earlier
> version of CounterStrike AI. I have to carefully define what rushing IS
> before I can tell you how effective my ai is AGAINST rushing attacks.
> Rushing is getting an early strong attack in on the opponent to KILL him
> before he can properly counter it. Rushing at the 4th minute can mean
> anything though -- I ussually define an ai rush "launch" as the moment the
> ai commits all of its armed units to their first attack, in other words the
> moment they LEAVE its base, as opposed to when that attack first arrives at
> the enemy base -- which is extremely variable because of random distances,
> map qualities, and what the opponent is doing in the meantime. On a nonmetal
> land map in a 1k/1k game, any nonhacked ai that can launch its FIRST attack
> from its base before the 10 minute mark is a good start on a good ai. The
> BEST I've seen on a nonmetal map 1k/1k start is an ai that launched its
> first attack at just under 4 minutes. For comparision, I have seen
> CounterStrike ai launch its first attack in under 2 minutes -- which is
> earlier than BAI even on hard and hits harder. It's only when BAI builds up
> for a while that it can withstand the onslaught of CounterStrike ai.

Well currently it can't :)  It can hold a force that reaches it at the 5th minute (unless the player uses resource cheats of course), but not before that point.  The word probably should be "raiding".  Actually when I went online I was surprised by how long most player took to attack, including ARM.  I always have a 5 flash assault ready just before the 5 minute mark (assuming a greenhaven type map), where as some players take 7 minutes. Actually on one map, the lava one where all the metal is in the middle and narrow ramps lead there, my human opponents took so long to attack that I actually managed to hold them back by placing HLT's on the ramps.  How I laughed :)

My AI launches an attack between 5-7 minutes.  If it took more than 7 minutes I'd cry openly (version 5 alas.........).  I consider anything above 6 unacceptable as it generally means the AI wouldn't have been able to hold back a human who moved in swiftly.  However, if it takes as long as 7 it generally has a better resource base and attacks in greater strength.  it
makes it quite interesting, because in some games it will hit you fast, but lightly, anf other times will take longer, but hit you harder.  The whole AI started because I couldn't find one that could stop my raids on the 4th minute.  That was the original goal of Queller.  It's pretty much achieved it, now it plays "Offence is the best form of defence", basically it's
modelled around my playing style.

Counter Strike seems to be more of an offencive AI than BAI, which is more construction centred, and is pretty much a cheating BSRAI.

I would certinally be impressed by an AI that launched in under 4, Queller once scrapped 4 minutes, but only once.

> >However as I started to enjoy it more, I decided to carry on with the AI
> >and allow it to handle an entire game. Version 3 was probably the biggest 
> >leap, it could handle the early game, but also play a long game too.
> >Version 5 was a huge step backwards, but I'm trying to block out the
> >memories :)
>
> I had about a 6 month period where nothing I did for the ai made it any
> better than my earlier versions and tended to make it much worse. I feel for
> generalized maps, my ai is about as good as I can get it till the late-game
> (after 45 minutes, when the fusions start appearing.)
>
> >I personally consider the BAI to be a rather weak AI. Rather than using
> >good weightings and such, it relies heavily on having the cheats to allow
> >it to stomp all over it's opponent. I personally believe the troop mixes
> >could be vastly improved, and if it didn't cheat would probably be steam
> >rolled. I haven't looked at the Counter Strike weights.
>
> Funny thing is, BSR said it was tweaked DOWN to a level where it could
> barely be beaten. The released BAI was version 3, versions 1 and 2 were much
> much harder. I will not deny that such could be true -- his ai-only units
> could EASILY have produced more resources and units at a faster pace. My
> friend Meph made a hacked c-air that could build almost anything you could
> want at a rate 100 times faster than a commander can build, is invulnerable
> to almost all forms of attack, is faster than a hawk, makes more resources
> than even its incredible buildrate uses, and is buildable by all the level 1
> factories. Oh, did I mention that it cloaks? :p He used it to quickly test
> ai's and new units he was designing -- otherwise, it took too long to test
> modified krogoths. Now he can build one in about a minute from the moment
> the game starts. Although later versions were NOT buildable by the ai, it
> was meant as a near-ultimate player cheat, I rigged up an ai that could
> build it to see what would happen. The ai hit 500 unit max before the 5
> minute mark, and there were so many of these hack c-airs that they made
> targetting anything on the ground DIFFICULT -- I'd forgotten to limit them
> to 1 or 2. (Have you seen the movie Andromeda Strain? -- that's basically
> what happened. :p )

That's the problem with cheating AI's, and why I dislike them, there's no way to really compare them, and I don't like playing against opponents who can do things I cannot.

> "I personally consider the BAI to be a rather weak AI." <-- One of the few
> things you can write that makes me very angry, but probably not for the
> reason you imagine.

Angry at me?  Or just angry generally?  From what I read below, my point of view and yours seem to be the same, so I assume it is the latter.

> Shortly after BAI's release, BSR asked for open feedback
> on BAI to be posted on Cavedog's message board. So I wrote that it was my
> opinion that BAI was rather lame and stated my reasons. I called it that to
> see if he could take harsh criticism (but I didn't tell him this) and
> because I felt the same as you about BAI -- just more strongly. BAI takes a
> no holds barred approach to hacked ai-only units, not only are they
> ludicriously more powerful that anything the player gets - in the case of
> the construction units and factories - they don't even work if you capture
> them for yourself! This is because of a lack of player-accessable build
> menus. BAI wastes around 80% of all the resources it makes (this is mainly
> because it lacks weights and limits on some of its units, particularly its
> hacked resources) -- I didn't see the point of giving the ai extreme amounts
> of resources if it couldn't find a way to use them. BAI is also very poor at
> dealing with even a 500 unit max limit and often ends up hitting the limit
> before it gets a diversified base and attack force. My final conclusion was
> BAI could certainly kill all but the very best players (also hinting that
> you had to use game cheats or bugs in the ai to beat it) but ONLY because it
> haphazardly cheats its way to victory.

Like I said, I am not really a fan of *any* cheating AI, I find they generally only test your ability to pick the flaws in the AI.

For a lot of hacked AI's, the only way to beat them is to take advantage of the flaws.  In BAI, early on if it loses anything that produces resources, you can hurt it badly.  You can just sucker it's troops in one direction, then bring in some raiders from the other.

> BSR's response came as an almost complete shock to me, despite the fact I
> was critical of his BAI. He accused me of slandering him for no reason, also
> that I was stealing his ideas, and that I was totally lacking any creativity
> of my own. I also caught a few other negative posts because of my
> poorly-stated opinion of BAI -- esentially I was both laughed and insulted
> off the Cavedog message board. Not long after that, I never posted there again.

Have you ever visited the TAK forum? :)  It sounds like BAI zealots exist too.  I missed most of TA history, and only came back to it and skirmish recently.

> The ideas I was supposedly stealing from him was how BAI used hacked units
> to be a super-powerful ai -- via hacked ai-only units, particularly
> resource-providers and increased buildtime c-units and factories. Since I
> hadn't released an ai that almost exclusively used hacked units to provide
> resources (but said that I knew how) -- his only evidence was my own posts
> where after hearing a bit about a pre-release BAI I said that I had made a
> hacked Arm vehicle plant only the ai could build that provided its own
> resources, and had a 10x buildspeed over the original. If an ai was designed
> with that particular unit in mind, it could be a VERY tough ai to beat as my
> unsuspecting friends playing on a LAN versis some ai's found out! (That
> plant later became part of Counterstrike ai, albet with a much-reduced
> buildrate (2x regular) and with less resource production than it needed to
> run at full speed.) At the time I mentioned the hacked ai-only factory, I
> did not know for certain how BAI was designed -- just that it was
> (supposedly) nearly unbeatable.

I always find it odd that cheating AI's are complimented as "The best around", because, hell, anyone can do that, just tell to make plants that also produce like mega moho's and fusion reators, using construction units that build everything in 1 second.  There we go, new *ultimate* AI.

> CounterStrike ai is my counter to the accusations that I was copying BSR's
> BAI. Developement of it started about the same time as BAI, but almost
> everything similar to parts of BAI I made from earlier sources (mentioned
> later;) and are done for different reasons. BAI sustains its base production
> with massive resource-producing hacked units that it builds more and more of
> throught the game. Counterstrike builds 2 special resource storage buildings
> that provide a VERY good starting income, (+60 metal/sec and +1200
> energy/sec -- the fluxors!) and builds normal fusions/hacked cloaking
> fusions/normal moho metal makers to sustain its later production. Often, the
> fluxors end up providing less than half of its income by the 40 minute mark
> -- meaning that killing the fluxors will only slow it down (and may not even
> do that if because of the overbuild bug it decides to build 10 more using
> its many adv. c-units!) It uses no hacked c-units, only 2 hacked factories
> (of which neither is neccessary, but the hacked krogoth gantry is cool...),
> only 2 hacked resources per side which have very strict limits. There's a
> limit of 2 on fluxors and 0-4 cloaking fusions -- a fluxor is roughly equal
> to 2 ai-only solars and 1.5 ai-only metal extractors with the BAI on hard --
> and BAI doesn't stop with 4 solars and 3 metals either! The cloaking fusions
> make twice the energy of regular cloaking fusions, produce +10 metal, and
> cloak for only -1 energy all for the same cost as the originals. (But they
> must pay the cloaking cost even while being built -- giving the appearance
> of adv. c-units building nothing :)
>
> MadRat, Creeperman, and Speed are the 3 guys behind the REAL cool ideas in
> Counterstrike. MadRat came up with the hacked ai-only bombers that use
> much-increased LOS and better badtargetting lines to find targets from
> outside missile tower range. (Up till that point, we all thought the ai
> always played as though it had perm LOS!)

Ah yes, that.  My testers have been debatig over this one.  is it worth using scouting units?  I know that the AI units can see further than a human's can, because missiles regualy come in from out of your units LOS.  I just assumed that AI units could see as far as their guns, and that the AI strategic brain (if it can be called that), ignored the fog when making decisions.

Is this right?

> I took that ability one step
> further and applied it to all armed aircraft normally found in the game.
> Speed created the first known (to me!) ai-only unit, the ai nuker -- a nuke
> silo that the ai can build that works. Creeperman came up with the ai-only
> jamming units that use a 0-damage weapon so they chase the attack force
> around. (Now the ai's attack force doesn't always show up on radar...) He
> also helped me make the ai-only nuke silos and bombers even better. I came
> up with the ai-only EMP silo and Neutron silo -- a simple offshoot of the
> ai-only nuke silos. If nothing else, the ai-only EMP silo has an annoying
> habit of draining player's antinukes in multiplayer vs ai games. Lastly, I
> came up with ai-only cloaking fusions because the ai NEVER cloaked the
> original cloaking fusions but did landmines -- however I found out that
> unless I changed the cloakcost, the ai would have to pay -450 energy/sec
> (-550 for core's!) even while the cloaking fusion was being built which
> could take ~10 minutes of gametime thus using up about 250,000 energy to
> keep it cloaked before it was finished. Last but not least, I made a
> cloaking ai-only Arm Shooter (that cloaks for -1 energy while still and
> being built and -200 energy while moving) that is easily among the most
> annoying things the ai can do to a player -- big blue lasers shooting from
> out of nowhere. :p

By the way, may I use your AI Jammers?  As far as I understand, they don't cheat?  They would be great!  Can I?

> Strangely missing is an ai-only antinuke silo. This is because neither
> Speed, BSR, nor I could get one that would work -- it's a limitation of the
> code that defines how antinukes work.

Yes, I see people often talking about them.  Early on I inquired into anti-nukes, but no success, I was told it isn't possible.

> Hacked units to add abilities the ai should have but doesn't (cloaking
> fusions, bombers that actually drop bombs occassionally, nuke silos that
> nuke, jammers that don't act like buildings, etc.) is good but results in an
> ai that requires ALL players to have all the hacked units and ai to work. An
> ai that uses Ai profiles only can work even if other players DON'T have it
> -- which fortunately for me allowed me to test ai vs ai games via the
> internet and on LAN's.

How do hacked bombers help the AI bombers drop bombs?  I'd also be interested in those.  I ship all the things the AI needs with the AI. Bombers are a pain, they spend so long aquiring targets that they get shot down before doing anything.

> >Your AI Tweaking Guide? Do you mean the AI guide on TAEMC? If so, remember,
> >that's what started this conversation in the first place :) I had already
> >done most of what the guide said, though it did contain some points which I
> >had never considered. Then I watched Mostly Harmless in action, and with
> >that made some big changes to my AI. Before, I had considered a ratio of
> >twice as much metal being used as being produced, reasonable, then Mostly
> >Harmless showed this need not be the case, and so I have produced a better
> >resource system for my AI.
>
> Ideally, the ai's resource drain should start out much greater than its
> production and by the end of the game it should have almost the reverse --
> its resource drain should be less, but not MUCH less than its production.
> It's ok if the ai wastes much of its energy production later in the game,
> but it should probably be more or less equal on metal useage/drain -- or any
> player worth their salt will be able to outproduce the ai so badly it isn't
> even funny.

I've got this pretty much.  THe problem with the AI is if it starts to lose. When it starts to lose, it has less metal extractors, but it still continues to build level 2 plants.  Arg!

> >MH also showed me (through it's weightings), that resources could be
> >weighted higher than plants, and although this may seem odd, the idea had
> >never struck me before.
>
> Even in a 5-minute game it makes sense for the ai to have at least SLIGHTLY
> more solars and metal extractors than factories.

However I wasn't sure how interlinked the 2 were.  It seems almost impossible to stop the AI building a resource base first, therefore the 2 weightings may be completely unrelated, and just use commands such as "have +50 energy and +3 metal, you may now build plants".  I have no idea how interlinked these weights are.  Trial and error tells all.

> The ai profile numbers
> don't really correspond to ideal interpretations, because the ai game engine
> even "interrupts" and overrides the ai profiles at times (case in point is
> how many moho metal makers it will build even when weighted extremely low
> and limited to 1.)

Mohos are a problem, in fact Version 3 had this problem, the AI needed more metal, so I made it build Moho's, however I later found that the metal problems were being caused by units building the Mohos!

> The numbers in my ai profiles aren't always "sensible
> values" though, they are often what I had to plug in to get the desired
> results out. I finally settled on a weight range of 0.05 to 1 where the
> values could be roughly computed as percentage build chance -- with 1 = 100%
> (build any chance you get) and 0.05 = 5% (rarely build this.) If you use BIG
> weights and little weights like 9 and 0.1, even a small weight like 2 would
> get built more than the ai profile says it should. In Mostly Harmless, I
> tried to force the weights down to as tiny a value as possible - to minimize
> the chances of late-game items being built early and to maximize the
> must-build-now items.

My AI uses 0.05 up to 12.  It started based around 1, however as the AI developed ratios became less important because the AI was done in certain areas, and I knew what it would do.  All I needed to do now was to nudge it in the right direction.  Sometimes though you just wish you could give it some fixed scripts, build this, then this, then this, if you lose this go back and rebuild it.  Do not build this until you have this etc.

> Even if you only tested the ai on one map, say Painted Desert for instance,
> you could still make over 10 different ai's to deal with VASTLY different
> conditions on that map:

Yes, but the best one is the one that can handle as many of those situations as possible.

> An ai geared for 1k/1k starting resources is vastly different than one
> geared for 10k/10k. The 10k/10k one should attempt to build a lot of early
> factories and c-units to quickly convert the metal and energy into something
> USEFUL, the 1k/1k one should build more early resources (higher weights on
> resources, lower limits on factories) so it has enough metal and energy to
> make something USEFUL.

This is more like my AI.  It has some very tight plant restrictions, however am tempted to try a blanket limit rather than inividual limits, as individual ones allow it to build what it wants, but eventually it spreads into every field of attack and has too many plants.  On the other hand, a blanket may means it builds too many level 1 plants, and doesn't allow itself to build level 2 items.

> An ai geared to rush early and one that's geared to survive an early rush is
> also vastly different -- which is also VASTLY different than one that tries
> to get lots of resources quickly. An early rush ai (Rusher) makes minimal
> defenses and spends most of its resources making attack units. The defensive
> ai (Porcupine) would concentrate more on defenses - neglecting both attack
> units and resources some as a result of it. And the resource-grabbing ai
> (Octopus) would have practically no defenses (because that was time "better"
> spent making more resources) and fewer attack units than a Rusher ai -- at
> least early on.

My AI currently tries to raid quickly, however the weights currently work well enough so that the second the raiding party is gone, the construction units appear, some defence is built, and the base and attack plants expand. Currently I am trying to take advantage if the AI's one advantage, it's ability to do everything at once.  I try to get it to expand, build more, and attack all at the same time to keep the player too busy from doing everything too.

> An ai that expects a unit limit of 200 (as defined in the missions or
> pre-v2.0b1 patches) or 250 will almost always have very low limits on
> everything it can build - and generally has to eliminate many units the ai
> can use only when it already has a big base/attack force already built (like
> the Arm Annihilator or the Arm Shooter.) On the other hand, an ai designed
> for a 500 unit limit should attempt to build bases that look like cities and
> armies that look like ants on the move. :)

I haven't taken this too seriously, if the AI lasts that long it generally cripples it's resources, the AI will almost always lose at unit max, unless it holds more territory.

> The Arm ai and Core ai should be thought of as 2 separate ai's (they are,
> btw!) with Arm using lots of level 1 attack units and few base defenses, and
> Core using lots of base defenses and trying (hoping) to build level 2 units
> (and krogoths!) to counter Arm's strategy.

Currently my AI's are not vastly different.  They use different unit weights, but not much else.  I may make CORE more defencive, but I feel the AI always works better in attack as units all put similar drains on resources, where as decent defences don't.  Attack is easier to build a resource model for, defence it is near impossible.

> It could be that you want a generalist ai that builds a little of everything
> (best for high starting resources on a metal-rich map with few
> constrictions) or a specialist ai that builds far fewer types of units but
> builds a shitload of them (the extreme of that would be a flash-rushing ai.)
> It's almost as a rule specialist ai's designed for a single map and under
> known starting conditions will beat generalist ai's, all other things being
> held equal.
> Last but not least, there are 3 different skill settings for which an ai can
> be designed for - easy, medium, and hard. Cavedog's ai's treats each as a
> totally different ai - and as such on easy the ai might play more
> defensively and do even poorer on resources, on medium it might rush but
> lack base defenses, on hard it could both rush and build more resources and
> build more base defenses. Maybe it goes for more aircraft on hard too?
> (That's called an Eagle strategy -- strong air, weaker ground. It's probably
> the hardest player strategy to defeat in the late-game.)

I have simply given it a better idea of what's going on.  Easy is pretty much the same for every map.  Medium weights towards the more useful units, but is a bit slow to get going.  Hard is as good as I can get it.

> You have to KNOW what you want your ai to do, and you have to recognize what
> kind of opponents it's likely to be up against. What's more, you have to do
> some little things to make it better at fighting its counter even though it
> might weaken it against what it's BEST at defeating.

It's designed to take on everyone :)  Its designed to keep a constant flow of troops hitting the player.

> Take my Mostly Harmless Default.txt Arm ai, here's what I expected:
>
> (from the map/starting settings)
> 1k/1k starting resources, (it tries to build lots of early resources and few
> defenses assuming it cannot afford them)
> 250 or greater unit max, (it has limits too low for 500 unit max, but too
> high for anything less than 150 unitmax)
> mostly open all-land maps, (it ends up spreading out a lot)
> starting locations that are either defended on some sides, or in a corner
> (thus the lower number of defenses should prove enough for a single attack front)
> and ample metal spots. (it builds LOTS of metal extractors)
>
> (from the other player)
> The ai "expects" only one opponent - because its attack force always heads
> towards its nearest opponent, and its base defenses (if there ARE any!)
> aren't adequate against attacks from more than one side.

I dind't worry about this too much, in a FFA game, there are always 2 or 3 people who come out on top, as my AI begins to win, it's base expands and defences become heavier, while it maintains a good flow of troops.  1 on 1, the AI has so many flaws that it faces a challenge.  That is why mine attacks, to try to convince the player to hold the attacks or keep the
player too busy to raid.

> Also, attacking the ai from multiple sides with each attack having specific
> goals in mind (one to blunt its assault force, another to kill all its merls
> from behind, some to destroy wandering c-units and metal extractors, and a
> major attack to kill fusions and factories, etc) will generally make the ai
> look like a joke.

I tired heavy defence, but it stops the AI attacking, and generally defence is no match for a nicely mixed attack force as the AI has no idea where good placement spots are.

> The ai doesn't handle a player who raids resources instead of trying to rush
> into the center of the ai's base and kill its commander. (Ussually, the
> greatest concentration of base buildings and defenses is near the ai's
> commander.)

You can encourage it to defend resources by heavily weighting LLT's and Rockets.  My MEX's generally have a turret sitting around somewhere nearby. It's not enough to stop the raid, but does help slow it down.

> That the opponent's goal gives the ai at least 5-10 minutes of time in the
> beginning of the game where it doesn't have to contend with getting its
> pathetic base destroyed.

5 minutes for Queller, if it can be beaten then, I redo it.

> That the opponent doesn't rig up an "impenetrable" wall (actually quite hard
> to do on true LOS, unless you are also destroying the ai's base) and just
> use the metal from dead ai units to run more and more of their base, or
> worse yet build an undefended dt wall to hedge out the ai's ground forces
> entirely. (Even if the ai cannot get to you by ground units, it is too dumb
> to quit building them.)

I mention in my readme that for the most enjoyable game the player should not use DT's, simply because the AI does not adept to it's situation.  To combat DT's I would have to create a lot of Hammers and such, but then the player would build no DT's, come out with his Flash force, and paste the AI. There is no real solution, other than to ban DT's.

> That the opponent doesn't build a 100+ aircraft swarm and swamp the AA
> defenses of the ai and kill or cripple it quickly. (This is almost
> impossible for players to counter, especially over the internet with its
> lag, so the ai has even MORE trouble.)

I would hope that my AI doesn't give the player that chance.

> The ai cannot handle a concentrated attack on a specific target, such as a
> fusion -- it will NOT, as players might, concentrate in a manner to stop the
> attack. Heck, it might act like it's not being attacked at all. It tries to
> attack the NEAREST target to the leading edge of its attack force,
> regardless of what or where it is -- that is ALL it knows as far as mobile
> defense is concerned.

I thought this was a large over sight, it's not tough to tell the AI to recall it's force because a sizeable enemy force is attacking.  I wish they'd allowed the AI to have more than 1 attack group.

> (from itself)
> It tries to expand and cover all nearby metal spots with metal extractors -
> all while building at least half as many solars as metal extractors.
> It builds few factories, and instead relies on hordes of c-units to give it
> a rapid base expansion.
> It uses missile towers as its primary form of base defense, both from light
> air attacks and ground assaults.

yep :)

> Its attack force is designed mainly as an anti-base weapon, but should do ok
> against a similar attack force in the field.

Mine can do both.......I would hope.

> The early bertha gets the worm, thus it tries to get level 3 units earlier
> than may be wise.

The AI doesn't like building them, actually it doesn't handle them that well either.  I weighted it quite high, but it's not a top priority.

> If left alone for an extremely long amount of time, (like 4 hours) it might
> build enough aircraft factories to field considerable numbers of aircraft
> and replace them in 10 minute intervals -- thus resulting in a large swarm
> of aircraft hitting your base about every 10-20 minutes.

The only way the AI would ever bunch aircraft is if it covered the map with them, or had so many plants they were all started and finished at the same time.

> If you attack its base but fail to overrun it, any c-units remaining in the
> area (particularly the ai commander, ai decoy commanders, and ai c-airs)
> will probably reclaim the metal and use it against you.
> Its attack force is meant to be (but may not succeed in being) something the
> player cannot ignore, no matter HOW strong their wall/s are.
> It cannot be counted on to do anything as early as possible, but should
> manage to "get by" better than the average TA player found on boneyards.

That's the idea :)  I saw that once.  Queller 4 it was, a tester of mine was destroying metal extractors, but for everyone he destroyed, a construction unit on the other side of the map replaced it :)

> Because the ai spends so much time having its c-units just wandering around
> instead of building (I don't know about you, but I build groups of solars
> almost as quickly as buildtimes allow by moving little), someone who moves
> their c-units less per build order, uses a shorter route to metal spots for
> c-units, builds concentrated defenses only in a very few key areas, gets a
> fusion/moho metal maker combo going sooner (or geothermal/moho metal
> extractor combo on metal maps or light carriers/moho metal maker combo on
> water/land maps), makes a massive attack force from the resource advantage
> over the ai can defeat the ai with little more than CTRL+W (select all
> mobile attack units) and 1 left mouse button click on the minimap telling
> their units to move them to the far side of ai's base (with +SHOOTALL on
> to kill resources as well).

I put solars close together too, they get finished faster.  I've seen the AI Commander go on huge walks, only to place a Solar.  It really begins to get on your nerves when you want it to get a plant up fast.

I have used the steam roll tactic on the AI before, nothing like 50 Panthers rolling across the planes.

> Resources are more important than kill ratios - you might have 10 losses per
> every enemy you kill (if you choose your targets wisely!) but because you
> constantly increased production at 1.5 times your opponent from the very
> start till the finish you can afford such losses. Losing 20+ aircraft but
> destroying 1 fusion or using 10 flash tanks to kill 1 commander (you don't
> get credit for units killed by his explosion) is almost always a good deal.

If your AI has high losses it may be a good thing, it could mean that ti has reached the enemy base, and it procedding to attack the defences (because of Coffee break it always takes hig losses).

> So, as you might expect, I get the occasional email message stating that
> although my ai is ok or even good at hitting players with an early attack,
> it seems totally lacking in the late-game -- and much of its base can be
> destroyed with attacks of 10-bombers at a time.

Generally my AI plays the whole game well, however there is one problem with it, it cannot afford a majoy loss.  if it's goliaths come across enemy goliaths, it would expect a pretty even battle, but if the nemy come up with something cunning, the AI will generally not handle the loss of lots of big units, for few of the enemies. Some times I have known a Commander explosion to deat this (and other) AI's.

CounterStrike on the other
> hand ussualy gets emails about cheating too much, sometimes even in ways
> that it either doesn't cheat (such as being able to build aircraft faster
> out of its air plants than players) or cheating in ways that all ai's can do
> -- such as acting like it has perm LOS. I have learned to find such emails
> not as criticism so much as a source of humor, because I know many
> limitations to TA that prevent me from making an ai like I WANT to make.
> BTW, give a try on the modified krogoth campaign and the extra campaign I
> made for TA in my v1.1 bugfix -- I'm sure you'll find the ai at the least
> VERY annoying. :) Campaign missions allow you to tell the ai EXACTLY how to
> use its units, but the ai will only build original-game v1.0 units unless
> explicitly ordered.
>
> DON'T FORGET THE OVERBUILD BUG! My ai profiles have what appear to be
> too-low limits because I *KNOW* the ai will more often than not build far
> more than the limits I used. Had I put in the ACTUAL numbers I wanted, the
> ai would just produce that many and more!

How could I forget that :).

> Lots of construction aircraft tend to allow the ai to build metal extractors
> quickly, but the ai cannot build them as fast as it can c-vehicles. I use
> c-air's for the ai's "explosive expansion" phase starting roughly 10 minutes
> in the game and lasting up to 30 minutes. The ai ussually hits its metal
> extractor limit before the 40 minute mark, if you're haven't been killing
> its c-units and metal extractors. And 40 metal extractors making +1.0/sec
> metal each adds up to a LOT of metal produced for very little metal and
> energy investment (less than the cost of 3 moho metal extractors or 1
> fusion plant.)

I avoid c-planes, because if the AI has them, it means it has a air-plant, which means it has no real force (level 1 air is truely pathetic).   leave air till later.

> If you've ever played Magic the Gathering the card game, designing an ai for
> TA is very similar to how to design a GOOD magic deck. Almost anything the
> ai can build is benificial, but some things are more benificial than others.
> And one thing's for sure - an ai cannot be made to do everything and do it
> quickly unless it cheats like a dog. (BAI and Counterstrike would be like a
> magic deck that gets twice as much mana or more per land and draws 2 or more
> cards per turn.) A small deck of 1-3 mana-cost green creatures can easily be
> made, but will have poor long-game surviveability. (IE: an Arm Flash-rushing
> ai that can handle a max unit limit even as low as 100 because flash-rushing
> is about all it does.) A blue "denial" deck with lots of powerful flying
> creatures has a very slow start - it is also very hard to play because of
> the difficulty in knowing what spells to counter. (IE: a Core Vamp-swarming
> base-from-hell ai -- the ai is poor at concentrating its air attacks into
> one big punch, or even getting lots of aircraft to begin with.) EVERYTHING
> has a counter, but some things are cost prohibitive to counter such as an
> Arm Pelicans swarm in the open ocean (mostly because of the Pelican bug.)
>
> This ended up being longer than I intended.
> David

It was nice reading it, took a long time to reply too :)
--------
>From: "Quitch" <quitch@fickle.demon.co.uk>
>Reply-To: "Quitch" <quitch@fickle.demon.co.uk>
>To: "David Sellick" <switeck@hotmail.com>
>Subject: Re: Bugfix 1.1
>Date: Tue, 10 Aug 1999 01:14:16 +0100
>
>May I ask you something?  Yes, good :)
>
>How do your improved bombers work?  I mean, how do they improve the TA
>bombers (in what fashion), and how did you get them to improve in that way?

TA's ai somehow uses the bomber's LOS radius as its "scanning" radius for potential targets. Because previously the original bombers' LOS were shorter than their bombing runs, the bombers would more often than not overfly their intended targets THEN turn around and try to bomb it -- which equals almost no chance of success if even 1 missile tower is around. The altered bombers have LOS set to 1,000 instead of 350 -- this is incredibly unfair for the players to have because they'd be using bombers instead of peepers. Another benifit of the huge LOS radius for the bombers is while they're on their stupid wandering patrols they are FAR more likely to find any lone enemy unit and bomb it than even a player would. I improved on MadRat's designs (he made the first improved LOS ai bombers) by playing around with bad targetting categories so the bombers would be MORE likely to destroy vital targets first (assuming they get only one pass.)

>Could I please use them in my Queller AI?
>As long as they don't cheat that is.

They cheat, as mentioned above - plain and simple. However, if you DON'T use them, I suggest not using bombers at all in the ai because the original bombers suck so bad in the ai's hands. You can use them, but remember that your ai will technically be a hacked ai if you do. This means if you want the ai to work right you'd have to have it fully installed on ALL computers (in exactly the same form) or it won't use any ai-only units at all, which in the case of BAI and Counterstrike means they don't do anything.

>I've said it before, but I'll say it again as a nice ending statement :)
>I love all the work you've done for TA, especially the notes on how to speed
>up TA, and those great Jammers that move!

Creeperman made the Jammers -- I don't think I changed them any. They are very annoying I admit -- epecially combined with the cloaked Arm Shooters. Not too expensive to counter if that's all there is, Shooters and Jammers -- but that almost never travels alone.
--------
>From: "Quitch" <quitch@fickle.demon.co.uk>
>Reply-To: "Quitch" <quitch@fickle.demon.co.uk>
>To: "David Sellick" <switeck@hotmail.com>
>Subject: Re: Bugfix 1.1
>Date: Tue, 10 Aug 1999 20:05:34 +0100
>
>AH yes, I was scanning the bomber files.  I must admit they are a lot better, and they actually...........bomb stuff, wow! :)
>
>I never think of such things as "cheating" units, more a sort of "compensation" for the fact that the AI doesn't understand peepers or is just generally thick.  It bombs, humans can bomb, so in my opinion it doesn't cheat.
>
>Don't worry, my readme explains to people how to install the units.
>
>Thanks!Quitch(y)
>Creator of the Queller AI for TA, CC and UH
>"Anti-AI tactics are the last refuge of the coward"

Oh, one last thing... BEFORE you may use the ai-only units out of Counterstrike and release your ai so others can use it, I want to see your ai myself.

Please send me a copy.
David
--------
You certainly can, here is the version 9 work in progress. Enjoy! 
Quitch(y)
--------
Topic: Your ai's problems

>From: "Quitch" <quitch@fickle.demon.co.uk>
>Reply-To: "Quitch" <quitch@fickle.demon.co.uk>
>To: "David Sellick" <switeck@hotmail.com>
>Subject: Re: Bugfix 1.1
>Date: Tue, 10 Aug 1999 21:14:32 +0100
>
>You certainly can, here is the version 9 work in progress.  Enjoy!
>
>Quitch(y)

Time for a dose of humility...
Ok, your ai has a good start -- especially good considering it's not based on mine or BSR's or anyone else's I'm aware of.

But it has a couple major problems that can be changed.

1.Move the Weight Arm 0.1 and Weight Core 0.1 lines to the BOTTOM of the ai profile -- they OVERRIDE any and all other weight settings (forcing the ai to use only the built-in exe-part to decide what to do next).

2.Lower limits to something at least vaguely reasonable. Were you to add up all the POSSIBLE units the ai is allowed to build, I'm sure you'd find it somewhere over 1,000. Try to get that number down to around 500 -- which is the HIGHEST unit max the game can be set to in skirmish. (btw, it can be set to 1,000 in missions ;)

3.Your ai seems to like hovers too much. Hovers are expensive, especially considering I saw your ai in skirmish (1k/1k, Painted Desert) trying to build 2-4 hover plants at a time.

Good luck pulling this all off.
David
--------
From: "Quitch" <quitch@fickle.demon.co.uk>
Reply-To: "Quitch" <quitch@fickle.demon.co.uk>
To: "David Sellick" <switeck@hotmail.com>
Subject: Re: Your ai's problems
Date: Fri, 13 Aug 1999 22:08:42 +0100

Ah, the Weight ARM and CORE. I must admit that the layout of the file is strongly influenced by the first AI I saw, which happened to be the BSRAI. You're right about basing though, it did start as I merely changed the BSRAI weightings, and I had no intention of playing with defence or resources. However, AI has always been my obsession, and here was my chance to do something about it!!! Well I savaged BSRAI, deleted all the values, spent a day entering *every* code, and Queller was born (I still have every Version, Version 1 seems very boring now). 
What exactly are you saying weight ARM and CORE does? BSR believes it only affects units you don't weight and has it at the top of his file (he said it was something the two of you disagreed over), you believe it overrides weights. From my tests it seems to affect construction. I set it to 0, the Commanders did nothing, at 0.1 my weights seem to be followed as far as I can see, and at 10 the Commander will take a while before building a plant, and when he does, it will almost always (no matter how you weight it) churn out a Construction unit first. Could you please explain in more detail what exactly you think those values do? I tried removing them once, but then Construction came back with avengence, even though it's weighted very low. 
Ah limits :) Yes I never looked carefully at them :) So far no one has ever reached a point with this AI where it has hit them, or the unit max. Even if the totalled values reach well over 500, a lot of them will never see the light of day, for example Wind Power will virtually never appear in my default profile. A game always has stages where the number of units is lower than the unit limit, however not all games have a stage where the unit limit is hit, therefore I believe it is safer to overlimit rather than underlimiting. However I will take a look at them, however none of my testers has hit the limit against it yet. 
As for hovercraft, I think that was probably bad AI luck. I have weighted them quite low, and set a limit to 1 for their platform. I found they were expensive and built too early would cripple the AI. However, built at the right time they are invaluable, they get to fights quickly, have a longer than normal range, and deal out more damage. They are however weaker, so I stick to missile hovercraft. If it was building 2-4 platforms, then it was just unlucky. 
Do I get permission to use your aircraft? The version I sent you I'm afraid wouldn't, because I messed up the limits, and didn't set the old aircraft to 0 at hard (although I did at medium and easy for some reason). 
Quitch(y)
--------
>From: "Quitch" <quitch@fickle.demon.co.uk>
>Reply-To: "Quitch" <quitch@fickle.demon.co.uk>
>To: "David Sellick" <switeck@hotmail.com>
>Subject: Re: Your ai's problems
>Date: Fri, 13 Aug 1999 22:08:42 +0100

Some semi-relavant stuff -- in a FILE ATTACHMENT called useful.zip!
BTW, the default.txt is my latest v12 of my nonhacked ai.

I hope it is of some modest value to you as it is poorly formatted stuff I've accumulated over the last 2 years.

David
-------
>I found the AI discussion quite fascinating, and wish I'd been there to input.  Mind you, I would have disagreed with almost everyone about the whole C-units in the beginning issue.
>
>It almost seems to be an AI designers obsession to get C-units out and about immediately, yet this is *exactly* what 3.1 did, and you never saw it win a game.  Considering that most ARM players rely on the quick kill, any AI which is going to insist on building first, is in one fell swoop alienating half the TA population.  Something I have tried to achieve (and in my opinion done) with my AI, is that is will construct, but only under the cover of an attack force keeping the enemy busy.  Also, looking at how the limits and weights work together, only convinces me to keep my current limits.  Keeping the weights and limits of combat units high, means that the AI will only build construction units if it has lots of combat troops, and this means it is probably winning at that point.  In my view, this is great!  When winning it can afford to construct, and will be under cover of a large army.  Should the AI start to lose i.e. lose combat troops, it will build just troops, and stop wasting time on construction units, so that new combat units may aid it in retaking lost ground.
>
>  Actually I need to ask you an AI speed issue.  Obviously the time between an AI moving from one task to another is determined partly by the speed of your machine.  However in some cases I have seen a Commander pause for 1 second, then build a solar, and then right after r that, pause for 5 seconds before building another.  Why the longer pause?
>
>Also, does arrangement of the weights and limits in an AI file change the speed of the AI, or does it read through the whole difficulty level before making a choice?  Or, does it read through the *entire* file before choosing?  Will it go quicker at the start if I put resources at the top of the file, or would I need to put resource weights *and* limits there, or would it make no difference?
>
>Also, if Weight ARM and CORE changes weights universally, why does a higher weighting make construction units more likely, even though I have not changed the limits?  This has been tested over several games and is not a fluke.
>
>I look forward to your reply.
--------
Artificial Intelligence Theory, part (X+1-Y)*Z

"Ah, the Weight ARM and CORE. I must admit that the layout of the file is strongly influenced by the first AI I saw, which happened to be the BSRAI. You're right about basing though, it did start as I merely changed the BSRAI weightings, and I had no intention of playing with defence or resources. However, AI has always been my obsession, and here was my chance to do something about it!!! Well I savaged BSRAI, deleted all the values, spent a day entering *every* code, and Queller was born (I still have every Version, Version 1 seems very boring now)."

My original ai profiles came from the ai list made by Andrew Griffin -- I looked at Cavedog's ai profiles, but realized I needed more order in mine to keep track of what I was doing. Mine has had a very long evolution - and I have saved many of the steps in that evolution. The few 3rd party ai profiles that predated my first I did not find out about until after I was well on my way to making my own style.

This would qualify as a 300-level college course (if not higher!) by this point. And would meet the requirements as both a math course and a philosophy course. Prerequisets are previous ai study, familiarity of TA's ai, strong math background (with some knowledge of statistical analysis, how to read and do proofs, and geometry), ability to grasp nontangeble concepts, and a touch of faith.

I've practically flooded you with dissimilar theories about how TA's ai works. But I haven't always pointed out what should be differentiated into facts, theories, opinions, wild-ass guesses, and plain bullshit. The ai for TA is one big black box with only a tiny output window. I have only CLUES as to why it does anything and am forced to make some wild-assed guesses for why it does what I see. Some of these wild-ass guesses are failsafe modes and default weights affect everything (often adversely.)

TA has some pretty complex ai routines hard-coded into the exe, but it's my opinion the ai comes to the wrong conclusion after "evaluating" (ie: calculating) the situation. Case in point, a nonmetal map with no metal patches and 1k/1k start the ai commander builds a few resource structures and stops. Or when it's out of metal and energy -- so it tries to build more solars/metal extractors, even though it probably should just QUIT doing stuff instead of trying to do far more.

"What exactly are you saying weight ARM and CORE does? BSR believes it only affects units you don't weight and has it at the top of his file (he said it was something the two of you disagreed over), you believe it overrides weights. From my tests it seems to affect construction. I set it to 0, the Commanders did nothing, at 0.1 my weights seem to be followed as far as I can see, and at 10 the Commander will take a while before building a plant, and when he does, it will almost always (no matter how you weight it) churn out a Construction unit first. Could you please explain in more detail what exactly you think those values do? I tried removing them once, but then Construction came back with a vengence, even though it's weighted very low."

(opinion:) There's a default weight on everything -- even if there's no weight line for them in the ai profile and there's no general weights either. (like weight ARM 0.1) My support of such an idea is (experimental data:) nonweighted items still get built by the ai, (at least OCCASSIONALLY!) even without both a specified weight line for them OR a default weight line. At the moment, I can presume such a default weight exists -- but not its size/magnitude. (ie:a default weight exists, but I don't know its exact value.) This means even when you do not HAVE a weight ARM line in your ai profile that the ai engine STILL uses one! (That unknown default weight affects that ai by an unknown amount in unknown places in unknown ways resulting in unknown consequences.)

"From my tests ... I set it to 0, ... at 0.1 ... and at 10 ... what exactly you think those values do?"
Your test was more exacting than mine. I only tested the weight Arm 0 and 0.1 possibilites as well as if there was any differances between having the weight Arm 0 and 0.1 line at the top or bottom of the ai profile.

"BSR believes it only affects units you don't weight and has it at the top of his file."
However, that does not explain why nothing gets built when the weight is set to 0 -- at least not if the weight 0 is at the very top. But, if you put weight ARM 0 at the BOTTOM of the ai profile everything weighted specifically above it would still be built. Things that aren't specifically weighted above it are not though!

Thus I conclude:
1.if it's at the top it affects everything.
2.if it's at the bottom it affects everything not already mentioned.

Now, on to how you can use this to your advantage. Consider the issue of minimum and maximum priority. Minimum priority would be something you'd probably want on an Arm vulcan -- better to get a bertha or 2 (or 10!) built first. Maximum priority is something you'd want the ai to do FIRST and FOREMOST. (An example of that would depend on what GOAL you have in mind for your ai.) But how do you do that?

To discover how to give maximize priority to something in an ai profile, I ran a simple test with 3 limit lines of the same unit:
Limit ARMFLASH 255 (at the beginning of the ai profile)
Limit ARMFLASH 1 (at the middle of the ai profile)
Limit ARMFLASH 0 (at the END of the ai profile, in the plan HARD section)

And the ai started building a shitload of flash tanks! Obviously, the Limit ARMFLASH 255 line was being used, and the 2nd and 3rd lines were ignored. Thus, I concluded that the TOP of an ai profile was evaluated first and the ai would disregard any later entires in the ai for the same unit.

From running many unrelated ai tests, the ai engine often does the first thing it reads in my ai profile (that has a reasonable weight on it) -- even if later things in my ai profile have equal weights on them. (This could just be luck!)

So, I conclude that:
Any lines mentioned LATE in an ai profile would be ignored if any above lines would contradict them.
Lines of seemingly equal value (like 2 weight 0.5 lines on 2 different units) would have the FIRST line get a slightly higher result because the ai engine "read" that line first.
If I want something "do it immediately!" I weight it very high and put it at the TOP of the ai profile.
If I want something done after almost everything else ("only after hell freezes over"), I weight it at 0.05 and put it at the bottom of the weight list.

This is like a VERY complex geometry proof with many missing steps -- not missing only because I am lazy, but because I'm missing much critical information about TA's ai to tell you much more.

What I don't know, (and possibly where the disagreement with BSR occurs) is just how MUCH the default weight affects OTHER weights in the ai profile (even if it's further down than the other weights in the ai profile.) That's like saying does the weight ON a weight weigh more than a bigger weight with no weight on top of it? You figure out a scientific way to test that with ALL other variables fixed to known values, get some meaningful results, and we'll fight over the meaning of the results for a few weeks -- then we'll BOTH know the answer to that question. :p

I don't want the absolute maximum priority line in my ai profile to be "set everything to the same weight." That's sort of the problem with Cavedog's ai profiles to begin with -- nothing seems very important to it. It'd alternate building a solar, metal extractor, defender, factory, hlt, metal maker, and then a guardian. This doesn't work too well because not enough metal and energy can be created by 1 solar, 1 metal extractor, and 1 metal maker to build all the above-mentioned buildings without considerable delays from a lack of resources. I figured a far-greater ratio of resources to resource-consumers was needed. (It's worth noting that resources themselves are ALSO resource-consumers at least while being built -- this is why I limited the number of construction units so low, so they weren't outbuilding their resources!)


"at 10 the Commander will take a while before building a plant"
The possible reason the ai went into guru mode when you set the weight to 10 is because of a massive probability overload (1 to infinity minus 1 against :p ) -- maybe at some fixed value for the weight on a unit (say 4+?) the ai goes from a "it's a good idea" to "must be built now, in preferance to everything else!" Only problem with that is it is CONSTANTLY trying to evaluate the situation on what to build, thus each idea is disabling the last -- after ~5 minutes, the ai gives up trying to choose what to build and defaults to building a factory -- unless one of the infernal failsafe modes activates.

I think I can explain this another way - consider the default ARM/CORE weights as a multiplier value on the existing weights already in the ai profile on individual units. In the case of weight ARM 0, anything times zero is zero -- thus, nothing happens. If weight ARM 10, then all other weights (for Arm at least) is multiplied by 10 -- which no doubt results in a numerical answer that is typically over 1, which may be the "build me now!" threshhold. The ai engine ends up with everything being a "build me now!" status and has a brainfart. So it sits there doing nothing because it doesn't know what to do first until some overriding routine says - "dammit just build factories/solars/metal extractors if you don't know what to do!"

"Ah limits :) Yes I never looked carefully at them :) So far no one has ever reached a point with this AI where it has hit them, or the unit max. Even if the totalled values reach well over 500, a lot of them will never see the light of day, for example Wind Power will virtually never appear in my default profile. A game always has stages where the number of units is lower than the unit limit, however not all games have a stage where the unit limit is hit, therefore I believe it is safer to overlimit rather than underlimiting. However I will take a look at them, however none of my testers has hit the limit against it yet."

That ai discussion I was in:
 Forum: AI Corner 
Topic: Weights and Limits
reveals a lot of possible theories on how and why the TA ai responds as it does. While going straight for construction units in preferance over attack units may seem bad, waiting over 10 minutes to start building *ANY* construction units is even worse. I won't argue that getting construction units extremely early is a bad thing or waiting up to 10 minutes to get them is a bad thing either, but doing anything outside of that range is a bad thing for the ai. The ai commander simply cannot build any level 2 structures OR spread out metal extractors in multiple directions, and thus must be suplimented by c-units.

What's the point of having a unit buildable if you cannot count on it being built EVER? Also, I consider the limits on attack ground units in your ai profile to be too high because the ai can only use so many units effectively at one time. 40 flash tanks attacking at once sounds great, but the most that can effectively surround an enemy unit is about 10 -- figure in 10 reinforcements and a max limit of 20 doesn't seem too unreasonable. On the other hand, Samsons offer extra AA protection wherever they are at -- so their max limit can be higher, but will still eat up too big of a chunk of max units if it's more than 40 even with unit max at 500. Ai Jeffys suck because they don't scout ahead of the ai's other attack units. Ai minelayers suck because the ai doesn't know to build mines OUTSIDE its base instead of inside. Ai Stumpys are good, but their effective number is even fewer than flash tanks because they move slower and will no doubt be clogged up with all the flashes in their way -- they also cost more and take longer to build, possibly resulting in the Arm vehicle plants being underutilized.

"As for hovercraft, I think that was probably bad AI luck. I have weighted them quite low, and set a limit to 1 for their platform. I found they were expensive and built too early would cripple the AI. However, built at the right time they are invaluable, they get to fights quickly, have a longer than normal range, and deal out more damage. They are however weaker, so I stick to missile hovercraft. If it was building 2-4 platforms, then it was just unlucky."

The time I see for missile hovercraft (and merl hovers) is when the ai has a good economy going -- and this is almost never very early.

"Do I get permission to use your aircraft? The version I sent you I'm afraid wouldn't, because I messed up the limits, and didn't set the old aircraft to 0 at hard (although I did at medium and easy for some reason)."

YES, you may use them! Remember though that your ai is a hacked ai that is hurt considerably if the hacked aircraft are missing -- on hard it would build no aircraft at all! This doesn't mean your ai cheats, but it greatly reduces boneyards compatibility. (ie: your ai will not work at its best without them.)

David
--------
From: "Quitch" <quitch@fickle.demon.co.uk>
Reply-To: "Quitch" <quitch@fickle.demon.co.uk>
To: "David Sellick" <switeck@hotmail.com>
Subject: Re: Artificial Intelligence Theory, part (X+1-Y)*Z
Date: Sun, 15 Aug 1999 14:21:56 +0100

"I've practically flooded you with dissimilar theories about how TA's ai works." 
And I love it :) 
"Some of these wild-ass guesses are fail-safe modes and default weights affect everything (often adversely.)"

When I see a fail-safe act in an effective manner I will die of shock. That must have more than 100 resources is terrible, metal takes longer to get there, and it also takes no account of current production. I've seen the AI beaten in earlier Versions just because after building the plant, it had less than 100 metal, and wouldn't build there for a long time.

"There's a default weight on everything -- even if there's no weight line for them in the ai profile"

So I've seen. I've been meaning to have a shot at this to see if I can work out what it is, or if it differs from unit to unit.

"So, I conclude that:
Any lines mentioned LATE in an ai profile would be ignored if any above lines would contradict them.
Lines of seemingly equal value (like 2 weight 0.5 lines on 2 different units) would have the FIRST line get a slightly higher result because the ai engine "read" that line first.
If I want something "do it immediately!" I weight it very high and put it at the TOP of the ai profile.
If I want something done after almost everything else ("only after hell freezes over"), I weight it at 0.05 and put it at the bottom of the weight list."

Hmmmmm, I'm not so sure. After all, to make a decision a unit would have to evaluate all it's possibilities. For a Commander to build a solar, he doesn't just hit the solar weight and decide to build it (IMHO), he would also need to check the limit, and the weights of all the other items he can build. This in turn would lead to him having to go through the entire file until he had run across everything he can build and their limits, then he can made a probability judgement as long as no overrides are kicking in. Therefore to allow a swift decision, everything that unit could build would have to be near the top of the profile, rather than just one or two items. Also, limits would need to be there, as even if the unit does decide immediately, it cannot build until it has checked the limit status. So that makes me wonder, when making a decision, would it read through the entire file, read through all the units it can build, or just do a probability calculation as it comes to items?

Also, when reading the file, does the AI read from one "plan" to the next "plan" and loop, or to the end of the file? Is positioning at the top of the plan enough, or do you need to move your hard profile to the top of the "file"?

"What I don't know, (and possibly where the disagreement with BSR occurs) is just how MUCH the default weight affects OTHER weights in the ai profile"

Surely the default is ignored if a new weight has been entered? Actually that does make me wonder if the default weight have a modifying effect on the weight you put in.

"What's the point of having a unit buildable if you cannot count on it being built EVER?"

There isn't any, however when I said they may not see the light of day together, I meant in a probabilistic sense, on Hard nothing is turned off except useless units. Therefore if the game is going on forever, at least the AI will keep some slight variation going, maintaining interest. By this stage, it is unlikely to cause even the slightest detrimental effect to have the odd Zipper running around.

"Also, I consider the limits on attack ground units in your ai profile to be too high because the ai can only use so many units effectively at one time. 40 flash tanks attacking at once sounds great, but the most that can effectively surround an enemy unit is about 10 -- figure in 10 reinforcements and a max limit of 20 doesn't seem too unreasonable."

However, that merely assumes there is only 1 unit. If you have waves of defence, then sure 10 Flashes can surround an item, but those Flashes will take casualties. If you limit the Flashes too low, the AI will have to wait for another wave to hit. In the mean time the player rebuilds the defences, and reclaims the metal. However, having large excess means the AI can "roll" over the player in a gigantic wave if the player allows it the breathing space. This prevents the player being able to establish new defences at old positions, and erodes his territory, and meanwhile the reinforcements will arrive, and the AI will seem to the player to have an endless number of troops. Psychological warfare is also very effective.

"The time I see for missile hovercraft (and merl hovers) is when the ai has a good economy going -- and this is almost never very early."

Trying to measure the AI by time though is a hit and miss business. The AI may have a better income at 5 minutes than 7, because he gets hit by a raid. AT 15 he may seems to be doing well, then at 20 the player rush hits him. Rather than time, surely you should consider by weight? If you weight metal at a certain level, then you can be fairly sure than the AI will have X metal extractors and Y solar panels before he builds hover capabilities.

A point I have wondered, is should the AI sacrifice a plant from another category to allow it hovercraft? Currently my AI builds hovercraft in "addition" to other plants and the economy holds if it is left late enough. However, I have wondered if a lab or plant should be sacrificed for these craft.

"YES, you may use them! Remember though that your ai is a hacked ai that is hurt considerably if the hacked aircraft are missing -- on hard it would build no aircraft at all! This doesn't mean your ai cheats, but it greatly reduces boneyards compatibility. (i.e.: your ai will not work at its best without them.)"

There are of course 2 solutions:
1. Create non-hacked versions.
2. Mention that not every Quake addition worked with QuakeWorld :)
--------
"When I see a fail-safe act in an effective manner I will die of shock. That must have more than 100 resources is terrible, metal takes longer to get there, and it also takes no account of current production. I've seen the AI beaten in earlier Versions just because after building the plant, it had less than 100 metal, and wouldn't build there for a long time."

You think that's bad -- I've seen occassions where the ai's metal (in storage) went NEGATIVE. You can only imagine the delays before the ai did another thing.

There's a default weight on everything -- even if there's no weight line for them in the ai profile.
"So I've seen. I've been meaning to have a shot at this to see if I can work out what it is, or if it differs from unit to unit."

Failsafes fuck up any attempt you make at putting precise values on the built-in default weight. Resources and factories will always seem predominate. Probably the reason your ai turns into construction fever when you remove the weight ARM/CORE lines is because of the failsafes. This is why you gotta watch the ai's resources with +VIEW #.

"Hmmmmm, I'm not so sure. After all, to make a decision a unit would have to evaluate all it's possibilities. For a Commander to build a solar, he doesn't just hit the solar weight and decide to build it (IMHO), he would also need to check the limit, and the weights of all the other items he can build. This in turn would lead to him having to go through the entire file until he had run across everything he can build and their limits, then he can made a probability judgement as long as no overrides are kicking in. Therefore to allow a swift decision, everything that unit could build would have to be near the top of the profile, rather than just one or two items. Also, limits would need to be there, as even if the unit does decide immediately, it cannot build until it has checked the limit status. So that makes me wonder, when making a decision, would it read through the entire file, read through all the units it can build, or just do a probability calculation as it comes to items?"

"Hmmmmm, I'm not so sure."
What part are you not so sure about? (1,2,3,or 4)
1.Any lines mentioned LATE in an ai profile would be ignored if any above lines would contradict them.
2.Lines of seemingly equal value (like 2 weight 0.5 lines on 2 different units) would have the FIRST line get a slightly higher result because the ai engine "read" that line first.
3.If I want something "do it immediately!" I weight it very high and put it at the TOP of the ai profile.
4.If I want something done after almost everything else ("only after hell freezes over"), I weight it at 0.05 and put it at the bottom of the weight list.

"to allow a swift decision, everything that unit could build would have to be near the top of the profile"
If you notice, in my ai profiles, ALL the weights are near the top and ALL the limits are towards the bottom. There are NO weights in the plan easy/medium/hard sections -- because I see no need. It doesn't matter so much if the very low weights are at the bottom of the ai profile so long as they are BELOW higher weighted items that could be built in preferance to it.

Maybe it's not obvious, but other than the universal weights (weight ARM/CORE #.#), there should be NOTHING in your ai profile that's conflicting -- like a limit ARMFLASH line 2 or 3 times -- because the later lines will be ignored in any case. The ONLY exceptions to this are in the plan easy/medium/hard sections since only 1 could ever be active at a time. (It is imossible for the game to be set to both easy and hard ai difficulty on a single computer, although I have done that on accident in multiplayer -- my ai was set to easy, my opponent's ai was set to hard.)

"Also, when reading the file, does the AI read from one "plan" to the next "plan" and loop, or to the end of the file? Is positioning at the top of the plan enough, or do you need to move your hard profile to the top of the "file"?"
It is clear that Cavedog took at least a few shortcuts, which means even if a unit COULD do other things it often does the first thing it comes to in the ai profile. HOW it reads the ai profile is almost anyone's guess, but my guess is that it reads the ai profile only ONCE -- placing it into a simplified form in memory with each unit represented by a number and weights and limits being values under them. +reloadaiprofile supposedly allows you to change which ai profile the ai is using on the fly, but I think the only way to actually CHANGE the ai is to save the game, change the ai, then reload the game -- the ai profile used is NOT stored with the savegame info and is instead reloaded every time the game is started.

"Surely the default is ignored if a new weight has been entered? Actually that does make me wonder if the default weight have a modifying effect on the weight you put in."
The test you've already run seemed to suggest that -- if failsafes can be ruled out as a possibility!

What's the point of having a unit buildable if you cannot count on it being built EVER?
"There isn't any, however when I said they may not see the light of day together, I meant in a probabilistic sense, on Hard nothing is turned off except useless units. Therefore if the game is going on forever, at least the AI will keep some slight variation going, maintaining interest. By this stage, it is unlikely to cause even the slightest detrimental effect to have the odd Zipper running around."

True, so long as you can keep its numbers from skewing the ai into something it's not good at -- it's probably even a good idea.

"Also, I consider the limits on attack ground units in your ai profile to be too high because the ai can only use so many units effectively at one time. 40 flash tanks attacking at once sounds great, but the most that can effectively surround an enemy unit is about 10 -- figure in 10 reinforcements and a max limit of 20 doesn't seem too unreasonable."

"However, that merely assumes there is only 1 unit. If you have waves of defence, then sure 10 Flashes can surround an item, but those Flashes will take casualties. If you limit the Flashes too low, the AI will have to wait for another wave to hit. In the mean time the player rebuilds the defences, and reclaims the metal. However, having large excess means the AI can "roll" over the player in a gigantic wave if the player allows it the breathing space. This prevents the player being able to establish new defences at old positions, and erodes his territory, and meanwhile the reinforcements will arrive, and the AI will seem to the player to have an endless number of troops. Psychological warfare is also very effective."

The ai's pathfinding ability gets worse as it gets more units. With 10 units, it's great. 100 units, it's ok. At 200 units, it's poor. At 500 units, it's absolutely terrible -- at least on a slow computer. Even my P2-400 crawls along at a -3 speed at times when total number of units crawling around on a map exceeds 1,000. So with each addition flash you're letting the ai build, you're making any existing ones that much less effective. Before long, 2/3 or more of the would-be attack force is hung up on trees and hills somewhere on the map -- often in tiny groups that are easy to kill. This loss of resources would better serve the ai if it had instead spent that metal ramping up its production into the level 2 and 3 units.

And even on the largest of maps, I've never seen my ai hit max attacking units, drain its pipeline dry, get them all killed, and take too long to "restock" its attack force because it ussually fails at one or more of those tasks. Either it isn't at max attacking units or even if at max units dying too quickly to cause the "ant trail" to stop. If my ai has a maxed-out attack force, it either DOESN'T lose (because it has overkill for TRUE los games) or it CAN'T attack. (because DT are in the way) :)

The main reason why my ai's attack force comes in waves has NOTHING to do with unit limits or weights and everything to do with resources. It's run its metal in storage down to <100 and is probably in nano-stall as well (not getting enough metal/energy to finish current projects at full speed.) So factories quit adding to the size of the attack force - whether it's being killed off or not.

Barring all that, ussually even when the attack force is routed there are a handful (<10) survivors that flee to return later with friends -- the problem with that is they don't wait to get another 100+ strong force to do so and often reattack when their numbers are between 10 and 20. If the ai loses the initiative to a player, it has more than likely lost. Because if its attack force is dealt with, its other ways of hurting its opponent are FAR less effective. Cases in point: aircraft -- players are better w/ them, berthas -- ditto, nukes -- players don't nuke lone units much (and miss)... All told, this means that the ai's attack force is its main attack type.

You're not going to make a purely nonhacked ai (on a nonmetal map at least) that *CAN* throw another 100+ troops out in less than 5 minutes -- which is plenty of time for even a slothful player to reclaim much of the wreckage. That's why the ai's attack force should match its effective economy. Maybe BAI can get away with a max-size attack force of 200+ units -- because it has metal and energy running out its ears. Your ai, if it's nonhacked, WON'T. By the one hour mark, it will be LUCKY to have a metal production of +100 metal and an energy production of +4000 energy. Sure, an ai having 100+ flash tanks running around sounds great - but getting the resources to do that, there's the bottleneck. And if the ai's spending too much of its metal getting flashes, it isn't spending enough metal getting more metal extractors (and etc...) to get more metal to make more flash tanks. If 50% or more of the ai's resources are going into its attack force, that just means only 50% is left for its base to grow on. Quite possibly, the 50% spent to make the attack force ends up feeding its opponent metal!

A large attack force tripping over itself moves FAR slower than a smaller attack force, because there's more pathfinding problems and more "slow elements" causing the leading edge to keep turning around to let the slower units catch up. (Thus jeffys end up being almost a total waste, because they're almost always waiting on other units to catch up.)

Having a max attack unit limit of 500 is of little use if it's always getting beaten back because it took too long to attack and made a slow, stumbling attack when it did attack. If its attack force ends up doing little more than feeding its opponent, it has defeated itself! If it gets a huge attack force only to lose them all in one big combat, it has lost. If its attack force is too small, it will be just as bad.

"The time I see for missile hovercraft (and merl hovers) is when the ai has a good economy going -- and this is almost never very early."

"Trying to measure the AI by time though is a hit and miss business. The AI may have a better income at 5 minutes than 7, because he gets hit by a raid. AT 15 he may seems to be doing well, then at 20 the player rush hits him. Rather than time, surely you should consider by weight? If you weight metal at a certain level, then you can be fairly sure than the AI will have X metal extractors and Y solar panels before he builds hover capabilities."

OF COURSE you should consider by weight -- but you have to consider what that weight means timewise! Weighting factories extremely high so they are built early results in them being started early but finished later after a LONG period of nano-stall. (Example of this is when the ai commander tries to start a second factory immediately after the first.) Weighting them low means they're built later, but when they're started they are finished quickly. Weighting them extemely low means the ai gets overrun before the first is built. DON'T weight all the ai's factories equally -- or it will no doubt start 3 or more at a time. What I ended up doing is deciding whether I wanted kbots or vehicles first, and weighting them the highest.  Then I decided whether I wanted the secondary ground troops to get started before air plants -- if so, they're weighted higher, if not then lower. Then I decided if I wanted hovers -- and if so, I probably weighted them low. Aircraft are also weighted low. The end result (hopefully) is early on the ai cranks out a lot of mostly kbots or vehicles -- and slowly but surely these get reinforced by other types giving them the ability to do more tasks. If the ai gets delayed by a raiding set-back, it probably won't continue adding the additional types of attackers because it will be busy replacing the resources and initial factories (which are weighted the highest) FIRST.

The early ai attack force should be good at raiding and good at fighting small groups of mobiles in the field. It should be poor at attacking a hardened base because that is an expensive addition to the attack force that shouldn't be needed yet. It should also be poor at slugging it out with a large enemy attack force -- this will encourage a player to WANT to throw everything at the attacking force instead of raiding. And if they're slow in trying to stop the attack force, they'll fail at that too. Capabilities should be added slowly to the attack force -- till it reaches a point where Mr. Kroggy and friends doesn't slow it down. This means any level 1 units that aren't good at raiding and fighting in the field gets weighted LOW -- at least out of the primary/first factory. This is also why I decided my Arm ai needed merls out of the adv. vehicle plant before the other units because the others would be redundant or only a little better than units already in the field. Merls on the other hand would add a new attack type. And it sucks shit to barely be holding back an attack force and ABOUT to turn the tide when a merl missile crashes down on your lone guardian/hlt killing it. A panther or bulldog on the other hand would just be something else that could be skirmished to death with lower base losses -- they can arrive a little more slowly but end up reaching respectable numbers too.

If the ai has to worry about being knocked way down by raids, it shouldn't build hovercraft at all! 

You miss the point -- hovercraft for the ai cripple it if it builds them BEFORE it has a strong economy going, because the factory is so exensive and each hovercraft drains resources considerably for their return-on-investment. Same with aircraft. They're GREAT for the ai to have in the late-game, when it SHOULD be able to afford it. So I weight the hover plant low -- it is started after many other build weights are satisfied, so there should be enough resources for the factory to be built extremely quick -- maybe even with 2+ c-units assisting!

"A point I have wondered, is should the AI sacrifice a plant from another category to allow it hovercraft? Currently my AI builds hovercraft in "addition" to other plants and the economy holds if it is left late enough. However, I have wondered if a lab or plant should be sacrificed for these craft."

The LESS the ai has to do, the better it is at doing it. This is why my ai takes a minimal approach to lots of things. Sure, that means it lacks much variation but it means what it's left doing it may do as well or better than an average player. I have dropped hovercraft from my default.txt ai for that reason.

Increasing one weight is almost the same as lowering ALL other related weights (weights on units in the same category.) So, raising the default weight on otherwise unweighted units is the equivalent to lowering all existing weights.

"There are of course 2 solutions:
1. Create non-hacked versions.
2. Mention that not every Quake addition worked with QuakeWorld :)"

Or do like I did -- make 2 separate ai's, one that requires hacked units and one that doesn't. However, the one that doesn't require hacked units really likes it if they're installed. :)
--------
Subject: Re: Artificial Intelligence Theory, part (X+1-Y)*Z

I wiped all but the last replies, it was becoming a big file :)

"You think that's bad -- I've seen occasions where the ai's metal (in storage) went NEGATIVE. You can only imagine the delays before the ai did another thing."

My opinion of the CD AI has just lowered......if that's possible.

"Probably the reason your ai turns into construction fever when you remove the weight ARM/CORE lines is because of the failsafes."
Well I checked just to make sure they weren't "overriding" my weights and were just modifying them, which after a bit of testing seemed to be correct, I kept them as everything works smoothly.

"If you notice, in my ai profiles, ALL the weights are near the top and ALL the limits are towards the bottom" 
However, before an AI can build something, it would need to check that this won't break the limit correct? If this is so, and assuming it can't skip lines, it would need to read through the rest of the weights until it found the limit line. Wouldn't this mean it read other weights on the way down, or do you think it locks out other weights after meeting one it likes?

"(It is impossible for the game to be set to both easy and hard ai difficulty on a single computer"
Shame that, it would be nice to ally yourself with a hard AI, and take on about 8 easy ones, or have 3 medium allies, and together take on 4 hards. Ah well, another AI overlook. Mind you, I've never seen any game do this either, I guess it's a coding issue........or laziness concerning the SP aspect.

"The main reason why my ai's attack force comes in waves has NOTHING to do with unit limits or weights and everything to do with resources."
This of course depends on it's current situation. It *may* do (hypothetically), however, of course it can, and usually is, resources.

"What I ended up doing is deciding whether I wanted kbots or vehicles first, and weighting them the highest"
This is what I have attempted to do. The AI grabs vehicles, and later gets supporting Kbots.

"This is also why I decided my Arm ai needed merls out of the adv. vehicle plant before the other units because the others would be redundant or only a little better than units already in the field"
Hmmmmm. I take a more all round approach. The ARM weight slightly more towards Flash tanks, because that's their strength, but have Samsons too, so as to out range defence. The Flashes then engage enemy forces, keeping them away from the Samsons who batter them from afar. I'm personally not a big ARM Vehicle Lv2 fan, and in standard TA the ARM just bite at Level 2. I wasn't, and still am not very sure about Merls. In the field all a lot of players do, instead of fighting like the AI, is to select their troops and tell them to move to the other side of the AI attack force. This makes the Merls redundent, and even dangerous to their own side. On the other hand, they are great against defence. Many is the time against the TA AI I would curse as my great HLT defence was taken out by rockets I couldn't see. However, this leaves ARM weak up close at level 2, and Goliaths are going to paste Flash tanks.....and pretty much anything else from Level 1 they come across. 

"So I weight the hover plant low"
When I first started on my CC AI I hadn't played a single game with CC. However I am learning more and more about it, and as I do, the Hovercraft plant weighting seems to get lower and lower :) Especially as the AI treats them as part of it's normal ground force.

"Sure, that means it lacks much variation but it means what it's left doing it may do as well or better than an average player" 
I don't like the narrow the AI's choices too much. I can.....and pretty much have.....made it a very good Painted Desert player. However because it was so good on this open map, it became inferior on more closeted maps. Solving one problem often leads to another. I like to let the AI build everything...although I weight things that I deem as less useful than most, pretty damn low, I keep it's options open.......if in probablity terms, restricted.
--------
"Probably the reason your ai turns into construction fever when you remove the weight ARM/CORE lines is because of the failsafes."
Well I checked just to make sure they weren't "overriding" my weights and were just modifying them, which after a bit of testing seemed to be correct, I kept them as everything works smoothly.

"However, before an AI can build something, it would need to check that this won't break the limit correct? If this is so, and assuming it can't skip lines, it would need to read through the rest of the weights until it found the limit line. Wouldn't this mean it read other weights on the way down, or do you think it locks out other weights after meeting one it likes?"

The ai profile is probably not read in a format that'd make sense to you or I. While the map is loading (I don't know which process it occurs in), the ai profile is used to create a binary table of units, limits, and weights. It seems that the order of the ai profile has SOME influence on the table, but you're asking me things that MOST of TA's designers don't know. And I sure as hell don't. The limit check is probably even a SEPARATE check run on the ai profile (be it in tabular form or text file) -- this means the order of the weights and limits need not be related in the LEAST. Heck, I've even seen the ai start new units when that unit limit has ALREADY been reached! (This is a rare occurance though, but I promise you that it happens.) This suggests that the weight check is completed before the limit check is even started -- the ai has already decided what to build, and the max limit didn't talk it out of doing it in time.

"Shame that, it would be nice to ally yourself with a hard AI, and take on about 8 easy ones, or have 3 medium allies, and together take on 4 hards. Ah well, another AI overlook. Mind you, I've never seen any game do this either, I guess it's a coding issue........or laziness concerning the SP aspect."
It's not really a bad issue to me -- because I can overcome that problem on a LAN or in internet games. If the game could handle multiple difficulty settings (one for each ai), it would be more complex codewise and run slower. I think it's slow enough as is.

The main reason why my ai's attack force comes in waves has NOTHING to do with unit limits or weights and everything to do with resources.
"This of course depends on it's current situation."
Presume it's at individual unit limits but not unit max. Ok, factories shut down because they have nothing left they are allowed to produce. The ant trail dries up -- but quite probably some units at the frontlines will die off, causing the factories to open back up. So at worst, the delay is short. More than likely, only one factory TYPE will be hitting unit limits at a time, meaning they'll probably alternate production -- thus minimizing resource drain if in a bind. (And the ai's ussually in a bind. :p) Level 1 vehicle and kbot labs tend to have the highest turnaround rates because their units move fast and die fast too -- good thing they can be replaced pretty fast!

On the other hand, if you DT wall the ai out - it will stack up units behind it for whole screenlengths. And if it's allowed to produce 200+ ground troops it won't quit till it reaches THAT or unit max (500 total units.)

"It *may* do (hypothetically), however, of course it can, and usually is, resources."
May do hypothetically what?

"Hmmmmm. I take a more all round approach. The ARM weight slightly more towards Flash tanks, because that's their strength, but have Samsons too, so as to out range defence. The Flashes then engage enemy forces, keeping them away from the Samsons who abtetr them from afar. I'm personally not a big ARM Vehicle Lv2 fan, and in standard TA the ARM just bite at Level 2. I wasn't, and still am not very sure about Merls. In the field all a lot of players do, instead of fighting like the AI, is to select their troops and tell them to move to the other side of the AI attack force. This makes the Merls redundent, and even dangerous to their own side. On the other hand, they are great against defence. many is the time against the TA AI I would curse as my great HLT defence was taken out by rockets I couldn't see. However, this leaves ARM weak up close at level 2, and Goliaths are going to paste Flash tanks.....and pretty much anything else from Level 1 they come across."

After playing for a LONG time, I have come to the conclusion that Samsons end up being far more useful than Flash tanks. Flash tanks still have their purpose, but I've seen many games where someone built flash tanks first and get raided by a jeffy or 2. (and that's a funny sight, unless it's being done to you...) And other times, Rapiers have shown up to maul the attack force because it had such poor AA. Being an EAGLE player myself, I can say from personal experience, "You can NEVER have too much air defense!" The ai on the other hand can only have too much air defense if AA costs it too much of its resources, but it's still pretty true even for the ai.

"When I first started on my CC AI I hadn't played a single game with CC. However I am learning more and more about it, and as I do, the Overcraft plant weighting seems to get lower and lower :) Especially as the AI treats them as part of it's normal ground force."
Hovers and fast attack ground units are at the mercy of the slowest unit they're traveling with, which ussually negates any benifits they may offer -- except mobility over water.

"I don't like the narrow the AI's choices too much. I can.....and pretty much have.....made it a very good Painted Desert player. However because it was so good on this open map, it became inferior on more closeted maps. Solving one problem often leads to another. I like to let the AI build everything...although I weight things that I deem as less useful than most, pretty damn low, I keep it's options open.......if in probablity terms, restricted."

I see your point, and will be looking into adding a few low limits and weight @ 0.05 on a few semi-useful units I left out of my ai. Maybe they will prove their worth.

Ok, on from ai theory to practice.

Your Queller ai (I only tested the TA:CC version, since it's what I have) has some SERIOUS problems that really made it perform awful when I tested it. For instance, it wouldn't build vehicle plants because you limited them to 0!

If something is abbreviated AAI*, that means it's a HACKED unit made for the ai and does not exist in TA:CC without adding something. Same goes for CAI*. You have AAIVP and CAIVP limited to 2 -- except I'm not using hacked units to test this ai!

I have never seen the ai use minelayers in an effective manner. And I've only VERY rarely seen players use minelayers in even a semi-effective manner against other players.

As mentioned earlier, having the ai build moho mines on a nonmetal world is pretty much suicide for the ai. What's more, having the ai build only a few solars and metal extractors is just as bad - because that's where the ai gets most of its resources, at least early on. Variety isn't bad, but an ai starving for resources is. Your ai is set up to build a LOT of big ticket items (like berthas, annihilators, fusions, a vulcan, guardians, HLT's, LLT's, flakkers, and ambushers) but will no doubt have only a fraction of the resources needed to do that without nanostall. Perhaps it would be more effective with fewer LLT's? I'm not saying you need to remove all the different attack unit types - I just think you should limit them to something reasonable. I don't remember the last time I ever got a ground attack force larger than 100 units, yet you have your ai set to make a ground attack force of over 800 units per side. (I did the math ;)

All said, even with the strikes against it your ai did rather well versis mine -- it even won 2 games out of 6 on Painted Desert at 1k/1k settings. But that means it might be a fluke, and even if it isn't I'm still working on my ai too. ;)

David
--------
Subject: FSilo

In your counter strike UFO files, there are pictures and animation's and such for "Fake" silos. FSilo, FEMP etc. What do these do? I don't recall any fake silos in TA, did you add them? If so, why? Was it to make the AI's real silos harder for a player to target? Or are these files needed in case a player captures an AI silo? 
--------
I added fake nuke silos -- the player can build them too. They fire the missile frigate "merl" rockets -- so they're not just useless buildings.

We need to get on boneyards at the same time and test ai's head to head. What's your nick there and about what time are you typically on there?
--------
From: "Quitch" <quitch@fickle.demon.co.uk>
Reply-To: "Quitch" <quitch@fickle.demon.co.uk>
To: "David Sellick" <switeck@hotmail.com>
Subject: Re: FSilo
Date: Tue, 17 Aug 1999 18:53:08 +0100

I'm going to skip the usual e-mail order, as I think the AI on AI games are more important. You say I weighted VP to 0, obviously because I wanted it to use the hacked plants. Didn't I send you the plants too? If that is so, then your tests are somewhat bias. Attached is an improved version, I have been tidying it up, changing the layout, halving most of the limits etc. In my opinion it plays a good game. it is a little unpredictable with it's construction units, but is almost done. I should have sent you the Queller AI ZIP, and it's self-contained, it has *everything* the AI needs.

Please remember though that in an AI on AI game, there is a major factor that can render such tests useless. Range. An AI with all missile troops will 9 out of 10 times hammer one that uses only close range troops. However, against a human, the one with close range troops is more likely to win that the one will missile troops because the human will sit in the middle of them. My AI could be made to fight other AI's as a speciality, however it isn't, it is for fighting humans. It is in this light that I view any results, even if my AI wins. I guess I'll have to do my own tests though, to compete :)

"Heck, I've even seen the ai start new units when that unit limit has ALREADY been reached"
IMHO, from what I've seen, it seems to only occur if the unit is built as the very *next* unit, and if the choice is made quicker than the limit is updated, which can occur if the plant moves straight from finishing one unit, to beginning another.

"On the other hand, if you DT wall the ai" 
I simply try to discourage the use of DT's against the AI. If a human builds DT's you need to add more shell weapons, if you add more shell weapons the human doesn't build DT's, and pastes the shellers in the open field. Remove the shellers and the human uses DT's again and frags you with defence. Try to hit a balance and he will smash part of your force in open combat, with dedicated troops, and finish you off with defences. I just tell people if they want a tough game, don't use DT.

"which usually negates any benefits they may offer"
You forget that Hovercraft have a longer range than vehicles, they also reach any current front faster, and they do more damage.

"All said, even with the strikes against it your ai did rather well verses mine -- it even won 2 games out of 6" 
I judge that as a pants performance. For me to be even remotely happy I would deem that it should have won at least half, and probably more. In actual fact, as I said earlier about AI on AI games, my AI may actually do better against Mostly Harmless (I presume you used that AI), than my current version. version 4 CORE I think it was, used nearly all Slashers, and against another AI would be deadly. However, Version 9 BETA (:)) uses things such as PeeWee, which while useful against a human due to their limited vision range, are next to useless against another AI. I noticed that Mostly Harmless CORE Kbots are nearly all Storms, rocket troops.

I use the online name of Quitch, I use Quitch everywhere and as of yet have never found another........even on ICQ. The only reason I'm QUitchy here is because when I applied for my forum password, I changed e-mail address, so never recieved it. Therefore I had to sign up under a new name, hence why the Y is in brackets.

My connection to Boneyards is only about 50ms faster than a direct IP. This because BY has no European servers (I'm English). I don't know if BY has a lack of bandwidth as well.
--------
"You say I waited VP to 0,"
Don't you mean you weighted VP to 0? I waited a long time to see it build VP's and it didn't!

"Didn't I send you the plants too?"
Whether you did or you didn't, I didn't use them but I have them now. And know to use them as well w/ your ai.

"I wanted it to use the hacked plants."
I thought I was running a nonhacked ai, DOH!

I think you put normal jammer limits too high. The whole reason why Creeperman made hacked jammers in the first place was they would clog up the factory they came out of because they didn't get far enough away from the exit. And you have each kind's limit set to 4. Factor in the overbuild bug, a cramped base, and any terrain restrictions and at least one factory is certain to end up being jammed -- and not in the way you intended. :p I limited them to 1 in my ai profiles -- adding more won't increase the jammed area except marginally. And if a bertha/nuke/whatever nails it, well the player PROBABLY knew to shoot there in the first place. Mobile jammers on the other hand can be Limited as high as 4, (for very large maps like Yerrot Mountains) but 2 or 3 is more reasonable. They try to move quickly to just behind the ai's attack force, and unlikely to be jamming any units traveling to the frontlines. And if the ai's got 4+ mobile jammers traveling with its attack force, it'll have slightly increased pathfinding problems (particularly on Greenhaven) and will have spent more metal on a unit that neither provides resources nor has the ability to damage its opponent/s. Once the ai's attack force starts shooting, all the jammers hide is its numbers.

"IMHO, from what I've seen, it seems to only occur if the unit is built as the very *next* unit"
It's still highly likely that the programming to check for weights is separate from the limits check programming -- obviously one CAN be done in disregard to the other, even if only in a narrow case.

"I simply try to discourage the use of DT's against the AI. If a human builds DT's you need to add more shell weapons, if you add more shell weapons the human doesn't build DT's, and pastes the shellers in the open field. Remove the shellers and the human uses DT's again and frags you with defence. Try to hit a balance and he will smash part of your force in open combat, with dedicated troops, and finish you off with defences."
I disagree with your assumptions. Ballistic units, at least in the level 1 class, tend to be more expensive than the direct LOS units (such as flash tanks and instigators) -- they often take a bit longer to build, giving the player MORE time to DT in. If the player's intent is to wall the ai in, it'll be easier to do if the ai spends most of its time making ballistic units.

"I just tell people if they want a tough game, don't use DT."
As do I. I ussually allow DT around fixed defenses (and recommend it!) -- but not to be used as a fence that prevents the ai ANY way to reach the player's base. Killzones kick ass too much to disallow them entirely. :)

"You forget that Hovercraft have a longer range than vehicles, they also reach any current front faster, and they do more damage."
I did not forget that hovers have slightly longer range than SOME vehicles at a greater cost on average and less armor to boot. The merl hovers do LESS damage than merls, the "tank" hovers have no real equivalent, (for what it's worth, they don't fill a nitch either) and the missile hovers do more damage and have greater range than a Samson, but have less range than a missile tower. Unfortunately, they have about half the armor of a Samson, making their combat survivability near nil versis players and not too good versis ai's that rely mostly on ranged weapons.

"I judge that as a pants performance."
pants?!? I am unfamiliar with the term... is it British slang? And does "pants" mean good, mediocre, or bad?

"For me to be even remotely happy I would deem that it should have won at least half, and probably more."
I've been working on ai's longer than you have -- I should have a slight home-court advantage. :p I was testing a modified version of your ai -- because it was lacking any hacked units. The lack of vehicle plants even seemed to work to its advantage because it spent its extra metal making more kbots to attack with! I was testing only my Arm vs your Core in the 6 skirmish games... It won the first 2, the next 4 just verified (in my mind) that I had changed my ai enough that it wasn't as likely to happen again. :) In the last 2 tests, I had also changed YOUR ai some more to improve its odds versis my ai -- ie: reduced LLT's to 0, altered the attack force composition to what I know is both effective in the ai's hands AND is good against my ai. (It's what *I'D* use to counter it...)

"my AI may actually do better against Mostly Harmless (I presume you used that AI)"
You can bet your ai didn't beat Counterstrike 2 times out of 6. :p Counterstrike sometimes gets the jump on BAI (set to hard) and beats it. Your ai might use Counterstrike's special hacked units, but without having them weighted appropriately (and nonhacked equivalents weighted extremely low or removed) your ai is at an exteme disadvantage. As Mephisto (a guy I worked with on ai's, maps, and units) said - an ai that's hacked can easily break almost all the rules of the game.

"However, Version 9 BETA (:)) uses things such as PeeWee,"
I thought I was testing your latest ai at the time?!?! For sanity's sake, could you put a version number line IN the ai profile (at the bottom even, where it should affect the ai the least. And could you start putting units of the same kind in alphabetical order (use mine as a template if you gotta -- I think you left out a unit or two as well from yours ;) and make it easier to test one side of you ai versis one side of mine in skirmish? As it is, I have to do about 30 cut-and-pastes to make an ai vs ai profile.

"I noticed that Mostly Harmless CORE Kbots are nearly all Storms, rocket troops."
I'm a good enough Core player to tell you that Core's in-close level 1 ground troops SUCK. AK's = weasal's weapon, peewee's armor. Instigator = was this supposed to be a flash counter? (it isn't!) Raider = ok, it's as good as a Stumpy -- which isn't all that great. Thud = ok, he's a good unit - but he'd better have LOTS of friends or be in very hilly terrain. :) Leveler = it's a great unit in a player's hands who knows to put it behind something tough, but the ai doesn't do this. This leaves only Core Storms, Crashers, and Slashers out of all level 1 kbots and vehicles. I added some Thuds for indirect fire ability (and as damage shields) -- ussually that ability isn't needed early on since even good players take a tiny amount of time to put up a good DT wall.
 
"My connection to Boneyards is only about 50ms faster than a direct IP. This because BY has no European servers (I'm English). I don't know if BY has a lack of bandwidth as well."
BY has only one sever - on the west coast of USA, I'm much closer to the east coast. BY also has internal overhead which generates extra lag. So direct IP games from me to you is probably better than BY, especially if we use my bugfix. BUT... it'll be easier to find you online using BY. Once found, we can swap IP numbers and play via IP -- I did that many times using TEN. Pingtimes that way were typically 200 less, BY is probably going to show just as big a drop. 
--------
Subject: Re: Ai practice
Date: Wed, 18 Aug 1999 20:08:01 +0100

"I think you put normal Jammer limits too high"
Nah :) Your argument only assumed that there was one plant, however my limits work around the fact that the AI can have two. Also, the Jammers are weighted low enough, that it will have max plants before max Jammers. All the plants have 2 exits, and this allows them to support Jammers. Sure, more than one Jammer per plant doesn't help coverage, but setting a limit to 1 only covers a single plant, plants are very rarely all in one space, especially on maps like Painted Desert, when they are leagues apart. Allowing it to have 4 Jammers for 2 plants dramatically increases the chances that both plants will have coverage. Allowing a maximum of 4 makes for best coverage, with least chance of blockages. The odd extra Jammer is useful, because players will often do a quick raid to hit the Jammer, so they can Bertha away. Having a spare Jammer (usually on the other side of the plant) makes this kind of thing harder to do. 
I agree that mobile Jammers need be limited around about 2. More is a waste.

"Once the ai's attack force starts shooting, all the Jammers hide is its numbers." 
And stop guardians trying to cripple the force every time it has a break, as they won't be able to see much of it. However, again, it's that kind of thing that hampers it in an AI on AI game, because the Jammers are rendered useless.

"As do I. I usually allow DT around fixed defences (and recommend it!) -- but not to be used as a fence that prevents the AI ANY way to reach the player's base. Killzones kick ass too much to disallow them entirely"
I discourage it for a harder game, simply because the AI is too stupid to adept, or ignore. However, I have added a few more artillery types to deal with wreckage.

"pants?!?"
Sorry, I forgot. It means terrible, or crap is probably closer to it.

"I've been working on ai's longer than you have -- I should have a slight home-court advantage. :p"
However, as you said you were lazy I assume you follow a slightly BSR path? he works on his AI once a week. I assume that you have bouts of creativity, then leave it alone again? I have been working on this AI for the past month and a half everyday. Sometimes it's small changes, sometimes it's the layout, some times I have to add something. It's fun. Though you may have been working on AI longer, mentally I was working on it since day 1. Every time I played a Skirmish against the old AI, I would think "I wouldn't do that", or "Urg, the AI should have built more of this" etc. With the cost of online gaming over here, you tend to look harder at the AI than most :)

"I thought I was testing your latest AI at the time?!?! For sanity's sake, could you put a version number line IN the AI profile" 
I believe I sent you the WHOLE AI, and this shows you didn't read all that beautiful documentation. The version number is in the AI Update file. You tested an early Version 9, the one I just sent you is a later part of version 9 :) That's the problem, by the time you get the AI, it's already out of date.

"As it is, I have to do about 30 cut-and-pastes to make an AI Vs AI profile."
I believe you have a LAN, no? It isn't made for quaint AI on AI testing, it was for human games. I did originally have it ARM stuff then CORE stuff, however what seemed to happen is all the ARM Commanders would get underway about 4 seconds before CORE. I'm not having that. The AI files to me make perfect sense, may I point out that to me your Mostly Harmless AI layout is enough to induce nightmares. I will be running a LAN test of Ai's tomorrow..........maybe not, I get my exam results, and I may not be in a fit state of mind.

"This leaves only Core Storms, Crashers, and Slashers"
I'd weight down Crashers, they take a while to build which means if the Kbot lab is the first lab then the AI gets stomped by a fast attack. I worry a lot about the beginning of the game, after all the whole idea originally was to have an AI that could withstand early assaults.

"Swap IP addresses"
My IP is 193.237.119.112 It's fixed. I guess BY will expand to another couple of servers. I would have thought that the success of Kingdoms, and it's pants AI, would drive more people onto BY.
Have you ever considered EGN? You get it from www.egnus.com
It looks a bit like ICQ, but it is for games. When someone on your list enters a game, the games symbol appears next to their name, and if you double click on it, it hunts down the IP, and you can join them. It is excellent for games like Tribes as you can hop in with your friends. If you had a lot of TA friends, you could do the same. 
If you ever look for me anyway, as I said, I'm *always* Quitch, in Tribes (only in game and in chat rooms) I'm XXX_Quitch.
--------
I still think you put normal Jammer limits too high.
"Your argument only assumed that there was one plant, however my limits work around the fact that the AI can have two."
My argument said nothing about factory numbers -- I actually assumed there'd be 2 plants.

"Also, the Jammers are weighted low enough, that it will have max plants before max Jammers."
Probably... To me, a weight of 1 or greater is pretty high.

"All the plants have 2 exits, and this allows them to support Jammers."
In a crowded base -- or one w/ lots of trees/rocks/hills/water around it -- only one exit will probably be viable to begin with.

"Sure, more than one Jammer per plant doesn't help coverage, but setting a limit to 1 only covers a single plant,"
Overbuild bug... Also, there is one jammer of 2 kinds so there should be 2 jammers total -- and one type will probably have 2 jammers even though limited to 1. Ideally, 2 or 3 jammers of each type would be sufficient to cloak much of an ai's base from radar coverage. But even with the limit set to 10, if the player had total radar coverage they'd probably see many parts of the ai's base which lacked radar jamming. I set mine to limit 1 because I don't consider base jamming very valuable to the ai. Its attack force needs jamming more, because it's more likely to be inside an enemy's radar range.

"plants are very rarely all in one space, especially on maps like Painted Desert, when they are leagues apart."
On a map like Painted Desert, advanced factories may well be far apart -- but unless the ai's opponent very agressively builds radars all over the map the jammed factories will all be outside radar range. Ussually, I find outlying mini-bases using direct aircraft flybys. Peepers are excellent for this, and render fixed jammers a waste of both resources AND time.

"Allowing it to have 4 Jammers for 2 plants dramatically increases the chances that both plants will have coverage. Allowing a maximum of 4 makes for best coverage, with least chance of blockages."
With 2 factories making a total of 4 jammers between the 2, there's a reasonable chance that one factory makes 3 and the other only 1 -- the factory that made 3 will probably be jammed. Don't forget that the overbuild bug could occur and 5 (or more) total jammers end up being built (so 1 factory will probably make 3, the other 2.) If there were more than 2 factories built, that would be less likely -- but if more than 2 factories were allowed they would be much more likely to be near one another and thus have far less need of redundant jammers. I have seen quite a few cases where even 1 or 2 jammers were enough to stuff a factory - even on maps like Painted Desert. The worst example of this is if the base is extremely congested when the jammer is finished, the jammer may remain where it was built.

"The odd extra Jammer is useful, because players will often do a quick raid to hit the Jammer, so they can Bertha away. Having a spare Jammer (usually on the other side of the plant) makes this kind of thing harder to do."
But... if the player knows to bertha the area to begin with, an extra jammer or 2 on the ai's part will be an annoyance at MOST -- and will give a bertha a higher probability of hitting something. I've had many games where I knew the area the enemy player's base was in and used random hits by my bertha to "train" it to vetran status -- I had scouted the area only once or twice and either didn't have any radar coverage over the area or it was jammed, (or both) yet my bertha kills started racking up. If that player had jammers in their base, it would've only increased my bertha's rate of fire and damage quicker (that's what getting kills does for units.)

"I agree that mobile Jammers need be limited around about 2. More is a waste."
Considering there's 2 types, and probably one will overbuild by 1, you'd end up with 5 jammers at a time. On a really big map and a high attack force turnaround (like when its attacking a base), every jammer will be jamming something different.

Once the ai's attack force starts shooting, all the Jammers hide is its numbers.
"And stop guardians trying to cripple the force every time it has a break, as they won't be able to see much of it."
That presumes radar-targetting -- which doesn't work nearly as well as sending in peepers as spotters. You do use peepers, don't you? :p And as you said, many players like to mix it up with the ai by driving their attack force right into the ai's attack force -- in this case, jammers would just be a nonshooting enemy. When my base is under attack, I like to give the enemy attack force as many distractions as possible while my guardians do much of the work. Any jammers I spot will probably be one of the last things I kill. So mobile jammers in an ai versis player game hides its attack force only until it fires (which is extemely valuable) - after that it hides only its numbers. (Which is a useful, but a short-lived effect.) And that's only until the player decides to counterattack. After that, the jammers are of little value at all.

"However, again, it's that kind of thing that hampers it in an AI on AI game, because the Jammers are rendered useless."
In an AI vs AI game, jammers are totally useless as you've said.

"As do I. I usually allow DT around fixed defences (and recommend it!) -- but not to be used as a fence that prevents the AI ANY way to reach the player's base. Killzones kick ass too much to disallow them entirely" 
I discourage it for a harder game, simply because the AI is too stupid to adept, or ignore. However, I have added a few more artillery types to deal with wreckage.

"pants?!?" 
Sorry, I forgot. It means terrible, or crap is probably closer to it.

"However, as you said you were lazy I assume you follow a slightly BSR path? he works on his AI once a week."
BSR is not great at making ai's for TA. You're already better than him at that. His skill lies in making 3rd party units, especially for doing the scripts for them. BAI is proof of that, BSR used his unit skill to overcome his ai's (BSRAI's) weaknesses. Although my nonhacked ai may not seem all that great, I consider it almost as good as possible considering the tools I have to work with. This is at least true for default, airbattle, and seabattle -- some of the other ai profile types are so oddball (and I often hate their maptypes as well) that I haven't tweaked them to nearly the same degree.

"Though you may have been working on AI longer, mentally I was working on it since day 1."
I have spent roughly 10-20 hours a week for the last year and a half working on my ai. That computes to ~100 8-hour workdays worth of work on the ai. My ai tweaking guide, and stuff I've told you is the results of all that labor -- which you are using. I have been playing TA since early Nov 97. My first ai profile is dated Dec. 1, 97. I emailed Cavedog a buglist (which included many ai problems!) mid-January 98. Your time spent playing TA may have me beat -- but only by a handfull of days. It is a minor insult to me for you to say that my ai should be something you can easily better unless you're already a VERY good TA player. Considering you just got TA:CC recently, that is possible but unlikely.


"I believe I sent you the WHOLE AI, and this shows you didn't read all that beautiful documentation."
You assume too much about me, it's embarrassing. :p

"The version number is in the AI Update file. You tested an early Version 9, the one I just sent you is a later part of version 9 :) That's the problem, by the time you get the AI, it's already out of date."
Correct, I knew it was version 9 -- but not an "early" version 9. Thus, some confusion occured when you told me your version 9 did better! My ai profiles all have (or at least they should ;) an internal version number in them. Only the minutest of changes get by without a version number change of +0.1. In my Old_ai.zip backup of all my ai profiles, if I find 2 different ai profiles with the same version number one gets the "a" designation the other gets a "b" designation. I think I have a v8.7g -- if that means anything to you. Major changes get bigger jumps in version number - ussually to the next whole number. That's why my default ai is now at v12.1, even though it's only seen about 5 major revisions.

"I believe you have a LAN, no? It isn't made for quaint AI on AI testing, it was for human games."
We don't play TA much anymore, and when we did we played versis Counterstrike ai because any nonhacked ai was a joke. Kenneth Thomas and I (the 2 best players) can make a joke out of 3 counterstrike ai's vs us 2. It is unlikely that your ai would fare well...

"I did originally have it ARM stuff then CORE stuff, however what seemed to happen is all the ARM Commanders would get underway about 4 seconds before CORE. I'm not having that."
That's probably because the CORE weights were halfway or more down in the ai profile (or the weight ARM/CORE 0.1 lines causing havoc) -- but my ai profiles now don't do that, instead the weights of one side are followed immediately by the other side's weights. Having limits buried at the bottom of an ai profile seem to have little effect on it, at least so long as an ai profile is under 20kb.

"The AI files to me make perfect sense, may I point out that to me your Mostly Harmless AI layout is enough to induce nightmares."
Your ai already shares a very similar layout to mine: weights at top, limits at bottom. So I don't see mine as looking all that different than yours. I long ago gave up trying to make 6 separate ai's in each ai profile -- thus my ai's profiles' easy/medium/hard settings are only variations on factories, c-units, and superweapons. I see little point in changing the ai considerably just so it plays BADLY. :p

BAI for UH on the other hand (and BAI v5.0 by Creeperman) have very long ai profiles that reflect vastly different ai strategies depending on skill setting.

Your ai profiles are neither in alphabetical order nor build menu order nor in order based on what an individual constructor can produce. (level 1-3 structures are only separated into resources and weapons.) The only reason mine has anything broken up and placed at different locations is because they vary with skill setting or because they're hacked units that aren't always available. The ai Blank.txt file has documentation for the ai layout I use. It may not seem as straightforward as BSR's BSRAI, but I discovered that sorting by level and then sorting by type made more sense than sorting only by type. The weights on groups of units built by the same constructor tend to only affect one another, NOT weights on dissimilar units. Thus the weight of a fusion has no effect on the weight of a solar.

As may be noted in some of my ai profiles, I weight some units to 0. This is because some of my tests seem to indicated that even units that are unbuildable (because of a limit of 0) still influence other unit's build weight ratios.

"I will be running a LAN test of Ai's tomorrow..........maybe not, I get my exam results, and I may not be in a fit state of mind."
My most stressful time is during exams. Afterwards, I am ussually relieved of most of the stress regardless of my grades. After many years of college, I have aquired an "I'm done with it, fuckit." attitude. Once the class is over, and I passed, I don't stress myself out over it -- it even makes me better able to remember what I learned in the class. Exam block from overstudying (studying 6+ hours the day before) is both much more stressful and much more damaging. The really hard classes tend to give lots of homework - if you do that and study for the class each day you should easily pass all but the most evil of classes (where they put a lot of stuff on the final that wasn't mentioned in class in advance.)

"I'd weight down Crashers, they take a while to build which means if the Kbot lab is the first lab then the AI gets stomped by a fast attack."
Thuds build even slower, and Crashers are good anti-jeffy units. Crashers are weighted pretty low in my ai -- it loves Storms FAR more. :)

"I worry a lot about the beginning of the game, after all the whole idea originally was to have an AI that could withstand early assaults."
There's only SO much an ai can do versis a good early raid by players. Even BAI on hard isn't immune to them. Creeperman said it was "easy" to totally cripple Counterstrike with an early raid. As a player, ussually what defeats me is someone cripples me with a good early raid/rush. If hacked ai's have trouble, nonhacked ai's have to use almost self-defeating tactics to give it a chance against early attacks.

"My IP is 193.237.119.112 It's fixed."
Mine's dynamic -- it (last I looked) was 206.27.93.*** -- with the 3 *'s being anything from 0-255. It seems my ISP's server uses a truely random assignment on it. I will be on BY off and on today from 11 am to 3 pm my time -- which is 5 pm to 9 pm GMT.

"It is excellent for games like Tribes as you can hop in with your friends."
I don't play Tribes -- it's not my style, I like slower, more strategy-based games for the internet. Fast-action games are best suited for LANs and modem to modem connections.

"If you had a lot of TA friends, you could do the same."
I don't.

"If you ever look for me anyway, as I said, I'm *always* Quitch, in Tribes (only in game and in chat rooms) I'm XXX_Quitch."
I go by Switeck pretty consistantly now. But if I'm ever on as "somebody else" I go by David Sellick. In games, I use my nick - when dealing with companies, tech support, and message forums I often use my real name.

"It looks a bit like ICQ, but it is for games."
I hated ICQ's interface. It is NICE to know who's online, but it is terrible for how it connects to them. Reminds me of playing IP games over AOL -- 5+ second lag. I'd send a message, and it'd take 10 seconds just for me to get an "OK" message back. (This is from someone who's a faster typist than I am...)

All that aside, your ai seems LIKELY with minor changes (that even I am unsure about) to beat my ai on most land maps. Default map ai's are probably the easiest to make decent. On the other hand, a good seabattle ai on a map with either resources out of the norm or something else odd can be a real bitch. Getting weights to something that results in a good ai is an art, the limits are more of a science -- anyone could get "lucky" but you're using a method that makes that far more likely than anyone else I know.

David
--------
Many times, on the smaller maps, jammers deterred me not in the least from using a bertha to maximum effect, because scouting was easy. On the larger maps, I almost never have radar coverage that includes the enemy's base -- so jamming a nonexistant radar was a moot point.
--------
Subject: Re: Ai practice
Date: Fri, 20 Aug 1999 20:17:59 +0100

I won't quote you on Jammers, the argument breaks down too much (into sections, not as an argument). 
Basically, I believe that the gain from Jammers outweigh the risks. 50% of the time the AI will park the Jammer away form the factory anyway. As for it not moving off the pad, that is something any unit can do. In the old Vehicle Pads (before I made my own) Slashers seemed to have made an art out of staying on the pad. An AI, as you pointed out in an earlier e-mail, concentrates it's defence around the Commander. This makes it very hard to assault as wreckage blocks units. As the defence increases people use more units, thus the pathfinding becomes worse, and more units die. In the end, I assure you, 9 out of 10 players will build long range hitters, and start to smash the entire base that way. 
Since the AI has no style :) in the way he plays, other than his ever subtle "Hammer Blow" attack, he needs every advantage he can get. If one of those advantages is coming from a direction the player though held nothing, than I hold that to be good. It is rare for an AI to block his plants with Jammers, I've not seen mine do it yet. 
I admit that Peepers have their advantage over radar, however as most AI attack forces use the advantage the AI has when using missile troops, most of the AI attack troops double also as anti-air, as do it's defences. 
Also, if the player closes in with his units, getting LOS, again, he won't be able it fire long range weapons because he'll risk hitting his own troops. 
Talking of factories I still need to allow my AI VP's to be used when captured. 
May I ask, your fake ARM silo was labelled armfsilo in the files was it not? I still have a grey box where the picture should go, and I am trying to track down the rouge file that in the UFO that is telling the menu the picture exists.

"This is at least true for default, airbattle, and seabattle -- some of the other ai profile types are so oddball (and I often hate their maptypes as well) that I haven't tweaked them to nearly the same degree." 
I spent absolutely ages testing default. About 7 Versions worth I believe. Only when I considered it great did I move on. Because I felt it had been so thoroughly tested I used it, as I had intended, as a base for the other AI's. Some require more or less a fresh start, such as ThePass, but others, like Wind, just need a very slight adjustment. Mind you, since then, I've still fiddled with default despite meaning to leave it alone :)

"Although my nonhacked ai may not seem all that great, I consider it almost as good as possible considering the tools I 
have to work with"
I consider Mostly Harmless the best non-cheating AI out there. I spent weeks tracking it down for a reason :) Of course, I make sure my testers say mine is the best :) Advertising you understand. Heh heh.

"It is a minor insult to me for you to say that my ai should be something you can easily better unless you're already a VERY good TA player" 
Already? I was playing TA from close to day 1. Those PeeWees got me hooked :) I couldn't evaluate exactly how good I am as BY is too slow for me really, however I raid, attack from multiple sides, use the range of units, scout. I believe I would qualify as an above average player. I just never played against TA:CC players, and there are extremely good players without CC out there. The problem is, when I play AI's, I get lazy :) I attack in one group, slogging it out with the AI's one group. Mind you, I find it makes the games more entertaining as it makes up for the AI's bad points by not exploiting them. 
I apologise for causing offence. However, I did consider your AI to build construction too early, and that a quick hit would take it out. However, your AI does get going quickly after that. It is to this end that I have been playing with the construction weightings, I'm trying to get them inside 8 minutes, without stopping the AI being ready to prevent raiders by the 4th. Also, while they sometimes help the AI by expanding and grabbing resources, others they cripple it with 4 simultaneous Sentinels. I'm still playing with them. However I do believe that currently my AI waits a tad (slightly) too long to get them.

"You assume too much about me, it's embarrassing. :p"
Meaning?

"Major changes get bigger jumps in version number - usually to the next whole number. That's why my default ai is now at v12.1, even though it's only seen about 5 major revisions."
This was my original intent. I may do it when I release the AI, but when it came to writing it, I wasn't sure which value should change. had I just made a minor change, or was the knock on effect big? In the end, I plumped for a single number system to keep things simple. Anyway, I only release to a select group of people, and only after lots of changes.

"It is unlikely that your ai would fare well..."
YOU UNDERESTIMATE QUELLER!!!! Well, no :) You're right, but the more humans (in a team) involved, the less chance an AI has. An AI goes for the nearest target, where as humans team up and destroy the AI's in concentrated combined assaults. 
My LAN games are always exciting, because my partner is my 7 year old brother :) He doesn't play a very aggressive game, though he does like standing his Commander on a hill and D-Gunning everything to hell. Actually we lost our last game, he sent his attack force to aid me, leaving has base undefended. He got killed, leaving me to face 2 AI's. I may have done it, except the 2nd AI rather than assault my force (as I thought it would do), struck behind me, catching my behind the lines vehicle factory unready. My force, attacked from both sides, was trashed. Oh dear :)

"Afterwards, I am usually relieved of most of the stress regardless of my grades. After many years of college, I have acquired an "I'm done with it, fuckit.""
Ah, a fellow spirit :) Now I'm moving onto another level of education, I like to take a relaxed approach to these things. Always look to the future.

"instead the weights of one side are followed immediately by the other side's weights"
That's exactly what mine do.

"Your ai profiles are neither in alphabetical order nor build menu order nor in order based on what an individual constructor can produce."
Neither are yours, Vehicle Plant is about your 4th item, and I don't remember geothermal coming before it, after all your first construction unit is a Commander, and he can't build Geothermal.

"I see little point in changing the ai considerably just so it plays BADLY. :p"
I was tempted not to bother with other levels at all, it's hard enough to make an AI tough. However, my brother plays medium, so I do it for him :) For Easy I just weighted everything to 1, I think that'll work :)

"but I discovered that sorting by level and then sorting by type made more sense than sorting only by type. The weights on groups of units built by the same constructor tend to only affect one another, NOT weights on dissimilar units"
I don't know if you have my latest version, but you'll notice that everything that is constructed by a plant is at the top half, and units take up the bottom half. However, when I go in and modify my AI it is generally with a set task in mind, and the set tasks usually break down into the groups I have. More resources needed, less plants needed etc.

"Thuds build even slower"
However they take more damage than Crashers, and are probably a better allocation of funds as they introduce a new element. Crashers are weaker Storms, with a longer range. Thuds on the other hand introduce a more powerful weapon, tougher armour, and best of all, their terrain ability. A Thud will keep a player busier than a Crasher will. Seeing shells land among your troops usually scares people into action.

"in my ai -- it loves Storms FAR more. :)"
If I was a a TA unit, I'd be a Storm. They look so cool :) Well, maybe I'd be a PeeWee :)

"As a player, usually what defeats me is someone cripples me with a good early raid/rush. If hacked AI's have trouble, nonhacked AI's have to use almost self-defeating tactics to give it a chance against early attacks."
In a 1K/1K game, getting a raid to the enemy base before the 4 min mark on a non-metal map is near impossible. There is one solution, Comm-boosting. I have written in my ever growing "How to make the AI more difficult" section, that players should not boost production, simply because there is nothing the AI can do to match that kind of production.

"Fast-action games are best suited for LANs and modem to modem connections."
I'm a team obsessive (only I could find room on Ashap Plateau for 2 on 2 team games). Tribes is great for team work, and coded for the Internet from day 1. It is incredibly smooth, and there are always about 600-900 servers active at one time.

"It is NICE to know who's online"
It was great, suddenly realising the Internet was a HUGE place, and not a cold wilderness :). Actually, I always wanted something that would tell me exactly how many people were on the same page as me. Whenever I go to a sigh, I just have to wonder how many people are look at the exact same page I am.

"I will be on BY off and on today from 11 am to 3 pm my time -- which is 5 pm to 9 pm GMT."
Weekends only for me, that's the annoyance of local call costs.

"All that aside, your ai seems LIKELY with minor changes (that even I am unsure about) to beat my ai on most land maps"
Today, AI's, tomorrow.........the PLAYERS!!!!!!!! :) Must admit, I still love your AI. Actually, I think I'll start mixing. As me and my brother are 2 different people (no really! :)), maybe your AI and mine can ally against me and my brother :) 
My AI still has a way to go, I've got to balance construction against attacking slightly better.

"Default map AI's are probably the easiest to make decent"
I'd say Metal is easier. Rather than having to think about how the AI wandering all over the map to metal spots will effect you, you know he will build them like solars. It also means your profile adepts better as the metal spread is always the same on a metal map. The biggest problem I find, is some metal maps are near flat, others have a few hundred metal things sticking up. Times like this I wish for a MetalFlat and MetalHilly profile. I like sub profiles, I've already made a WindAcid (I had to change your Wind on Acid Pools to WindAcid simply because my normal Wind tries to build a shipyard in the Acid), and a WindMetal profile.

Recently (oh no, another story!) I played on Metal Heck 2. Myself and my ARM ally against the CORE foe. We lost (no, I lost to my own AI!). I took too long to smash my opposite number, and had to use Merls to dig him out. By the time I did, the other CORE had an airforce that smash any attack I sent out (by the way, why did you halve all the maxdamage values of your AI aircraft?). Eventually he crushed my partner, and then me. I couldn't believe it when I saw a Freaker coming my way. They are weighted at 0.05, so seeing one of them was *not* a good sign :) Everytime I tried to make a move towards his base he bombed my buildings, and my attack force was trying to contain the other CORE. Just made a mess of that then :) It's fun to lose.

Tried the special Skirmish maps from TAMEC? OK, so most of them were made taking into account the fact that the AI was thick, and trying to play 9 Quellers or Mostly Harmless's on them is probably insane, but they're worth a look if you fancy an AI challenge. Anyway, I like Tight Spot, I'm going to play a FFA on it soon :) Problem is I spend so long watching AI on AI battles that I hate to drag myself back to playing myself, and when I do, Normal seems so slow now as all battle I watch are +10.

"Getting weights to something that results in a good ai is an art, the limits are more of a science -- anyone could get "lucky" but you're using a method that makes that far more likely than anyone else I know."

Well if I'm using a method that lets me get lucky more often, I'll start playing the lottery again :) So how do get lucky so easily? Updating every 5 seconds? I generally limit something 10X more than it was weighted, or did, I halved them all recently. 
Well, I'm back to glorious AI testing to get construction just right. Hurrah! I really should get round to making a webpage, I promised the testers a message board too.
--------
"Basically, I believe that the gain from Jammers outweigh the risks."
Hey, apparently since I added them back to my ai I believe that too -- I just am far more conservative in their useage than you.

"As the defence increases people use more units, thus the pathfinding becomes worse, and more units die."
I'd like to note that the ai routines also get poorer - both in the pathfinding *AND* in the build choices (pauses between building gets longer) as unit count goes up. (This doesn't explain the stupid commander bug where the ai's comm stands around before doing anything -- that's probably a too-long ai profile, an exe bug, or the results of an overtaxed cpu -- pause and slow down gamespeed and see if that helps.) Once at unit max, the ai is still making choices on what to build and "throwing away" the ideas when it finds out it can't. So if and when it can build units again (because it's back below max), it could be anything.

"Since the AI has no style ... needs every advantage he can get."
My main complaint about ai nonmobile jammers is not so much their chance of blocking a factory -- which I have exaggerated the likelyhood (except on a map like Pincushion) -- it's their cost. The ai is almost always playing catch-up with a good player on resources and effective fighting ability (ie: a bigger attack force/stronger base.) The EARLIEST I've ever seen my ai reach what I call the "happy" point -- where it has multiple fusions and moho metal makers capable of sustaining almost any build project -- is just under 30 minutes (on Painted Desert, 1k/1k start.) And that was a fluke -- the ai's comm assisted the first fusion and suffered a short but brutal (not much else got done) resource crunch. More commonly, the ai takes 40 minutes to 1 hour to reach the "happy" point -- and then only if its base is untouched the whole time! This is how hard it is for a nonhacked ai to get resources. The resources lost building jammers is not great - but the ai has to play a very tight game to succeed against good player. And the time lost by the ai building jammers instead of advanced c-units is the real game-loser. Because unless the ai can struggle up to level 3 resources quickly, it will NOT get the level of resources that a good player can get even by staying with level 1 and 2 units -- this is because ai's forget about geos.

"It is rare for an AI to block his plants with Jammers, I've not seen mine do it yet."
If it were to occur, it would probably occur about 1+ hour into the game after the ai's base is almost completely untouched for that amount of time.

"I admit that Peepers have their advantage over radar, however as most AI attack forces use .. missile troops, most of the AI attack troops double also as anti-air, as do it's defences."
I use Peepers like the 5th guy in the beam-down party on the old Star Trek episodes:
'Captain, I've found something! ARGH!!!!'
Many times I forget about the peepers and only notice the LOS bubble disappearing off radar. This gives me the location of enemy troops, but not the numbers or composition -- except that there's probably some AA units in it. Then I send in another peeper and pay better attention. :) What sucks is when I commit an airforce to wiping out the units in the area and in the middle of the attack realize my aircraft are getting killed by flakkers -- and I can't find the flakkers!

"Also, if the player closes in with his units, getting LOS, again, he won't be able it fire long range weapons because he'll risk hitting his own troops."
No, the player *WILL* fire long ranged weapons whether the risk of hitting his own troops is there or not. And the friendly fire problem isn't all that bad -- unless you use berthas as base defense like I do. :p
 
"Talking of factories I still need to allow my AI VP's to be used when captured."
The factory needs a build menu GUI and an ANIM of all units it can build. See my Counterstrike's ARMAIVP for details -- it's in Ai_Cheat.ufo.

"your fake ARM silo was labelled armfsilo in the files was it not? I still have a grey box where the picture should go,"
That means the ANIM for it is missing. One of my earlier versions has that problem -- I haven't figured out how to make my own ANIM's for build menus yet.

"This is at least true for default, airbattle, and seabattle -- some of the other ai profile types are so oddball (and I often hate their maptypes as well) that I haven't tweaked them to nearly the same degree."
I spent absolutely ages testing default. About 7 Versions worth I believe. Only when I considered it great did I move on. Because I felt it had been so thoroughly tested I used it, as I had intended, as a base for the other AI's. Some require more or less a fresh start, such as ThePass, but others, like Wind, just need a very slight adjustment. Mind you, since then, I've still fiddled with default despite meaning to leave it alone :)

"Although my nonhacked ai may not seem all that great, I consider it almost as good as possible considering the tools I 
have to work with"
I consider Mostly Harmless the best non-cheating AI out there. I spent weeks tracking it down for a reason :) Of course, I make sure my testers say mine is the best :) Advertising you understand. Heh heh.

"It is a minor insult to me for you to say that my ai should be something you can easily better unless you're already a VERY good TA player"
Already? I was playing TA from close to day 1. Those PeeWees got me hooked :) I couldn't evaluate exactly how good I am as BY is too slow for me really, however I raid, attack from multiple sides, use the range of units, scout. I believe I would qualify as an above average player. I just never played against TA:CC players, and there are extremely good players without CC out there. The problem is, when I play AI's, I get lazy :) I attack in one group, slogging it out with the AI's one group. Mind you, I find it makes the games more entertaining as it makes up for the AI's bad points by not exploiting them. 
I apologise for causing offence. However, I did consider your AI to build construction too early, and that a quick hit would take it out. However, your AI does get going quickly after that. It is to this end that I have been playing with the construction weightings, I'm trying to get them inside 8 minutes, without stopping the AI being ready to prevent raiders by the 4th. Also, while they sometimes help the AI by expanding and grabbing resources, others they cripple it with 4 simultaneous Sentinels. I'm still playing with them. However I do believe that currently my AI waits a tad (slightly) too long to get them.

"You assume too much about me, it's embarrassing. :p" 
Meaning?

"Major changes get bigger jumps in version number - usually to the next whole number. That's why my default ai is now at v12.1, even though it's only seen about 5 major revisions."
This was my original intent. I may do it when I release the AI, but when it came to writing it, I wasn't sure which value should change. had I just made a minor change, or was the knock on effect big? In the end, I plumped for a single number system to keep things simple. Anyway, I only release to a select group of people, and only after lots of changes.

"It is unlikely that your ai would fare well..."
YOU UNDERESTIMATE QUELLER!!!! Well, no :) You're right, but the more humans (in a team) involved, the less chance an AI has. An AI goes for the nearest target, where as humans team up and destroy the AI's in concentrated combined assaults. 
My LAN games are always exciting, because my partner is my 7 year old brother :) He doesn't play a very aggressive game, though he does like standing his Commander on a hill and D-Gunning everything to hell. Actually we lost our last game, he sent his attack force to aid me, leaving has base undefended. He got killed, leaving me to face 2 AI's. I may have done it, except the 2nd AI rather than assault my force (as I thought it would do), struck behind me, catching my behind the lines vehicle factory unready. My force, attacked from both sides, was trashed. Oh dear :) 
"Afterwards, I am usually relieved of most of the stress regardless of my grades. After many years of college, I have acquired an "I'm done with it, fuckit."" 
Ah, a fellow spirit :) Now I'm moving onto another level of education, I like to take a relaxed approach to these things. Always look to the future.

"instead the weights of one side are followed immediately by the other side's weights" 
That's exactly what mine do.

"Your ai profiles are neither in alphabetical order nor build menu order nor in order based on what an individual constructor can produce."

"Neither are yours, Vehicle Plant is about your 4th item, and I don't remember geothermal coming before it, after all your first construction unit is a Commander, and he can't build Geothermal."
That's the weights section -- and BLANK.TXT explains the sorted order. The weights are sorted in Level 1 + Level 2 buildings because that's the build menu of construction units. The weights are from HIGHEST to lowest, to capitalize on their priority in the ai profile. This is meant to give geothermals a slightly higher build probability than they would otherwise, and also to decrease a "lucky" build of a weight 0.05 unit. So my weights are sorted by priority and by build type.
My limits section are organized by build type, then level (1,2,3), then alphabetical order.
Buildtype is sorted in this manner:
Aircraft, Hovercraft, K-bots, Ships, Seaplanes, Vehicles, Buildings, Naval Buildings
Oftentimes, I delete the whole section out (as is the case of missing ships section in my Default.txt ai!) because the ai has no constructors to build those units. This shortens the ai profile, and hopefully makes it run faster. Much of my ai profile is done in such a way to minimize its length, particularly the much-simplified easy/medium/hard sections.

"I was tempted not to bother with other levels at all, it's hard enough to make an AI tough. However, my brother plays medium, so I do it for him :) For Easy I just weighted everything to 1, I think that'll work :)"
I mean I don't make a separate ai profile for easy and medium that plays with a different "style" -- ie: more air, less ground, fewer defenses, etc. I just limit superweapons, c-units, and factories and let the built-in resource modifiers of the skill settings (50% on easy, 70% on medium, and 100% on hard) make the ai more pacifistic.

"but I discovered that sorting by level and then sorting by type made more sense than sorting only by type. The weights on groups of units built by the same constructor tend to only affect one another, NOT weights on dissimilar units" 
I don't know if you have my latest version, but you'll notice that everything that is constructed by a plant is at the top half, and units take up the bottom half. However, when I go in and modify my AI it is generally with a set task in mind, and the set tasks usually break down into the groups I have. More resources needed, less plants needed etc.

"A Thud will keep a player busier than a Crasher will. Seeing shells land among your troops usually scares people into action."
Thuds are incredibly short-ranged, (shorter than Hammers by far) slow, and less willing to fire on the move. Crashers on the other hand are AA. A Thud is something I can dance my flash tanks around and ignore till all the Storms and Crashers are dead because its ballistic shots have difficulty hitting flash tanks moving near their full speed. A Crasher crossing the map is far more likely to blow a metal extractor on passing than a Thud as well. (A Storm is good at this too ;) I try to use the free will bug to its fullest. Thuds make the grade, barely. Crashers get the nod only because they're AA, not for their troop-fighting abilities. An attack force with inadequate AA is worse than an attack force with too much. :p

"If I was a a TA unit, I'd be a Storm. They look so cool :) Well, maybe I'd be a PeeWee :)"
I'd be an Arm Pelican on a flat land map. It'd seem out of place extermely, but would in fact be in its element. Cowardice would be my "weapon" of choice -- keeping near max-range and always behind heavier attackers. They lack the "glamour" of Mavericks and Krogoths, but end up being more useful for a variety of tasks -- and doing it well -- for their cost. They're not so good for attacking bases, but are good at providing AA for such attacks.

"In a 1K/1K game, getting a raid to the enemy base before the 4 min mark on a non-metal map is near impossible."
A Jeffy-rush is the bane of almost all ai's. And it works well on players too, so much so that my early build strategy works to defeat that strategy!

"There is one solution, Comm-boosting. I have written in my ever growing "How to make the AI more difficult" section, that players should not boost production, simply because there is nothing the AI can do to match that kind of production."
Comm-boosting = using your comm to assist in building resources? If so, yes -- generally the ai doesn't do that. BUT, I often get "work gangs" of 2-5 c-units building massive projects and strings of projects at buildrates unmatched even by BAI on hard. (I lose because I run out of resources ussually, and workgangs take time to build and organize.) My Commander is just one unit in one work gang, more useful than most -- but not the linchpin to my wins. Comm assistance doesn't help me much against counterstrike. :p I ussually require multiple players to beat a single Counterstrike ai unless we wall ourselves in and bertha and nuke it to death. (Sometimes, even that doesn't work.)

"I'm a team obsessive (only I could find room on Ashap Plateau for 2 on 2 team games)."
I've played 2 vs 2 on Ashap Plateau a couple times -- it works best with LOW resource starts (1k/1k or LESS) and low unit max -- like 20 to 100.

"Weekends only for me, that's the annoyance of local call costs."
Argh -- I've been idling on BY for 2 weeks now for NOTHING!

"Today, AI's, tomorrow.........the PLAYERS!!!!!!!! :)"
Players prove well... unpredictable. I've seen my ai lose to players who aren't all that good versis other people simply because they did something oddball. One of the most common tricks is a group of 5 flash tanks are sent out on a single mission -- kill the ai's commander!

"My AI still has a way to go, I've got to balance construction against attacking slightly better."
That's a problem mine has, and will keep. The balance is fluid, weights are not. If the ai gets attacked - or is about to get attacked what was good before sucks now. If the ai's commander decides to go on a metal extrator building spree, it's good to get a c-unit building solars quick before the com drains energy down to negative amounts. But just as the c-units start work on solars, shit happens...

"I'd say Metal is easier." (to make an ai decent at)
You got me there, but I forget about metal maps because I play them so little.

"Times like this I wish for a MetalFlat and MetalHilly profile."
Ideally, each map would do best with an ai profile specifically tooled for each of them. This is why some of C_A_P's maps draw praises for ai behavior -- it's because I've had a chance to make an ai set up with that map's limitations in mind, metal extractors are limited to values within limits of the map and weighted to get what I feel is the proper level of wanderlust for the c-units.

"Tried the special Skirmish maps from TAMEC?"
Yes, "Skirmish" is in fact easy versis 9 of my ai's.

"Problem is I spend so long watching AI on AI battles that I hate to drag myself back to playing myself, and when I do, Normal seems so slow now as all battle I watch are +10."
I watch mine at +21 speed. :p (Registry hack -- discovered how when making a resource registry hack.) Originally, the computer's speed was +53 (value 63 in decimal) which was too fast for my lowly P2-400.

"Getting weights to something that results in a good ai is an art, the limits are more of a science -- anyone could get "lucky" but you're using a method that makes that far more likely than anyone else I know."

Well if I'm using a method that lets me get lucky more often, I'll start playing the lottery again :) So how do get lucky so easily? Updating every 5 seconds? I generally limit something 10X more than it was weighted, or did, I halved them all recently. 

The way to get "lucky" more is to first find out if what you're testing can affect the ai when modified in the EXTREMES -- then try to hedge your bets down to a sure win.

"Well, I'm back to glorious AI testing to get construction just right. Hurrah! I really should get round to making a webpage, I promised the testers a message board too."
Please send me a few of their email addresses, or have them contact me.

David
--------
Subject: Re: Ai practice
Date: Mon, 23 Aug 1999 22:45:03 +0100

** Why have you halved the max damage values of aircraft for your AI aircraft?

"(This doesn't explain the stupid commander bug where the ai's comm stands around before doing anything -- that's probably a too-long ai profile, an exe bug, or the results of an overtaxed cpu "
** I've had this happen (a lot recently) quite early. Twice recently my AI Commander has just stood there after building 3 metal extarctors. He waited 50 seconds before doing anything else. I have *no* idea why.

"My main complaint about ai nonmobile jammers is not so much their chance of blocking a factory -- which I have exaggerated the likelyhood (except on a map like Pincushion) -- it's their cost" 
** On the other hand they drain fewer resources while built, which may save the AI by having less being drained, allowing the AI to gain resources again, and other factories to produce as quickly as possible. 
"'Captain, I've found something! ARGH!!!!'" 
** Probably a killer alien with the ability to turn into the episodes standard beautiful woman. 
"No, the player *WILL* fire long ranged weapons whether the risk of hitting his own troops is there or not." 
** In that case there is no need to worry because he's just as likely to kill his own troops as he is yours. 
""your fake ARM silo was labelled armfsilo in the files was it not? I still have a grey box where the picture should go,"
That means the ANIM for it is missing."
** However what I'm trying to do is wipe it out completely. What do I need to remove to make it forget that a Fake Silo can appear on the menu.

"let the built-in resource modifiers of the skill settings (50% on easy, 70% on medium, and 100% on hard) make the ai more pacifistic." 
** Ah, I never knew that.

"A Thud is something I can dance my flash tanks around and ignore till all the Storms and Crashers are dead because its ballistic shots have difficulty hitting flash tanks moving near their full speed"
** However all a Crasher is is a weaker armoured and weaker armed Kbot than a Storm, AND it builds slower too. If you want AA the main weapon of the Level 1 CORE is the Slasher, so Crashers become unimportant. However Thuds add something new, they can fire over wreckage and DT's. Crashers add nothing new, and are not worth the cost when you can have Slashers.

"Comm-boosting = using your comm to assist in building resources?"
** No, I meant using your Commander to speed up the building of troops. An AI can't do that, but using this technique a player can launch an assault in under 2 minutes.
"Argh -- I've been idling on BY for 2 weeks now for NOTHING!"
Grrrrrrrr :)

"One of the most common tricks is a group of 5 flash tanks are sent out on a single mission -- kill the ai's commander!" 
** Well as most players have CC, I've basically ordered them to get your 1.1 bugfix if they have CC, simply because this solves the whole Commander wiggle problem.

""My AI still has a way to go, I've got to balance construction against attacking slightly better."
That's a problem mine has, and will keep. The balance is fluid, weights are not."
** After watching your AI in action, I must admit your construction strategy beats the pants (that's a good thing) off of mine. I tried your CORE against my ARM in a 2 on 2 on Metal Heck 2 (yes I used your WindMet profile, though I renamed it in line with Queller to WindMetal). My 2 ARM's pasted your first CORE, while their defences held off the second one. However your second CORE built up his base so fast that not only did his force hold off the combined attacks of both Quellers, his army started to gain ground! However, eventually he started to hit his plant limits, and so Queller started to catch up. Eventually it was becoming bogged down in the centre, MH's planes were keeping my AI occupied, while artiller shelled my base. In the end I got bored of the Stalemate. In the end my AI would probably have won though weight of numbers as it had more AI's, however with the speed down to -1 I wasn't waiting that long :)
I'd say that my AI starts off better, in all modesty :) I prefer the starting routine of my Commander to yours, it builds a better combination of resources before it builds a plant. Too often yours would build few resources, then try for two plants. Too often it's first plant was an AP. However after that your AI woukld play better if the map was a 16x16 or greater. It would wheel out construction units, expand like mad, and evetually just out produce poor little Queller. I'm glad it did though, because it has opened my eyes to a problem Queller has, it simply doesn't expand fast enough. Left to itself or winning battles it will eventually expand, but to begin with it sits in a rut. I don't know how your AI would handle a starting raid, in fact even on a 16x16 Queller attacked before Mh was ready, however it quickly started churning out units to stop the attack, and eventually pushed Queller back.

"You got me there, but I forget about metal maps because I play them so little." 
** What annoys me is the way the AI Commander insists on still walking the same kind of distance he would on a Default map before building a metal extractor. It's terrifying really.

"Yes, "Skirmish" is in fact easy versis 9 of my ai's."
** I would expect it to be so against any AI. Firstly you have walls which render missile troops in the AI's hands useless. Secondly you can create killzones.

"Originally, the computer's speed was +53 (value 63 in decimal) which was too fast for my lowly P2-400."
** Subtle :)

"Please send me a few of their email addresses, or have them contact me."
And what should I tell them they're contacting you for? 
--------
** Why have you halved the max damage values of aircraft for your AI aircraft?
Rub it in that I forgot to answer one why don't ya! I think I'll save that question for next time. It is of little consequence to the ai anyway.

"(This doesn't explain the stupid commander bug where the ai's comm stands around before doing anything -- that's probably a too-long ai profile, an exe bug, or the results of an overtaxed cpu"
** I've had this happen (a lot recently) quite early. Twice recently my AI Commander has just stood there after building 3 metal extarctors. He waited 50 seconds before doing anything else. I have *no* idea why.
The lovely failsafes at work. Ai comm says: I'll build a metal extractor. Failsafe override -- not enough energy, order canceled -- restart decision process! Ok, I'll build a factory. Failsafe override -- not enough energy, order canceled -- restart decision process! Ok, I'll build a LLT. Failsafe override -- not enough energy, order canceled -- restart decision process! And so forth and so on. Until it choses Solars -- and even then it may kick in the same failsafe for all I know UNLESS it has energy max. The ai considers production values more than storage so if it's making +25 energy (the comm) and using -9 for 3 metal extractors it sees that it doesn't have NEARLY enough energy income to sustain a build project. In any case, Cavedog programmers on TA second-guessed themselves on the ai when they made failsafes in the first place because their ai profile method worked so dismally. (and still does!) And their failsafes cause the ai to do some very stupid things as a result as well -- if the failsafes had reasonable results, it wouldn't be nearly as disasterous.

"My main complaint about ai nonmobile jammers is not so much their chance of blocking a factory -- which I have exaggerated the likelyhood (except on a map like Pincushion) -- it's their cost" 
** On the other hand they drain fewer resources while built, which may save the AI by having less being drained, allowing the AI to gain resources again, and other factories to produce as quickly as possible.
Yes, so they should be weighted as high or higher than other units out of the plant but lower than the advanced c-units out of the plant.

"'Captain, I've found something! ARGH!!!!'" 
** Probably a killer alien with the ability to turn into the episodes standard beautiful woman.
Anyway, you get the point - finding the bad guys is WORTH losing a peeper. Finding what they have is worth (to me) another peeper or 2.

"No, the player *WILL* fire long ranged weapons whether the risk of hitting his own troops is there or not." 
** In that case there is no need to worry because he's just as likely to kill his own troops as he is yours.
Not likely, the weapons themselves in TA are either fast weapons or smart weapons. With a few notable exceptions most units in TA have rather accurate weapons. A nonmoving target can expect to get hit almost dead-on even by the first shot of many different types of weapons -- barring terrain and debris. A bertha may not be particularly accurate -- but I'm ussually shooting at a big mob then (and at its center) so its overshoots and undershoots are quite unlikely to hit my units. For every 1 friendly unit I kill using such tactics, I kill around 10 enemy units. You have to also remember that you CANNOT hit your own units with your weapons, you can only hurt them with splash damage. (The same is not true for your ally's units. ;) This is what makes a Samson horde on a flat map so danged deadly -- they can all fire without hitting each other.

""your fake ARM silo was labelled armfsilo in the files was it not? I still have a grey box where the picture should go,"
That means the ANIM for it is missing." 
** However what I'm trying to do is wipe it out completely. What do I need to remove to make it forget that a Fake Silo can appear on the menu.
It's in the download dir under one of the build menu TDF files. That's a text file, it should prove very easy to read and understand for you at this point. Delete it out and the problem's solved.

"let the built-in resource modifiers of the skill settings (50% on easy, 70% on medium, and 100% on hard) make the ai more pacifistic." 
** Ah, I never knew that.
And there's a lot of things I don't know either (and am forgetting, since I no longer work with many of the inner details of TA daily.)

"A Thud is something I can dance my flash tanks around and ignore till all the Storms and Crashers are dead because its ballistic shots have difficulty hitting flash tanks moving near their full speed"
** However all a Crasher is is a weaker armoured and weaker armed Kbot than a Storm, AND it builds slower too. If you want AA the main weapon of the Level 1 CORE is the Slasher, so Crashers become unimportant. However Thuds add something new, they can fire over wreckage and DT's. Crashers add nothing new, and are not worth the cost when you can have Slashers.
Slashers gum up easier on debris than Crashers -- and you cannot count on the ai getting Slashers going even by the 10 minute mark if it built 2 kbot labs and then an air plant first. Crashers are supplimental AA. You haven't seen the multiplayer games with my friends where we take apart the ai's assault force with aircraft by killing the ai's PATHETIC AA in just 1 or 2 passes, then decimating the rest *AT LESURE* because the ai has such poor AA with its attack force. My ai gets tilted heavily to the AA side because that's how most players play -- both versis the ai, and versis each other. I've seen Storms shoot down aircraft, but they're rather unreliable in that regard - thus my ai builds Crashers.

** No, I meant using your Commander to speed up the building of troops. An AI can't do that, but using this technique a player can launch an assault in under 2 minutes.
So it's using the comm to speed up production of troops by guarding a factory. Yeah, that's the purpose and my "justification" for the Ai-only vehicle plant and Ai-only krogoth gantry in Counterstrike. But that is counter-productive to the EXTREME in the first few minutes of the game on a large map, if starting resources are low.

"Argh -- I've been idling on BY for 2 weeks now for NOTHING!" 
** Grrrrrrrr :)
That's what I'm saying!

"One of the most common tricks is a group of 5 flash tanks are sent out on a single mission -- kill the ai's commander!" 
** Well as most players have CC, I've basically ordered them to get your 1.1 bugfix if they have CC, simply because this solves the whole Commander wiggle problem.
I wish it did, but in fact it only makes it less of a problem -- the problem doesn't go away.

** I'd say that my AI starts off better, in all modesty :) ... Too often it's (my ai's) first plant was an AP. 
Whichever one doesn't build an airplant first ussually starts off better! And in my tests one ai is unfortunately about as likely to do that as another, regardless of weights! If it makes its first airplant as it's 3rd or 4th plant, it ussually does good. If you look at my default.txt ai profile, you'll find airplants at almost the very bottom of the weights list (for their section.)

** I prefer the starting routine of my Commander to yours, it builds a better combination of resources before it builds a plant. Too often yours would build few resources, then try for two plants.
That's unfortunately the build factory failsafe that kicks in if resources are too high. Some versions of my default.txt have this problem only rarely -- that is where they build 3-5 solars and 3-5 metal extractors before they build a factory. What often happens is the ai builds a solar RIGHT after a metal extractor and bottoms out its energy. The energy failsafe cuts on and the next 2-10 things it builds are solars! Most annoying, and ai profiles cannot avoid that problem entirely.

"You got me there, but I forget about metal maps because I play them so little."
** What annoys me is the way the AI Commander insists on still walking the same kind of distance he would on a Default map before building a metal extractor. It's terrifying really.
I think it's sad, not terrifying. If a nonhacked ai could play like Counterstrike -- or build a defended base as quickly as BAI on hard -- THAT would be terrifying.

"Yes, "Skirmish" is in fact easy versis 9 of my ai's." 
** I would expect it to be so against any AI. Firstly you have walls which render missile troops in the AI's hands useless. Secondly you can create killzones.


"Originally, the computer's speed was +53 (value 63 in decimal) which was too fast for my lowly P2-400."
** Subtle :)
I can send it to you if you want it.

"Please send me a few of their email addresses, or have them contact me."
** And what should I tell them they're contacting you for?
Feedback about TA Bugfix v1.1 and my ai's as well! Only time I ever get messages about TA Bugfix v1.1 is ussually from someone who can't install it.
--------
** Why have you halved the max damage values of aircraft for your AI aircraft?
Weapon specific damage values -- such as nukes, missiles, and EMP and TRON weapons. Because missile weapons do double damage to aircraft, halving the max damage values of AI aircraft roughly averages out. Had I not done this, AI aircraft would prove much harder to kill. The problem is mentioned in TA Bugfix v1.1 readme under seaplanes.
--------
Subject: Re: Air ai for special TA map
Date: Tue, 31 Aug 1999 21:25:01 +0100

Ah, way ahead of me :) I haven't started on it yet, I need to get my Queller Airbattle stuff going, and in fact have it scheduled soon, but I need to get the UH codes in really to Queller. 
Hmmmmm, I personally plan on weighting the ground units a bit higher, thought I agree, you can't have enough AA :) 
Aren't Jethros cheaper than Pelicans? If so, and if the routes between the starting points are valleys, wouldn't you be better with just Jethro's? Also on that note, wouldn't Samsons be better? The plant is cheaper and they build a lot faster, they also don't need to open their gun ports. I haven't checked damage mind you. 
Oh, and please note that these comments are made solely on your notes, I haven't looked through the weights. 
Quitch(y)
Creator of the Queller AI for TA, CC and UH
"Anti-AI tactics are the way of the coward!" 
> >From: Bob <datanut@urbancom.net <http://207.82.250.251/cgi-bin/linkrd?_lang=&hm___action=mailto%3adatanut%40urbancom%2enet>
> >Subject: Air to Air
> >Date: Wed, 25 Aug 1999 01:34:29 -0500
> >
> >HI
> >
> >I would like to summon your expertise on AIs.
> >I want to make a map that specifically needs air combat and ground
> >forces once you get to your objective.
> >The map I'm going to design will be divided up with unclimbable
> >mountains with clouds. All mountains will be voided. The span of the
> >voided area will make the big guns useless. There will be different
> >valleys that will encure different combat strategies. The way to get to
> >these valleys is the basis of this map design.
> >
> >I have experimented with the current air AIs and none is good enough.
> >I can taylor the map to accommodate the AI if I had one.
> >
> >Can you make an AI that uses the standard Cavedog units and force lots
> >of air vehicals to be built? Like a challenge? Will put your name in
> >bright lights on my map!
> >
> >Thanks
> >
> >Datanut

This is my best air ai so far -- at least from simple tests. It will offer a newbie a challenge, but I cannot promise much else. Aircraft build slowly and the ai is TOO stupid to know to guard its factories with extra c-units.

Weighted VERY low (at 0.05) for Arm is a single kbot lab -- meant to be built long after the ai's commander is started on his repair patrol. The hopes are that it will be built somewhere near the enemy. But that's a random thing, and you never can tell. If you don't like it, I can remove it for Arm. The way it's set up currently, the only things the Arm kbot lab makes is c-units and AA kbots (Jethros and Pelicans) -- this is so even if they're blocked they'll still be somewhat useful. It's my opinion that you can never have too much AA. :)

Core's ai does NOT build any factories besides aircraft plants, but I could change that to build vehicles (Slashers and mobile flakkers particularly) if you think it would have a reasonable chance of doing any good.
David
--------
Subject: Switcher
Date: Sun, 5 Sep 1999 16:34:49 +0100

I was wondering if you could do me a favour? 
BSR, after much whining by me (heh heh), has bee considering adding compatibility for your BUGFIX to his Switcher. The Switcher is a launcher UH uses, however the Rev31 file it looks for is different to the one the BUGFIX uses, so either you have 2 TA's on your harddrive, you move all the files around before you can play a game, or you only have the BUGFIX or UH. 
However I have been persuading him to get his launcher to recognise the BUGFIX (simply because I love the changes it makes), so that when UH is switched off, it will revert to a BUGFIXed state. 
If you could just e-mail BSR at smot@mindspring.com (unless you have a better e-mail address for him) telling him you don't mind him adding BUGFIX compatibility, and also what a wonderful guy I am (oops, how'd that get there :)). I'd be most grateful, thanks! 
Quitch(y)
Creator of the Queller AI for TA, CC and UH
"Anti-AI tactics are the way of the coward!"
--------
>From: "Quitch" <quitch@fickle.demon.co.uk>
>Reply-To: "Quitch" <quitch@fickle.demon.co.uk>
>To: "Switeck" <switeck@hotmail.com>
>Subject: Switcher

>I was wondering if you could do me a favour?
Don't you mean I need to do BSR a favor? :p
(It wouldn't be the first time...)

>BSR, after much whinging by me (heh heh), has bee considering adding compatibility for your BUGFIX to his Switcher.  The Switcher is a launcher UH uses, however the Rev31 file it looks for is different to the one the BUGFIX uses, so either you have 2 TA's on your harddrive, you move all the files around before you can play a game, or you only have the BUGFIX or UH.
>
>However I have been persuading him to get his launcher to recognise the BUGFIX (simply because I love the changes it makes), so that when UH is switched off, it will revert to a BUGFIXed state.
>
>If you could just e-mail BSR at smot@mindspring.com (unless you have a better e-mail address for him) telling him you don't mind him adding BUGFIX compatibility, and also what a wonderful guy I am (oops, how'd that get there :)).  I'd be most grateful, thanks!

I will email him, but it may not mean we will get along any better.
Hopefully he will include TA Bugfix support with Switcher. However, I do not expect further cooperation.

David
--------
Many thanks.
--------
From: Robbie Eatherington <rillip3@mailcity.com>
To: ", XXX_Wretched" <wretched_1@hotmail.com>, XXX_Wingman <jah50@hotmail.com>, XXX_Theta <theta@tampabay.rr.com>, XXX_Sabrewolf <elijah@cableone.net>, XXX_REDZ1R & XXX_RIX007 <ahicks1@tampabay.rr.com>, XXX_Rebel <jcbusby@gw.total-web.net>, XXX_Oquendo <xxx_oq@netzero.net>, XXX_MINA <xxx_mina@hotmail.com>, XXX_MechWarrior2 <thor330@yahoo.com>, XXX_MechWarrior <warman33@hotmail.com>, XXX_MarshalJake <jasonw@post.cis.smu.edu>, XXX_Juvenile <missle4u@hotmail.com>, XXX_JJQUICK <badmisterfriars@hotmail.com>, XXX_Fulgore <noahm@cableone.net>, XXX_Fenrir <fenrir_86@hotmail.com>, XXX_EnPhorSeR <coppershrooms@msn.com>, XXX_DNA <hollngrs@bright.net>, XXX_DeathKnight <out124@aol.com>, XXX_dante767 <dante767@hotmail.com>, XXX_Convict_ <scarter1@tampabay.rr.com>, XXX_BloodyFang <dmj_88@hotmail.com>, XXX_BloodSport <too_toogood@yahoo.com>, XXX_BLaDeR <camsoup16@yahoo.com>, XXX_BlackWidow <xxx_blackwidow@yahoo.com>, XXX_Blackout <ty8ler@aol.com>, XXX_BLACKMETAL <xxxblackmetal@hotmail.com>, Xeno <mlsheltn@memphis.edu>, WWO_ReD9 <imat@email.msn.com>, WayConn <wayConn@bellsouth.net>, Switeck <switeck@hotmail.com>, StEeLe <BJ_StEeLe@hotmail.com>, Spunky the Necromancer <mgarrett@ns.gemlink.com>, Ros <glyn@gcdando.demon.co.uk>, RaPtoR <raptor@pivot.net>, Outlaw_Ned <gunslinging.outlaw@octa4.net.au>, NoN_wingsfan83 <wingsfan83@aol.com>, NoN_Phalout <phalout@hotmail.com>, NoN_Lummus <clea@sprynet.com>, NoN_DoSea <dosea@rain.org>, Mark Millington <Psiren@aol.com>, Mandy <warren@atlantic.net>, MaldorEldomic <cheeze2761@hotmail.com>, Luke Wakeling <luke@waking.globalnet.co.uk>, Loup-Garou <dandevil@aol.com>, JustNasty <justnasty@hotmail.com>, JustRuthless <xxx_scizzor@hotmail.com>, JHD_Fakenrite <danr@total.net>, JHD_Cpt_Dread <jyhad@home.net>, JHD_Acelevy <homebrew55@yahoo.com>, GI_Bly <jbailey@citlink.net>, Garth Barber <tomcat2001@hotmail.com>, Gangsta Ba <klcan@idirect.ca>, EMG Man aka StumpDude <rillip3@mailcity.com>, DeathWish2 <xxx_deathwish@hotmail.com>, DeathWish <Kramer9966@aol.com>, David <david@datatask.com.au>, Dan Samuel <dan@synthetic-reality.com>, Craig Poole <psykotyx@aol.com>, "Commander, Core" <l3nnyh@aol.com>, Bryan <bj_topdog_22@hotmail.com>, Brave Sir Robin <smot@mindspring.com>, BOLT <h.m@telia.com>, Bob <datanut@urbancom.net>, Andrew Roy Arbon <andrew@gatekeeper.demon.co.uk>, Aklein <aklein24@hotmail.com>
Subject: Quitch's e-mail is down...
Date: Mon, 08 Jan 1996 01:17:12 -0800
From rillip3@mailcity.com Sun Sep 26 06:42:57 1999 


He would like you to get ICQ (if you don't already have it) and add him
to your list (ICQ# 13295204).  He is also the only one with the name
Quitch, so you can search by name too.  He would also like to say that
he is thinking of running a Testers Vs QAIv10 game.

Quitch's e-mail at the University is busted.  He can recieve and read
e-mails, but can not send them.  Or at least that's what I think he said
the problem was.  Anyways, I suppose it really doesn't matter. I'll
probably be acting as a messenger for Quitch until he gets his e-mail
working.

Quitch also needs some help with graphics.  Since I am already doing the
story lines, and am completely inempt at graphics, I can't do anything
much.  If any of you guys do art, please scan it or do it in MS Paint or
something, send it to me, and I'll convert it to .jpg or .gif format for
use on the web site.

Still needs to finish the Arm intro...
EMG Man (a.k.a. StumpDude)
--------
Subject: Mostly Harmless v12
Date: Wed, 20 Oct 1999 11:30:05 +0100 (BST)

On the website for Queller (which I planned to release 2 weeks ago, my monitor blew up preventing me), I am going to have downloads for every AI I can find, and any that are sent in.

I am writing to you to ask your permission to replace your Mostly Harmless default and airbattle v11.7 profiles, with the v12 profiles you sent me.  If allowed I will include them in the ZIP rather than the 11.7 ones.

Also, what address did I use for you before?  My address book has been broken and I can no longer access it.

Quitch(y)
--------
Please could you give me the e-mail address of Datanut, thanks!

I have finally released a copy of Queller to the"public".
You can find it at here if you're interested.
http://members.xoom.com/queller/
--------
My ISP has been down - oh 2 week now. It got cancelled from lack of payment. We quit paying when we had trouble getting online (only 2 out of 20+ tries on different days and various times) for over a week straight, then after that the best d/l rate I could get and sustain was around 2kb/sec for my 56k modem!
Thus, you probably haven't heard from me in a while. I am attempting to rectify this by using my friend's ISP connection when I can and will get a new ISP very soon.

Here's my latest Mostly Harmless ai profiles -- if you're going to post my ai's post these instead!
(Only way to get the latest is get it from the horse's mouth!)
--------
My email address is to be ONLY my hotmail account. I DID have another email account (@gww.net) but it has been down more often than not. So do not use it.

You may post any ai I send you, but please ask before posting ai's by "me" that you get from elsewhere -- I get blamed for a lot that I have not done. :)
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: switeck@hotmail.com
>Subject: AI's
>Date: Tue, 9 Nov 1999 19:14:56 +0000 (GMT)
>
>At Estrella's I download two othe AI's by you,
>Switeck's Original AI and Switeck's Fair AI.  As far
>as I could tell they were older Mostly Harmless's with
>different names.  Is this presumption correct?
>
>Quitch

Quite true! They are all older versions of Mostly Harmless. TAMEC will probably get my newest ai's first -- if only because I can upload them myself. Anyone else that has ai's made by me either have specific ai's made for special maps (such as ones I did for CAP) or have older ai's than the ones found at TAMEC.

David
--------
From: Quitch <quitchy@yahoo.co.uk>
Subject: Re: AI's
Date: Wed, 10 Nov 1999 10:21:28 +0000 (GMT)

Oh don't worry about MH v12.6.  I got that the second it entered the build queue :).

By the way, since you've replied to this I know at least that this account works and so I need to ask you a couple of questions.

1. What is the e-mail address for datanut?

2. Could you please write a 50-100 word description for Mostly Harmless AND one for Counterstrike?  It's for the authors description section of AI Central.

Many thanks

Paul aka Quitch(y)
--------
My old ISP was down over 2 weeks, as of 2 weeks ago. It got cancelled from lack of payment. We quit paying when we had trouble getting online (only 2 out of 20+ tries on different days and various times) for over a week straight, then after that the best d/l rate I could get and sustain was around 2kb/sec for my 56k modem!
Thus, you probably haven't heard from me in a while. I am attempting to rectify this by using a crappy but free ISP in the meantime while I search for a new ISP.

Good to see you got my MH v12.6 the second it entered the evaluation queue. You may post any ai I send you or you get from TAMEC, but please ask before posting ai's by "me" that you get from elsewhere -- I get blamed for a lot that I have not done. :) One semi-popular misconception is that I made BAI. Even stranger, supposedly someone else made CounterStrike ai -- or "What's that?"

My ONLY used email address is my hotmail account. I DID have another email account (@gww.net) but it went down more than 6 months ago. So do not use it.

Datanut's email address is:
Datanut	datanut@urbancom.net	(Bob Ritz)


50-100 word description for Mostly Harmless:


"Mostly Harmless" Ai v12 is a nonhacked, noncheating ai.
However, It will use hacked units if you install them!

It works on all map types EXCEPT some maps that use special unique ai's.
NOW supports TAUIP v3.0 units on land maps.
Plus, update ai profiles for special ai  for a couple TAMEC maps by CAP.


Counterstrike description comes later -- give me a day...
David
--------
From quitchy@yahoo.co.uk Wed Nov 10 10:26:30 1999 
Subject: Re: AI's

Thanks for the descriptions!

Don't worry, I've been going to every forum where AI even gets breathed now :)  I list by the AI's who made them, and I know that Mostly Harmless is by you (and I wanted to beat TAMEC to getting it onto a main page :)).  I will ask in the future, after all I did about the Estrella ones, I didn't want to post AI's that were simply older (and therefore most likely inferior) copies of one of your current AI's.

Thanks for his e-mail.  I look forward to the Counterstrike description.

I heard about your AI in TAUIP, I read it on the TAUIP page (I can't remember who makes it), then when I first posted Queller and put up my Queller is the best non-cheating AI around, they started to preach to me the wonder of the TAUIP 3.0 AI :)  Didn't know you'd made it till recently, it sounds impressive.

Paul
--------
My ai is not a part of TAUIP, last I looked. Are you sure you're not confusing MH v12.6 with Banzi Ai or Alpha Ai? At one time, I think TAUIP came with CounterStrike Ai -- but CounterStrike Ai proved too metal-poor to sustain the very expensive 3rd party units that were hard-weighted in their FBI file to something ludicrious. MH v12.6 simply includes a list of units contained in TAUIP so it knows not to build most of them, and very few of the others. It will still choke badly for lack of metal and energy when used with TAUIP unless the hard-weighted lines are removed.
--------
From quitchy@yahoo.co.uk Fri Nov 12 06:57:11 1999 

By the way, where are the hard weighted lines?

Oh, well they said their TAUIP 3.0 AI was *very* good,
and then your AI came out saying it had 3.0 support,
so I put 2 and 2 together...........and it seems I
made 5.

Oh well.

I look forward to your Counterstrike description.

Paul
--------
Hard-weighted lines are lines that appear in unit FBI files (which are in the units dir inside a unit *.ufo file).
They are of this form:
ai_weight=weight unitname 0;

Many 3rd party units and some Cavedog units - such as CORAMPH - use this in their FBI files, which messes up any ai profiles which don't explicitly change the weights to something else balanced with the rest of the ai profile.

There are significant problems with Queller ai:
1.It doesn't ever build an ARMAAP (and probably not a CORAAP as well) when I have Sidedata.tdf placed in the Gamedata dir.
2.There are messed-up GUI menus in QuellerUnits.ufo.
3.Quite a few of the QuellerUnits.ufo download's *.tdf files exceed the 5 unit limit in their build menus. TA for some reason only uses the first 5 entries.
4.The default weight lines (Arm 0.1 and Core 0.1) are in fact messing up your complex weighting structure. Almost everything that's weighted to 0.05 elsewhere in the ai profile gets built only extremely rarely. I don't think I've ever seen a Flea, AK, Zipper, Pelican, Freaker, Gimp, Jeffy, Weasal, Triton, Lugar, Spider, Croc, Reaper, Titan, or Lancet yet. On the other hand, I've seen many other units which supposedly DO have low weights (relative to the other weights) on them. For instance, I've seen enough LLT's, Peewees, Snipers, Penetrators, and Phalanx that I am sick of them -- they seem to be built almost as much (if not MORE) than other higher-weighted units that can be produced by the same factories.
5.Something (cause unknown) is causing glitches and slowdowns in TA on my P2-400.

I have redone QuellerUnits.ufo so that Sidedata.tdf is NOT required and the missing gui menus are fixed. I have separated out the ai-only aircraft from that ufo and am sending my fixed Ai_Air.ufo from Counterstrike ai instead.
I also redid default.txt so that the default weights lines come after all the other weight lines, lowered the limits on many units that the ai gets limited utility out of (typically from 5 to 2), and I put all the limits into the same structure as my ai profiles use. The last part may prove an annoyance for you, but despite all the changes I made to default.txt -- it should play almost exactly the same as the Queller ai you've grown to like...

I wish you'd told me about the missing secondary weapons problems on ai-only stealth fighters when Counterstrike ai and TA Bugfix are used together instead of secretly making new versions of the ai-only stealths that also use new names (ABFHAWK and CBFVAMP) -- now I have to fix my Counterstrike ai and set the limit your fixed versions to 0 in all my ai profiles so my ai won't make an infinite number of your ai-only stealths if QuellerUnits.ufo is accidentally installed along with Counterstrike.

My "what is Counterstrike" post has been unfortunately delayed by the above problems. I am highly unsure about how to define Counterstrike so that I'm not describing all hacked ai's in general. I cannot say "it cheats" and be satisfied without considerable explaination for why it cheats.

On a side note, I discovered that the Arm Nuke Silo (Retaliator) corpse is worth ~2,700 metal -- making it ~1,700 more metal in value than the finished Silo costs. I am correcting that as well as other tiny glitches in the NEXT release of TA Bugfix.

Last but not least are 2 TA Demos that you might find amusing to watch. Both were done on TA v3.1 - so TA Bugfix should not be used to replay them.

David
--------
Subject: Re: AI's
Date: Mon, 15 Nov 1999 21:18:56 +0000 (GMT)

Ah, you're interfering with Queller again :)  Wait till I find fault with Mostly Harmless, I'll let you know ASAP, heh :)

OK, I'm gonna do this point by point:

> >Subject: Re: AI's
> >Date: Fri, 12 Nov 1999 14:57:10 +0000 (GMT)
> >
> >By the way, where are the hard weighted lines?
> >
> >Oh, well they said their TAUIP 3.0 AI was *very*
> good,
> >and then your AI came out saying it had 3.0
> support,
> >so I put 2 and 2 together...........and it seems I
> >made 5.
> >
> >Oh well.
> >
> >I look forward to your Counterstrike description.
> >
> >Paul
>
> Hard-weighted lines are lines that appear in unit
> FBI files (which are in
> the units dir inside a unit *.ufo file).
> They are of this form:
> ai_weight=weight unitname 0;
>
> Many 3rd party units and some Cavedog units - such
> as CORAMPH - use this in
> their FBI files, which messes up any ai profiles
> which don't explicitly
> change the weights to something else balanced with
> the rest of the ai
> profile.
>
> There are significant problems with Queller ai:

It's better than MH? :P  Just kidding ya.

> 1.It doesn't ever build an ARMAAP (and probably not
> a CORAAP as well) when I
> have Sidedata.tdf placed in the Gamedata dir.

This is a problem I have had.  It does build the CORE version (quite a bit, and with considerable success).

> 2.There are messed-up GUI menus in QuellerUnits.ufo.

I'm not unit editor, I just chucked some stuff together so I could get on with the AI.

> 3.Quite a few of the QuellerUnits.ufo download's
> *.tdf files exceed the 5
> unit limit in their build menus. TA for some reason
> only uses the first 5
> entries.

The what *dribble* :)

> 4.The default weight lines (Arm 0.1 and Core 0.1)
> are in fact messing up
> your complex weighting structure. Almost everything
> that's weighted to 0.05
> elsewhere in the ai profile gets built only
> extremely rarely. I don't think
> I've ever seen a Flea, AK, Zipper, Pelican, Freaker,
> Gimp, Jeffy, Weasal,
> Triton, Lugar, Spider, Croc, Reaper, Titan, or
> Lancet yet.

Good, you shouldn't, they are possibly the most useless units in the history of clone (or pattern) kind for an AI.  They're just in there for kicks, someone may use the 5000 unit patch :)

On the other
> hand, I've seen many other units which supposedly DO
> have low weights
> (relative to the other weights) on them. For
> instance, I've seen enough
> LLT's, Peewees, Snipers, Penetrators, and Phalanx
> that I am sick of them --
> they seem to be built almost as much (if not MORE)
> than other
> higher-weighted units that can be produced by the
> same factories.

Rubbish.  The occasional LLT, so what?  The first complaint I had when starting Queller was that it was boring, it would Flush and Slush, and people didn't like playing it because of that.  I have worked on Queller based on maths, THEN based on trial and error.  Removing the 0.1 destroys the balance.  Peewees are sweet, and usually don't get produced till the AI has
a solid ground force anyway.  I certinally see a lot more Panthers than any other vehicle, and Penetrators are fairly useful with the bugfix, especially as the AI can see at max LOS.  We just have different views on units.  I think your AI uses WAY too many Dominators.

> 5.Something (cause unknown) is causing glitches and
> slowdowns in TA on my
> P2-400.

Incompotent user at keyboard error? :)  No idea, I don't have any problems.

> I have redone QuellerUnits.ufo so that Sidedata.tdf
> is NOT required and the
> missing gui menus are fixed. I have separated out
> the ai-only aircraft from
> that ufo and am sending my fixed Ai_Air.ufo from
> Counterstrike ai instead.
> I also redid default.txt so that the default weights
> lines come after all
> the other weight lines, lowered the limits on many
> units that the ai gets
> limited utility out of (typically from 5 to 2), and
> I put all the limits
> into the same structure as my ai profiles use. The
> last part may prove an
> annoyance for you, but despite all the changes I
> made to default.txt -- it
> should play almost exactly the same as the Queller
> ai you've grown to
> like...
>
And I'm sure I'll change 'em all right back.  Queller is my work, and I have every intention of doing it myself, though your help is appreciated, but I don't want to make an AI only to realise that a lot of the credit should go to someone else.  No.

Another issue is that I didn't want to have an ARM and CORE AI that weren't equal.  Almost every AI I know has the problem that CORE will suually subdue ARM simply because they get Goliaths and ARM doesn't.  I personally don't think Queller overuses these units.

Also, you keep limiting everything to what seems to be the MH 200-250 support, where as Queller goes for a more 250-500 orientated approach.

> I wish you'd told me about the missing secondary
> weapons problems on ai-only
> stealth fighters when Counterstrike ai and TA Bugfix
> are used together
> instead of secretly making new versions of the
> ai-only stealths that also
> use new names (ABFHAWK and CBFVAMP) -- now I have to
> fix my Counterstrike ai
> and set the limit your fixed versions to 0 in all my
> ai profiles so my ai
> won't make an infinite number of your ai-only
> stealths if QuellerUnits.ufo
> is accidentally installed along with Counterstrike.
>
Well then you'll have to do that for millions of units.  Why would QuellerUnits.ufo be in there when Counterstrike is installed?  If that worries you then you'd have to do the same for every 3rd party unit, vehicles plants, BAI units etc.  Whats the problem?

Secretly?  I let you know about a week after making them.  Is it my fault if you've made a product which isn't caperble of running alongside your own AI?  My CC reccomend that people get your bugfix, so I need fighters for the bugfix, and fighters for normaly TA without them.

> My "what is Counterstrike" post has been
> unfortunately delayed by the above
> problems. I am highly unsure about how to define
> Counterstrike so that I'm
> not describing all hacked ai's in general. I cannot
> say "it cheats" and be
> satisfied without considerable explaination for why
> it cheats.
>
Well I look forward to seeing what you come up with.

> On a side note, I discovered that the Arm Nuke Silo
> (Retaliator) corpse is
> worth ~2,700 metal -- making it ~1,700 more metal in
> value than the finished
> Silo costs. I am correcting that as well as other
> tiny glitches in the NEXT
> release of TA Bugfix.
>
> Last but not least are 2 TA Demos that you might
> find amusing to watch. Both
> were done on TA v3.1 - so TA Bugfix should not be
> used to replay them.
>
> David
--------
No problem.  Thanks for your help with the files, as you can tell, units are not my speciality :)

By the way, in addition to you noticing problems with the ARMAP (the CORE one works though), have you ever seen Queller build Levelers?  I can't get it to make the damn things, and I have no idea why.  It must be something I've done because it would build them in early versions.

Oh and I look forward to watching the demo of you in action.

Paul
--------
Subject: AI_Air.ufo
Date: Mon, 15 Nov 1999 22:50:27 +0000 (GMT)

I'm afraid I can't use this file, for the reason you yourself pointed out, it doesn't supported bugfixed planes!!!

As you seem to know more about these files, how should I add the BF planes?  The download files can't support more than 5 entries, so should I make an Air 3?

I also spotted your QuellerUnits error and was going to e-mail you about it, until I remembered that you'd already spotted it yourself and sent me a corrected file.......nuts :)

I also have to violently disagree with your limits!  2 floating heavy lasers???  Fat chance!  A lot of default maps have water, and the AI needs thse because it doesn't waste time on ships (generally because the water doesn't go anywhere).  However these lasers also cover land.

99 Geos?  Sure encourage it, but up the weight don't give it an insanse limit that will lead to kamikaze units.

10 Instigators and 8 Raiders?  I think not.  That would get the AI pounded on maps like Greenhaven, what's it gonna do with those Slashers?  Smash hills into submission?

10 Cans?  I'd rather it had more Cans than Sumo's, smaller, faster.

Limit Fluxor 0?  Hah!  If they're silly enough not to remove that despite the fact the readme tells them to move em........well they deserve what's commin' to 'em :)

6 Fusions...........no :)  I think it needs spares, and enough to fire guns, but 6?  That's overkill in my opinion.

9 Guadians but only 2 LLT's?  This is Queller not MH, it can't afford the Guardians, this is because it is more laser orientated than MH.

Snipe 2?  What's it gonna beat a Krogoth with?  Force of will?  I gave it 5 because....well they have a crummy reload time, but are excellent for picking things off, and they get hit by players quick, it pays to have spares.  Building another after you lost it is too late.

Oh, and I still hate your layout :)

I will admit I have a weakness in that for some reason I always round to the nearest 5......hmmmmmm.  I think some limits could be changed and have planned it, though I wouldn't do what you have.  2 Ambushers?  The player will feel fear now if he tries to Flea rush <g>

Ah the wonder of arguing AI :)

Paul
--------
"Ah, you're interfering with Queller again :)  Wait till I find fault with Mostly Harmless, I'll let you know ASAP, heh :)"

I'd be surprised if you found MH more faulty than EMG Man did in his review of MH on your website. The 0 rating on ai recovery just stung -- that's something I worked especially hard on... it's like being told your very best wasn't even worthy of an F+ grade. Getting a final overall score of "6" didn't help -- I felt like he simply gave me that without MH actually meriting such a high score because he felt sorry for me.

Now, to explain my harsh assessments made in my eariler email... I refer back to roughly a week ago:

Because I'd recently uploaded my latest MH ai, I decided it was time to do the same with my Counterstrike ai (with my newest TA Bugfix loaded, v1.3 which I'd just released) -- which I hadn't touched in over 2 weeks. Halfway through one of my tests (after making minor changes to the ai profiles) I noticed for some odd reason the ai's hawks and vamps were only firing 1 missile apiece or firing both of their missiles superimposed on top of each other -- this is not uncommon if the unit's launchpoint is the same for both weapons and both weapons are identical. After almost an hour of watching stealths fire their missiles during that game I concluded they were most certainly firing only one missile. And then I realized with dread that not only was that mistake in TA Bugfix v1.3 but that I'd uploaded it to TAMEC and told quite a few people (such as Cavedog's boneyards manager Paradox who said he was interested in TA Bugfix because it might help Boneyard's lag problem) they could download it from there.

I then quickly uncompressed all of TA Bugfix and started reading armhawk.fbi looking to see if I forgot to edit the 2nd weapon to be the same as the first. It was, and I was totally baffled. So I did what I always do when I'm totally baffled -- I ran a few more tests.

While I was away from the computer, I then remembered that the "hawks" I was seeing were *NOT* part of TA Bugfix and were in fact the ai-only hawks in Counterstrike (CS) ai -- which offered me a bit of relief since this meant I hadn't made a BIG mistake (as far as I was concerned) in TA Bugfix v1.3. Once I realized that the problem was with Counterstrike's stealths and not TA Bugfix's, I created a new fixed Ai_Air.ufo (the one I sent you) in under 5 minutes.

By this point, I was thoroughly sick of messing with my CS ai, so I decided to take a look at something considerably different: Queller v10, which I'd recently downloaded. (Particularly for viewing the TA Demo of it in action!) When I first messed with your newest Queller v10 since installing it, I didn't expect that there'd be anything wrong with it -- other than a differing opinion than mine on what's the "best" balance in the ai profiles. I was letting 2 ai's tear into each other for a while and they'd spread across the map. As the battle turned more towards Core's favor when Core got its powerful level 2 ground units out, I started wondering when Arm's advanced aircraft were going to show up and turn the tide. It never happened. No wonder Core wins often, Arm doesn't get an Adv. Air Plant!

Anyhow, my first assumption was it was a FLUKE that Arm's Adv. Air Plant wasn't built. After testing 3 more times in VERY LONG GAMES (2+ hours each) of me vs 1 Arm ai (and all I did was build a helper and hole up in a corner!), I concluded that something must be wrong with the newest version of TA Bugfix (v1.3) because this had never happened before. <--Which sent me back into the panic mode I was in a day or so earlier. (see above for why) So I was looking at TA Bugfix's C-Air's FBI file, the Adv. Air Plant's FBI file, and a lot of download files and GUI files. No luck at finding the problem. So I removed TA Bugfix and repeated the test without it. 2 long tests later, I was at least sure TA Bugfix wasn't to blame -- and was once again relieved. So, I concluded it was either how I recompiled TOTALA1.HPI or part of Queller. Considering I'd been running TA without these problems not too long ago, I checked Queller first -- checking TOTALA1.HPI would be a bitch, I was saving that for last. :p

First, I then looked over sidedata.tdf in detail by compairing it to a similar sidedata.tdf (the one in TA Bugfix) line-by-line, but couldn't find ANYTHING that seemed at fault.

Because of that I thought it was something to do with the ai profiles -- maybe a missing number after a Limit or Weight screwed up the whole profile, or an old entry of Limit ARMAAP 0 was left over from the "easy" setting or something. Because your ai profile used tabs after each unit name but before the number, I thought there was some slim chance that it could somehow be the problem. Also there were a couple odd places where you use 2 or more tabs or a tab and a space (instead of 1 space) after each unit name, and I thought that TA may have mistaken the second space or tab as a "null" (equal to 0) value for that weight or limit. So I edited ALL the ai profiles so they used only 1 space after each unit name, then ran another long test game. Nothing changed, the problem was still there.

It was only then that I tried "extreme" changes to your ai profile such as using my template for limits, and filling in the values your ai profiles used -- thereby hoping to eliminate any possible tabs where spaces are expected or even a shifted space (which is created by holding down shift while hitting spacebar.) Another reason I did this was so I would know where all the unit limits were quickly without having to search for them in your ai profiles (whose organizational order I only sorta know, because I haven't edited them much.) I ran another long test game. Nothing changed, the problem was still there.

Then I tried moving the default weights to below the other weight lines. In the test of that, the ai built a lot of OTHER units I wasn't seeing. So I thought I'd "solved" the problem, since I was now seeing the occassional (but still rare) Flea and other units that your ai had limited to 2-5. At this point I thought I finally had proof that the default weights do something nasty to the ai -- but still no sign of the Arm Advanced Air Plant in my next test!

Back to the drawing board -- and I was running out of ideas!

I had a farfetched idea that maybe it was a weight combination -- such as when you set weights REALLY high on one item and VERY low on another, the low item never gets built. But strangely enough, the REVERSE was happening -- Arm's Adv. Air Plants were weighted higher than ALL the other factories combined. So, I could only guess that maybe it was *ANOTHER* 2's compliment error (which means that when numbers are too large, they become negative -- thus worth less than other seemingly-smaller numbers) in TA's programming. An example of this error is when you go from 32,000 kills to 33,000 kills -- at 32,767 kills, getting 1 more kill makes your total kills -32,768! Or, as might be the case with single-byte values, adding 1 to 255 results in an answer of 0 instead of 256! I thought something like this was occuring with your ARMAAP weights. So I fudged around with the weights (lowering them to something similar to the other limits) and ran a few more tests...

No luck there, and by this point I was majorly frustrated: I removed Queller and tried messing around a bit with my MH ai. It wasn't long (maybe a day) before I went back to trying to figure out what's wrong with Queller ai because I thought I should know why Queller ai wasn't working -- and I didn't! Simply put, I was feeling stupid -- and when I feel stupid I get careless: when I reinstalled Quller to run some more tests I forgot to install sidedata.tdf! In that test, the problem had mysteriously vanished. Since I didn't at first realize that I'd forgotten the sidedata.tdf file, I replaced my massively modified default.txt for Queller with the ORIGINAL default.txt and ran another test to see if my final ai profile change had done the trick. Once again, the Arm Adv. Air Plant was built! However, just like before, many of the 0.05 weighted units were not built. What's more, Arm wasn't building Vehicle plants either -- even though I had QuellerUnits.ufo installed. So I thought that there was some sort of conflict between data in QuellerUnits.ufo or QuellerCCUnits.ufo and what was in sidedata.tdf. Then, when I exited TA and went to check sidedata.tdf again for something I may have missed the first time, I realized I'd forgotten to install it! I installed it, ran another test game -- and the Arm Adv. Air Plant wasn't built. Thinking that maybe my assumption of conflicts between ufo files and sidedata.tdf might be correct, I did a complete HPI viewer dump of QuellerUnits.ufo, and went through each part of it looking for any clues I could find on the nature of the conflict. Which lead me to an unrelated issue:
That was the FIRST time I'd seen ABFHAWK.FBI.

Do you remember getting annoyed about Cavedog releasing the Core Immolator because it meant you had to re-edit ALL your ai profiles? I felt very similarly about ABFHAWK, because I'm trying to make MH at least barely compatible with Queller's special ai-only units. (How else am I going to compair the 2 head-to-head?)

Perhaps you *DID* send me a message weeks ago about this change to the ai-only stealths (found in Queller and Counterstrike) so they would work properly with TA Bugfix, but it could've been when your email was all messed up -- because I never remember getting it, nor do I remember reading about this change on the TA message boards I occassionally visit. (I visit them only slightly more often than I post messages there, and considering I post VERY few messages on some of them this may explain why I didn't see your post where you mentioned this...)

Anyway, I also removed the weight ABFHAWK entries in the default.txt I sent you. After fixing the original (AAIHAWK), I felt there was no need for it.

In the process of reworking QuellerUnits.ufo so that sidedata.tdf wasn't needed, apparently I had forgotten to save the modified AAIVP0.TDF before repacking QuellerUnits.ufo -- this is the error in the first copy I sent you. Levelers should be buildable now too. :p

I removed all the aircraft from QuellerUnits.ufo for fear there might be something I overlooked since my luck (and yours too unfortunately) hasn't seemed too good lately on finding errors. I substituted my new Ai_Air.ufo in its place -- which added being able to disable aircaft in Queller at will. (This is why my CS ai has all the separate Ai_*.ufo files instead of just 1 ai_only.ufo file -- so people can tweak the difficulty of the ai by adding and removing what they don't want.)

Even after all these changes, I think there might still be some deep-rooted error in either TA Bugfix, Queller, or possibly TA itself that's causing the glitches and occassional slowdowns in the game. Because I've had savegames fail to load on occassion, I think the problem may be a bad cob file somewhere because that is what Rex222 said caused problems with TAUIP and savegames. And unit cobs are where my talent is very weak and my only tools are KNOWN to be very buggy.

All the above described above has taken me about 1 week with roughly 1-4 hours spent on it each day. Much of my tests were done with savegames, which I have yet to erase. I have a Yerrot Mountains savegame with Queller that has over 20 hours on the game clock and I have about 15,000 kills. I did all this hacking because I found a bug and I cannot stand to not fix it.

I did not do this to rub your nose in problems (real or imagined) with Queller ai -- and claim my MH is better because of that. Nor WILL I say that without many more ai vs ai tests where mine wins (which it may not) -- and even that temporary advantage will only last till you improve Queller further!

Also, the reason why the Queller ai profile I sent you ends up more like mine than yours was I didn't bother to convert it back to your ai table structure. I might have bothered to change it all back to your organizational method if I could do so quickly, but it took roughly 30 minutes of tedious work to convert it in the first place -- and going the OTHER way is naturally more difficult for me, just as the reverse would be for you. I don't want you to USE this ai profile in Queller, I only intended it to be an example of what happens when default weights are moved (not removed) to the bottom of the weights list. In my opinion, the results are surprising.
--------
Me-->"I've not seen a Flea, AK, Zipper, Pelican, Freaker, Gimp, Jeffy, Weasal, Triton, Lugar, Spider, Croc, Reaper, Titan, or Lancet yet."

"Good, you shouldn't, they are possibly the most useless units in the history of clone (or pattern) kind for an AI.  They're just in there for kicks, someone may use the 5000 unit patch :)"

The original nuke silos, antinukes, radars, and the like win the award for useless to the AI. These units are at worst LESS useful than other units the ai can make, they are by no means useless. It's my opinion for instance that the Pelican kicks butt on land even when used by the ai. The main reason why most are missing from my MH ai has to do with MH being a metal-miser who is not supposed to do anything that's less than 100% effective. If I could get the ai's resource model to support greater production levels, I'd probably add back some of these "useless" units in the hopes they could be built when the ai needs them most.

The limits set on Dominators in MH is the same as in Queller - 10. I'm just more stingy about what tags along with the Dominators, so they seem to become common rather quickly. Long-ranged attackers is a very important addition to an attack force, and best added as soon as possible. Morties are meant to be even more common than Dominators for this reason. However, a trio of Dominators can kill a single nonmoving unit probably in one volley. At best, if you're not watching that unit, all you'll get is "unit such-and-such is under attack." By the time you hit F3, it will be dead -- and you may not know where it died. If you have radar, you'll probably spot the ai's attack force (assuming it isn't jammed...) but the first-strike attack is almost worth the inconvenience of using poor infighters. Arm uses Merls for this, but since Core "specializes" a bit more in Kbots I thought it made good sense to use the Dominators instead since they can be built far sooner than Diplomats.

In my latest MH, I have lowered the limit to 8 and dropped the weight a little -- as a result of reading the review of MH where it got poor marks on your website. I also made the ai a little less construction-unit happy, but still just as intent on expanding in a hurry. I also made it less interested in making lots of factories at the same time, until somewhat later in the game when the ai wants to build ~5-20 advanced air plants at once. (LORD help you after it completes them all!) This results in the ai sometimes having a small surplus of metal. This solves all but the vulnerable to early raids problem that were described in the review of MH. But that is a problem that even BAI on hard has on small maps, so I don't feel too badly if I can't fix that "problem" in MH! On the other hand, Counterstrike doesn't ussually have this problem because it has its first mobile attack unit built in as little as 1 1/2 minutes. That is if it doesn't build an air plant first. :p

Your ai is set to build over 150 kbots alone... I have yet to see it do that tho, even with me giving it lots of resources with +ATM over and over again.

That's how come I concluded the default weights do something nasty to your ai -- the ai engine is probably using failsafes for weights since the default weightings in your profiles "burned out" the later entries.
--------
Can I take it that this is a fixed version? :P

I reccomend tranquilisers.....lots of them :)

OK, thanks.  Of course, people might get all upset
that in normal TA these planes don't act the same as a
normal Hawk or Vamp, some people get upset at these
minor differences (then I tell 'em where to shove it)
<g>

Paul

> >Subject: AI_Air.ufo
> >Date: Mon, 15 Nov 1999 22:50:27 +0000 (GMT)
>
> >
> >I'm afraid I can't use this file, for the reason
> you
> >yourself pointed out, it doesn't supported bugfixed
> >planes!!!
>
> I went to a bit of trouble to FIX it!!!
> THIS IS THE FIXED VERSION!!!!
>
> David
--------
Woah, long story, point by point again:

> "Ah, you're interfering with Queller again :)  Wait
> till I find fault with
> Mostly Harmless, I'll let you know ASAP, heh :)"
>
> I'd be surprised if you found MH more faulty than
> EMG Man did in his review
> of MH on your website. The 0 rating on ai recovery
> just stung -- that's
> something I worked especially hard on... it's like
> being told your very best
> wasn't even worthy of an F+ grade. Getting a final
> overall score of "6"
> didn't help -- I felt like he simply gave me that
> without MH actually
> meriting such a high score because he felt sorry for me.

The problem is he reviewed MH first so he hasn't got a baseline to go by.  Personally I think MH is very impressive.  It plays a construction game and yet, within about 7-8 minutes gets this huge army from litterally no where.

I think you'll find that once EMG has tried other AI's he'll reevaluate his opinion.  THe problem is he does say he finds it easy to beat.  MH does have a vunerable period early on (IMHO) when it's first getting its con units out, and EMG is usually raiding at this point, so maybe he does find it easy.  I on the other hand am the man who lost 2 Jeffies by parking them next to an enemy Vehicle Plant, one after the other, repating the mistake.  DOH!

I think I gave MH a 7.  I don't know if that seems low, but I'm not sure what Queller would get (EMG would probably give it 9, and I think I'd give it an 8....probably because I gave you a 7 :P).  However the 7 was an averaging of certain aspects, it is very aggressive, but it is also vunerable to raiding, as you stated in your readme it doesn't try to defend it just attacks (and damn well too).

> Now, to explain my harsh assessments made in my
> eariler email... I refer
> back to roughly a week ago:
>
> Because I'd recently uploaded my latest MH ai, I
> decided it was time to do
> the same with my Counterstrike ai (with my newest TA
> Bugfix loaded, v1.3
> which I'd just released) -- which I hadn't touched
> in over 2 weeks. Halfway
> through one of my tests (after making minor changes
> to the ai profiles) I
> noticed for some odd reason the ai's hawks and vamps
> were only firing 1
> missile apiece or firing both of their missiles
> superimposed on top of each
> other -- this is not uncommon if the unit's
> launchpoint is the same for both
> weapons and both weapons are identical. After almost
> an hour of watching
> stealths fire their missiles during that game I
> concluded they were most
> certainly firing only one missile. And then I
> realized with dread that not
> only was that mistake in TA Bugfix v1.3 but that I'd
> uploaded it to TAMEC
> and told quite a few people (such as Cavedog's
> boneyards manager Paradox who
> said he was interested in TA Bugfix because it might
> help Boneyard's lag
> problem) they could download it from there.
>
> I then quickly uncompressed all of TA Bugfix and
> started reading armhawk.fbi
> looking to see if I forgot to edit the 2nd weapon to
> be the same as the
> first. It was, and I was totally baffled. So I did
> what I always do when I'm
> totally baffled -- I ran a few more tests.
>
> While I was away from the computer, I then
> remembered that the "hawks" I was
> seeing were *NOT* part of TA Bugfix and were in fact
> the ai-only hawks in
> Counterstrike (CS) ai -- which offered me a bit of
> relief since this meant I
> hadn't made a BIG mistake (as far as I was
> concerned) in TA Bugfix v1.3.
> Once I realized that the problem was with
> Counterstrike's stealths and not
> TA Bugfix's, I created a new fixed Ai_Air.ufo (the
> one I sent you) in under
> 5 minutes.

I caught onto this when after installing the bugfx and fiddling with Queller I ran my usual Conflict Crusher, just to make sure all the Unit ID's were tidy. However I had some new maps so ran the feature and weapon orphan checker, and that came up with the Hawk and Vamp missing weapon.  So I slapped together some new ones.  I was tempted to modify the originals, but
I thought people might get upset that the Hawk's second missile was now about 3 something faster than theirs.

It took me 20min to track down all the files I missed before they worked :)

> By this point, I was thoroughly sick of messing with
> my CS ai, so I decided
> to take a look at something considerably different:
> Queller v10,

You mean considerably better of course :)

> which I'd
> recently downloaded. (Particularly for viewing the
> TA Demo of it in action!)

Wasn't it fun watching an AI play such a great game :)
I was damn happy that it didn't make any really noticable screw ups (except for the fact that I guess it was probably in mild nano-stall for the seonc half of the game).  I just wish the air AI was better, those Rapiers could have gutted EMG's base in 5 minutes.  Mind you it was great when they found his Comm wasn't it :)

> When I first messed with your newest Queller v10
> since installing it, I
> didn't expect that there'd be anything wrong with it
> -- other than a
> differing opinion than mine on what's the "best"
> balance in the ai profiles.

Ah, the problems.  I've been trying to figure these out for ages.  The two items you later mention stopped working a while ago, the problem is I introduced new units AND the bugfix at the same time, so I didn't know which was the problem and hadn't put too much effort into finding out (till now).

> I was letting 2 ai's tear into each other for a
> while and they'd spread
> across the map. As the battle turned more towards
> Core's favor when Core got
> its powerful level 2 ground units out, I started
> wondering when Arm's
> advanced aircraft were going to show up and turn the
> tide. It never
> happened. No wonder Core wins often, Arm doesn't get
> an Adv. Air Plant!

Depressing isn't it.  I'd watched hundreds of games and never seen one.  It really begins to get to you.

> Anyhow, my first assumption was it was a FLUKE that
> Arm's Adv. Air Plant
> wasn't built. After testing 3 more times in VERY
> LONG GAMES (2+ hours each)
> of me vs 1 Arm ai (and all I did was build a helper
> and hole up in a
> corner!), I concluded that something must be wrong
> with the newest version
> of TA Bugfix (v1.3) because this had never happened
> before. <--Which sent me
> back into the panic mode I was in a day or so
> earlier. (see above for why)
> So I was looking at TA Bugfix's C-Air's FBI file,
> the Adv. Air Plant's FBI
> file, and a lot of download files and GUI files. No
> luck at finding the
> problem. So I removed TA Bugfix and repeated the
> test without it. 2 long
> tests later, I was at least sure TA Bugfix wasn't to
> blame -- and was once
> again relieved. So, I concluded it was either how I
> recompiled TOTALA1.HPI
> or part of Queller. Considering I'd been running TA
> without these problems
> not too long ago, I checked Queller first --
> checking TOTALA1.HPI would be a
> bitch, I was saving that for last. :p
>
> First, I then looked over sidedata.tdf in detail by
> compairing it to a
> similar sidedata.tdf (the one in TA Bugfix)
> line-by-line, but couldn't find
> ANYTHING that seemed at fault.

I have done this countless times praying that I'd misspelt half the file again :)

> Because of that I thought it was something to do
> with the ai profiles --
> maybe a missing number after a Limit or Weight
> screwed up the whole profile,
> or an old entry of Limit ARMAAP 0 was left over from
> the "easy" setting or
> something. Because your ai profile used tabs after
> each unit name but before
> the number, I thought there was some slim chance
> that it could somehow be
> the problem. Also there were a couple odd places
> where you use 2 or more
> tabs or a tab and a space (instead of 1 space) after
> each unit name, and I
> thought that TA may have mistaken the second space
> or tab as a "null" (equal
> to 0) value for that weight or limit. So I edited
> ALL the ai profiles so
> they used only 1 space after each unit name, then
> ran another long test
> game. Nothing changed, the problem was still there.
>
> It was only then that I tried "extreme" changes to
> your ai profile such as
> using my template for limits, and filling in the
> values your ai profiles
> used -- thereby hoping to eliminate any possible
> tabs where spaces are
> expected or even a shifted space (which is created
> by holding down shift
> while hitting spacebar.) Another reason I did this
> was so I would know where
> all the unit limits were quickly without having to
> search for them in your
> ai profiles (whose organizational order I only sorta
> know, because I haven't
> edited them much.) I ran another long test game.
> Nothing changed, the
> problem was still there.
>
> Then I tried moving the default weights to below the
> other weight lines. In
> the test of that, the ai built a lot of OTHER units
> I wasn't seeing. So I
> thought I'd "solved" the problem, since I was now
> seeing the occassional
> (but still rare) Flea and other units that your ai
> had limited to 2-5. At
> this point I thought I finally had proof that the
> default weights do
> something nasty to the ai -- but still no sign of
> the Arm Advanced Air Plant
> in my next test!
>
> Back to the drawing board -- and I was running out
> of ideas!

As I did.

> I had a farfetched idea that maybe it was a weight
> combination -- such as
> when you set weights REALLY high on one item and
> VERY low on another, the
> low item never gets built. But strangely enough, the
> REVERSE was happening
> -- Arm's Adv. Air Plants were weighted higher than
> ALL the other factories
> combined. So, I could only guess that maybe it was
> *ANOTHER* 2's compliment
> error (which means that when numbers are too large,
> they become negative --
> thus worth less than other seemingly-smaller
> numbers) in TA's programming.
> An example of this error is when you go from 32,000
> kills to 33,000 kills --
> at 32,767 kills, getting 1 more kill makes your
> total kills -32,768! Or, as
> might be the case with single-byte values, adding 1
> to 255 results in an
> answer of 0 instead of 256! I thought something like
> this was occuring with
> your ARMAAP weights. So I fudged around with the
> weights (lowering them to
> something similar to the other limits) and ran a few
> more tests...

Can't say I've had gams get that many kills :)  I give up after an hour usually.

> No luck there, and by this point I was majorly
> frustrated: I removed Queller
> and tried messing around a bit with my MH ai. It
> wasn't long (maybe a day)
> before I went back to trying to figure out what's
> wrong with Queller ai

That's because it's so loverble :)

> because I thought I should know why Queller ai
> wasn't working -- and I
> didn't! Simply put, I was feeling stupid -- and when
> I feel stupid I get
> careless: when I reinstalled Quller to run some more
> tests I forgot to
> install sidedata.tdf! In that test, the problem had
> mysteriously vanished.

Man I wished I'd cocked up too, well we can't all make mistakes :)

> Since I didn't at first realize that I'd forgotten
> the sidedata.tdf file, I
> replaced my massively modified default.txt for
> Queller with the ORIGINAL
> default.txt and ran another test to see if my final
> ai profile change had
> done the trick. Once again, the Arm Adv. Air Plant
> was built! However, just
> like before, many of the 0.05 weighted units were
> not built. What's more,
> Arm wasn't building Vehicle plants either -- even
> though I had
> QuellerUnits.ufo installed. So I thought that there
> was some sort of
> conflict between data in QuellerUnits.ufo or
> QuellerCCUnits.ufo and what was
> in sidedata.tdf. Then, when I exited TA and went to
> check sidedata.tdf again
> for something I may have missed the first time, I
> realized I'd forgotten to
> install it! I installed it, ran another test game --
> and the Arm Adv. Air
> Plant wasn't built. Thinking that maybe my
> assumption of conflicts between
> ufo files and sidedata.tdf might be correct, I did a
> complete HPI viewer
> dump of QuellerUnits.ufo, and went through each part
> of it looking for any
> clues I could find on the nature of the conflict.
> Which lead me to an
> unrelated issue:
> That was the FIRST time I'd seen ABFHAWK.FBI.

Dum dum dum.

> Do you remember getting annoyed about Cavedog
> releasing the Core Immolator
> because it meant you had to re-edit ALL your ai
> profiles? I felt very
> similarly about ABFHAWK, because I'm trying to make
> MH at least barely
> compatible with Queller's special ai-only units.
> (How else am I going to
> compair the 2 head-to-head?)

Heh, yeah I posted a whine thread :)  Luckily though Queller wasn't finished, and the Immolator wasn't a game changer, so it wasn't too much of a problem.  All these people telling Cavedog they want more units, and all these AI designers say no, no more, we've finished our profiles :)  What's more apprently CD missed the limits section of the Immolator so if it ain't in your profile, you'll see several million on the map.

> Perhaps you *DID* send me a message weeks ago about
> this change to the
> ai-only stealths (found in Queller and
> Counterstrike) so they would work
> properly with TA Bugfix, but it could've been when
> your email was all messed
> up -- because I never remember getting it, nor do I
> remember reading about
> this change on the TA message boards I occassionally
> visit. (I visit them
> only slightly more often than I post messages there,
> and considering I post
> VERY few messages on some of them this may explain
> why I didn't see your
> post where you mentioned this...)

I don't know how I contacted you.  After making one I had to track down your e-mail address again because my address book was down.

> Anyway, I also removed the weight ABFHAWK entries in
> the default.txt I sent
> you. After fixing the original (AAIHAWK), I felt
> there was no need for it.

Sob, my poor unit pack (see AI Central downloads page) :)

> In the process of reworking QuellerUnits.ufo so that
> sidedata.tdf wasn't
> needed, apparently I had forgotten to save the
> modified AAIVP0.TDF before
> repacking QuellerUnits.ufo -- this is the error in
> the first copy I sent
> you. Levelers should be buildable now too. :p

I look forward to seeing them.

> I removed all the aircraft from QuellerUnits.ufo for
> fear there might be
> something I overlooked since my luck (and yours too
> unfortunately) hasn't
> seemed too good lately on finding errors. I
> substituted my new Ai_Air.ufo in
> its place -- which added being able to disable
> aircaft in Queller at will.
> (This is why my CS ai has all the separate Ai_*.ufo
> files instead of just 1
> ai_only.ufo file -- so people can tweak the
> difficulty of the ai by adding
> and removing what they don't want.)

I'll probably put them back.  I prefer to have set conditions because that way you know the AI will perform.  Anyway for those who don't want them they can use Quell which comes without hacked units.

> Even after all these changes, I think there might
> still be some deep-rooted
> error in either TA Bugfix, Queller, or possibly TA
> itself that's causing the
> glitches and occassional slowdowns in the game.
> Because I've had savegames
> fail to load on occassion, I think the problem may
> be a bad cob file
> somewhere because that is what Rex222 said caused
> problems with TAUIP and
> savegames. And unit cobs are where my talent is very
> weak and my only tools
> are KNOWN to be very buggy.

Yes, it's called Boneyards :)  I've had saved game problems, sometimes one whole side will disapper, but I don't know if that's Queller or the bugfix or TA because they're all loaded at the same time.

> All the above described above has taken me about 1
> week with roughly 1-4
> hours spent on it each day. Much of my tests were
> done with savegames, which
> I have yet to erase. I have a Yerrot Mountains
> savegame with Queller that
> has over 20 hours on the game clock and I have about
> 15,000 kills. I did all
> this hacking because I found a bug and I cannot
> stand to not fix it.

Ouch!  Well you're certinally methodical.  Glad you fixed it, I'm a perfectionist and it was beginning to get to me that Queller wouldn't go adv air.

> I did not do this to rub your nose in problems (real
> or imagined) with
> Queller ai -- and claim my MH is better because of
> that. Nor WILL I say that
> without many more ai vs ai tests where mine wins
> (which it may not) -- and
> even that temporary advantage will only last till
> you improve Queller
> further!

I can tell you from countless games that MH beats Queller more than Queller beats MH.  When I first played them off against one another I went into a panic.  What was the point of making an AI when a superior one existed?  If I'd found Mostly Harmless before the BSRAI I may never have made Queller. Because of this I got EMG Man to take on MH, he said he found it a lot easier to beat than Queller, and so I rested.  I guess they will appeal to players depending on how they play.  I don't know how you think of them, but I personally believe Queller plays a better early game, but a weaker longer one (I'm still trying to get fusion and MMM's working together, something MH does very nicely indeed).  This reflects the fact that I like the game for 30min, then begin to tire.  By 60min I usually stop anyway.  I also find Lv2 harder to test because so many factors in the early game influence the late.  However I am working on the profiles so they are a bit more single minded in the late game, more Mav's and Panthers, less Penetrators and Shooters.  Having near solely Golaith's works for the CORE though :)  I just wish I could get Cans and Sumos out more.  however, MH will always do that better because it seem (again, IMHO) geared towards moving for levels 2 and 3, rather than level 1.

However I don't usually take notice of my AI verus yours anymore (for one mine loses :)), MH has a vunerable period early on when it's all construction, however another AI will not attack until MH has entered it's get troops like mad, period.  A player will raid before then.  This is why I feel Queller can quite fairly compete with MH on a level playing field (if not with the "Switeck" brand name that goes with it :))

> Also, the reason why the Queller ai profile I sent
> you ends up more like
> mine than yours was I didn't bother to convert it
> back to your ai table
> structure. I might have bothered to change it all
> back to your
> organizational method if I could do so quickly, but
> it took roughly 30
> minutes of tedious work to convert it in the first
> place -- and going the
> OTHER way is naturally more difficult for me, just
> as the reverse would be
> for you. I don't want you to USE this ai profile in
> Queller, I only intended
> it to be an example of what happens when default
> weights are moved (not
> removed) to the bottom of the weights list. In my
> opinion, the results are
> surprising.

I will admit I didn't like the changes, they turned Queller into Mostly Harmless, an inferior Mostly Harmless.  Queller is geared for a level 1 supported by level 2 military game, the Mostly Harmless style didn't work with it's weights which didn't give it enough resources.

I also tried taking out the Weight ARM and Weight CORE (though it's not a high priority to me), but construction was too heavy.  However I have yet to try it with construction units moved to the bottom of their factory.  Removing these weights also adds the nasty 2 factories one after the other beginning game which almost every AI I know suffers.  I have tried to avoid that with Queller because it can't stop a raid if it falls into that trap.  Of course, if I'd seen MH before BSRAI I may never have used those weights at all, and construction wouldn't be at the top of every queue either.  Old habits are hard to break :)

I doubt I'll change a lot now.  A few limits, I had been thinking about moving Flashes down to 20 before you suggested it, but then that may require a reweighting and I wasn't sure I wanted to do that. However I will be playing with the limits more.

My Queller TA and CC work is going to have to be put on hold for a while though.  I'm trying to get Queller UH together because it now has a deadline.  Eek!

Paul Cocker aka Quitch(y)
--------
Subject: Bugfix..........some of it anyway 
Date: Tue, 16 Nov 1999 19:37:45 +0000 (GMT) 

Well I've improved Queller's level 2 mix.  Many thanks indeed for your help with the gamedata file, it was wonderful watching Brawlers and Hawks tear through the sky!  All I need to do now, is try to get Fusions and MMM's working in tandem.  I'm damn proud of Queller now, and I've done a few limit adjustments to help it work better with the 250 limit.  Beautiful ;).

I did try removing the Weight ARM and CORE lines, however this turned Queller into a construction obsessive AI again.  When it comes down to it, Queller was worked around these lines, and they make Queller what it is (damn amazing <g>).  By the way, in my last e-mail I don't think I dealt with you Weights 0.05 comments.  Queller really shouldn't EVER build these
units. However I couldn't bring myself to turn off units that the AI could actually do something with. Since the profile is so small, I kept them in because some insane person might have tghe Yankspankers 5000 unit patch. Some people say they have 4 hour epic games, so they may actually see these units.

By the way, as you unit test with others feel free to invite me :)  I can bring Queller, heh, save you have a boring old 3.1 AI with the other person (eek!).  By the way, you and a friend (can't remember who), can beat 3 Counterstrike AI's, yet in the start of that game against MH you weren't exactly routing it :).  I still lose to Queller with EMG, it's because he always has CORE as his opposite number, and every time the Goliaths appear he's routed and crushed, and I'm always left suddenly in a corner with two forces coming from two directions.  Mind you, his raids are better than mine, where I lose Jeffies to Stumpies :). Oh where can I put my face?

I also have some suggestions for your next bugfix (if you make one). These come from the UH fixes, I thought you might want to know about anything you've missed. BSR will probably have the answer to any questions you have about this list. My apologies if I include things you've already fixed.

Omissions from AA weapon damage sections (don't ask me what).

(Not sure if this is one you'd like), Hovercraft go to rubble if destroyed at sea, thus they don't block ships.

The unload cursor currently shows the wrong animation.

Decoy Comm, on his F1 help his build time is currently less than the real Comm's, thus making him easy to spot.

Hammer script bugs (?)

There is a model bug for the Thunder, it makes it's bombing runs less accurate.

Cloakable fusion, missing decloak radius.

Moho Metal Maker (both sides), if shot while vapour will activate, bug.  Solar Collectors also have this too.

CORE adv con Kbot has a script bug that means it loses build and repair orders.

ARM Wind Generator blades spin in the wrong direction.

FINAL QUESTIONS:

1. You haven't answered this one yet. Do you want your name on AI Central made into an e-mail link?

2. What was the difference between bugfix 1.3 and 1.2. You should add a "What's new!" file, so that we can all bitch "I downloaded this for THAT!!!!?" :P

The ever smug and overuser of smiley faces:

Paul
--------
"Omissions from AA weapon damage sections (don't ask me what)."
That would be the lack of special damage to seaplanes. I have run many tests and done quite a bit of math and have concluded that what Cavedog did was give Seaplanes HALF as much armor as they would otherwise have to make up for missing special damage lines on all AA weapons. This makes Seaplanes VERY easy to kill with flakkers, since their hitpoints are very low compaired to other aircraft. Adding special damage lines in AA units FOR the seaplanes will eradicate any use whatsoever of using them if REGULAR aircraft are available, UNLESS ALL seaplanes have their armor at least doubled. I'd rather keep it like it is -- low armor, no special AA entries -- the game will behave roughly the same (except seaplanes will have extra vulnerability to flakkers) and may be a little faster because it doesn't have to check as many special weapon damages each time an AA weapon is fired. If you take a good look at TA Bugfix's special ai-only aircraft, they have HALF the armor of the originals for the same reason. THUS, the way to fight them is to build flakkers which do equal damage to ALL air units. This is especially unfair for the ai-only brawlers and rapiers because flakkers were already their worst enemy, but there's no way around it short of a rewrite for AA weapons damages for TA Bugfix *AND* a rewrite of the ai-only air units in Counterstrike. It could however use the ORIGINAL Brawlers and Rapiers, but then they'd lose the ability to find targets further off their patrol route.

"(Not sure if this is one you'd like), Hovercraft go to rubble if destroyed at sea, thus they don't block ships."
I do NOT know how to impliment it in such a way that the heaps are left on land but not in water. I want hover corpses to block passage on land, which is how it is. Except in extremely long games on water or very constricted maps would I see it as a major problem. The ship corpses are far worse because they're so much bigger. I think I can cut the ship's corpes' footprints so it is easier to get around them and possibly allow more wrecks to be packed closer together. Then a ship graveyard can become REALLY crowded. :)

"The unload cursor currently shows the wrong animation."
I'll "borrow" the appropriate fix from UH.

"Decoy Comm, on his F1 help his build time is currently less than the real Comm's, thus making him easy to spot."
I'd have made that exact change LONG ago except there's a big issue of someone using this as a cheat. It is not difficult to add a download.tdf file for the commander, and if the REAL commander costs no more than a decoy commander it would be VERY tempting to use it!

"Hammer script bugs (?)"
It appears to not recoil its guns correctly -- since it still works, I'm not even going to TRY to rescript it with my buggy tools and limited ability.

"There is a model bug for the Thunder, it makes it's bombing runs less accurate."
I almost consider that a price and penalty inherent in using that unit. I never knew the Thunder had much of a bombing problem other than it drops few bombs per run. I think I'll live with it the way it is. Core's bomber is better, too bad.

"Cloakable fusion, missing decloak radius."
Which seems to have absolutely no effect on gameplay from what I can tell. All units that cloak only seem to uncloak when either they TOUCH an enemy unit, fire their weapon/s, or energy levels drop below what's neccessary to keep them cloaked. This bug is only on the Arm Cloaking Fusion, and even it decloaks JUST FINE when an aircraft flies right over it -- considering how big its footprint is, that isn't too hard to do!

"Moho Metal Maker (both sides), if shot while vapour will activate, bug.  Solar Collectors also have this too."
Already fixed, ad nausium.

"CORE adv con Kbot has a script bug that means it loses build and repair orders."
Already fixed. BSR heard about the bug from ME (and others) btw.

"ARM Wind Generator blades spin in the wrong direction."
I'll "borrow" the appropriate fix from UH.

"FINAL QUESTIONS:"
"1. You haven't answered this one yet. Do you want your name on AI Central made into an e-mail link?"
YES! PLEASE put my email address linked to my nickname.

"2. What was the difference between bugfix 1.3 and 1.2. You should add a "What's new!" file, so that we can all bitch "I downloaded this for THAT!!!!?" :P"
I honestly don't remember everything I do. You could compair the old readme to the new one. Some of what I describe in only 100 or less words in the readme though can be MANY THOUSAND little changes -- as the case was for removal of bad catagories and bad targetting sections. (And I'm not exaggerating here when I say 1,000's of changes!) Suffice to say, it would be EXTREMELY hard for me to redo TA Bugfix from scratch -- or even to redo it if all I had was TA v3.1c and TA Bugfix's readme!

The BIG change I remember doing in v1.3 was removing all the replicating "damaged terrain" scar features -- these features had a corpse -- themselves -- which made their creation and destruction an INFINITE loop. This bug is actually closely related to the missing DT (holes in DT walls) which sometimes occur in multiplayer games. This "tiny" fix required me to edit almost every feature file in TA, TA:CC, TA:BT, and all the Cavedog downloadable maps.
--------
Subject: Re: Bugfix..........some of it anyway
Date: Wed, 17 Nov 1999 11:56:12 +0000 (GMT)

Your recordings didn't go very far did they :)  Every time the game was staring *FOOM!* your line went.
Heh.

More below:

Paul

> My replies in no particular order:
>
> "Omissions from AA weapon damage sections (don't ask
> me what)."
> That would be the lack of special damage to
> seaplanes. I have run many tests
> and done quite a bit of math and have concluded that
> what Cavedog did was
> give Seaplanes HALF as much armor as they would
> otherwise have to make up
> for missing special damage lines on all AA weapons.
> This makes Seaplanes
> VERY easy to kill with flakkers, since their
> hitpoints are very low
> compaired to other aircraft. Adding special damage
> lines in AA units FOR the
> seaplanes will eradicate any use whatsoever of using
> them if REGULAR
> aircraft are available, UNLESS ALL seaplanes have
> their armor at least
> doubled. I'd rather keep it like it is -- low armor,
> no special AA entries
> -- the game will behave roughly the same (except
> seaplanes will have extra
> vulnerability to flakkers) and may be a little
> faster because it doesn't
> have to check as many special weapon damages each
> time an AA weapon is
> fired. If you take a good look at TA Bugfix's
> special ai-only aircraft, they
> have HALF the armor of the originals for the same
> reason. THUS, the way to
> fight them is to build flakkers which do equal
> damage to ALL air units. This
> is especially unfair for the ai-only brawlers and
> rapiers because flakkers
> were already their worst enemy, but there's no way
> around it short of a
> rewrite for AA weapons damages for TA Bugfix *AND* a
> rewrite of the ai-only
> air units in Counterstrike. It could however use the
> ORIGINAL Brawlers and
> Rapiers, but then they'd lose the ability to find
> targets further off their
> patrol route.

I remember when I first got those AI aircraft (I was trying to work out what was special about them).  I thought you'd halved the damage so as to balance the LOS, so I doubled it on mine :)  When those Rapiers first ripped though my MT's with impunity, I knew something was wrong :)

> "(Not sure if this is one you'd like), Hovercraft go
> to rubble if destroyed
> at sea, thus they don't block ships."
> I do NOT know how to impliment it in such a way that
> the heaps are left on
> land but not in water. I want hover corpses to block
> passage on land, which
> is how it is. Except in extremely long games on
> water or very constricted
> maps would I see it as a major problem. The ship
> corpses are far worse
> because they're so much bigger. I think I can cut
> the ship's corpes'
> footprints so it is easier to get around them and
> possibly allow more wrecks
> to be packed closer together. Then a ship graveyard
> can become REALLY
> crowded. :)
>
> "The unload cursor currently shows the wrong
> animation."
> I'll "borrow" the appropriate fix from UH.
>
> "Decoy Comm, on his F1 help his build time is
> currently less than the real
> Comm's, thus making him easy to spot."
> I'd have made that exact change LONG ago except
> there's a big issue of
> someone using this as a cheat. It is not difficult
> to add a download.tdf
> file for the commander, and if the REAL commander
> costs no more than a decoy
> commander it would be VERY tempting to use it!

But isn't the biggest audience of the bugfix the skirmish one?  If they cheat.......well they're entitled to because it isn't really cheating if they do it against an AI (as long as they don't claim they won fiar and square).  Also, at LAN parties people usually play fair or monitor one another (at least at mine :)).  OK online it's harder, but you did rename
him so that the decoy couldn't be spotted, but then you leave in this so that he can be spotted?

> "Hammer script bugs (?)"
> It appears to not recoil its guns correctly -- since
> it still works, I'm not
> even going to TRY to rescript it with my buggy tools
> and limited ability.

Couldn't you "borrow" that too?  Mind you, the speedy opening of it's gun ports....is that part of the script?  If not then you could just use the UH script couldn't you?

> "There is a model bug for the Thunder, it makes it's
> bombing runs less
> accurate."
> I almost consider that a price and penalty inherent
> in using that unit. I
> never knew the Thunder had much of a bombing problem
> other than it drops few
> bombs per run. I think I'll live with it the way it
> is. Core's bomber is
> better, too bad.

What about cliffs?  Is that a TA bug?  If you drop bombs against something near the base of a cliff or hill, the bombers will start bombing too late and miss everytime (if they fly OVER the hill on the way in).

> "Cloakable fusion, missing decloak radius."
> Which seems to have absolutely no effect on gameplay
> from what I can tell.
> All units that cloak only seem to uncloak when
> either they TOUCH an enemy
> unit, fire their weapon/s, or energy levels drop
> below what's neccessary to
> keep them cloaked. This bug is only on the Arm
> Cloaking Fusion, and even it
> decloaks JUST FINE when an aircraft flies right over
> it -- considering how
> big its footprint is, that isn't too hard to do!
>
> "Moho Metal Maker (both sides), if shot while vapour
> will activate, bug.
> Solar Collectors also have this too."
> Already fixed, ad nausium.
>
> "CORE adv con Kbot has a script bug that means it
> loses build and repair
> orders."
> Already fixed. BSR heard about the bug from ME (and
> others) btw.

Yes, THUG keeps whinging about it ;).

> "ARM Wind Generator blades spin in the wrong
> direction."
> I'll "borrow" the appropriate fix from UH.
>
> "FINAL QUESTIONS:"
> "1. You haven't answered this one yet. Do you want
> your name on AI Central
> made into an e-mail link?"
> YES! PLEASE put my email address linked to my
> nickname.

I'll get it done tonight then.

> "2. What was the difference between bugfix 1.3 and
> 1.2. You should add a
> "What's new!" file, so that we can all bitch "I
> downloaded this for
> THAT!!!!?" :P"
> I honestly don't remember everything I do. You could
> compair the old readme
> to the new one. Some of what I describe in only 100
> or less words in the
> readme though can be MANY THOUSAND little changes --
> as the case was for
> removal of bad catagories and bad targetting
> sections. (And I'm not
> exaggerating here when I say 1,000's of changes!)
> Suffice to say, it would
> be EXTREMELY hard for me to redo TA Bugfix from
> scratch -- or even to redo
> it if all I had was TA v3.1c and TA Bugfix's readme!

Though no where near as big, unless I keep the Queller AI Update file constantly up to date, I forget what I've done.  I have a sneaking suspicion that some versions are almost the same because I go in, make radical changes, undo them, then think that I've been working on it long enough for a new test release :)

> The BIG change I remember doing in v1.3 was removing
> all the replicating
> "damaged terrain" scar features -- these features
> had a corpse -- themselves
> -- which made their creation and destruction an
> INFINITE loop. This bug is
> actually closely related to the missing DT (holes in
> DT walls) which
> sometimes occur in multiplayer games. This "tiny"
> fix required me to edit
> almost every feature file in TA, TA:CC, TA:BT, and
> all the Cavedog
> downloadable maps.

Eek!
--------
Subject: Re: New version of TA Bugfix out (v1.3) at TAMEC!
Date: Wed, 17 Nov 1999 23:31:53 +0000 (GMT)

What is in the old_weapons.ufo?  If I remember
rightly, the readme states that it can be deleted and
only contains weapons you've wipped.

For the AI problem, do you mean the files in the AI
subdirectory within the rev31 file are being
overridden?  If this is the problem, wouldn't it be
possible to introduce a file containing all the FBI's
(I assume this is where hard weights anf limits are
kept?) from the Totala1 file which have been
reweighted and relimited?

Paul
"Never underestimate the power of a Quitch"
--------
>What is in the old_weapons.ufo?  If I remember
>rightly, the readme states that it can be deleted and
>only contains weapons you've wipped.

Correct, it contains all the weapons that were removed by TA Bugfix. Some 3rd party units require and EXPECT these now-missing weapons to function correctly. So for them, I corrected the deleted weapons and stored the fix in a separate ufo file.

>For the AI problem, do you mean the files in the AI
>subdirectory within the rev31 file are being
>overridden?  If this is the problem, wouldn't it be
>possible to introduce a file containing all the FBI's
>(I assume this is where hard weights anf limits are
>kept?) from the Totala1 file which have been
>reweighted and relimited?

That's not the ai problem I mean -- the ai problem has to do with the file TOTALA.EXE CHOOSING TO IGNORE the weights in all my ai profiles whenever certain conditions occur. I reffered to this as "failsafe modes" in my Ai Tweaking Guide.
--------
Subject: AI_Air.ufo
Date: Mon, 15 Nov 1999 22:50:27 +0000 (GMT)

"I also have to violently disagree with your limits!  2 floating heavy lasers???  Fat chance!  A lot of default maps have water, and the AI needs these because it doesn't waste time on ships (generally because the water doesn't go anywhere).  However these lasers also cover land."
I need to correct that mistake... I probably should've set floating laser max to 0. Those lasers cannot be built by anything other than c-hovers, c-seaplanes, and c-ships. Unless I'm mistaken, your ai doesn't make any of these on a land map. :p

"99 Geos?  Sure encourage it, but up the weight don't give it an insanse limit that will lead to kamikaze units."
Ok, insane though it might sound -- the ai still will NOT regularly build geos and will NOT look more than a screen or 2 out normally (except occassionaly with c-airs) to find them. It doesn't seem that the ai hunts down thermal vents the same way it does metal spots, as a matter of fact the ai tends to ignore them -- even with me giving it stupid orders to suicidally build on ALL of them! If a thermal vent was in or near your base, presumeably you'd build a geo on it yourself because geothermal plants have the highest energy-to-metal-cost ratio of any high-output resource. For Wind generators to even TIE the Geothermal plant, they would have to make +24 energy/second *CONSTANTLY*! Solars don't even RATE on this scale. Even fusions are only 2/5ths as good, metal-wise. You are a complete TA munchkin if you can think of ANY energy-provider that ever gives nearly as much energy/sec per point of metal cost that geos do. (I only know of 2 myself!) This includes anything you would have to turn off to use as an energy source, like radar towers. Plus, if there's a geothermal ALREADY built on the vent, (or anything covering the vent) the ai will *NOT* send c-units out to build on it. With that all said, in an EXTREMELY long game where the ai has done EVERYTHING else it possibly can with its basic c-units, yes the ai might send kamakazi c-airs out to build on far-away thermal vents. But chances are if it gets to that point you are already losing. :p

I will admit this value on geos might have interesting results on Pincushion, considering the number of thermal vents there, but that's the breaks! And who knows, the ai might get 10+ moho metal makers on that map as well -- and have the energy (thanks to all the geos) to run them. You can probably only guess how many attack units per minute it can make with that amount of resources. :)

"10 Instigators and 8 Raiders?  I think not.  That would get the AI pounded on maps like Greenhaven, what's it gonna do with those Slashers?  Smash hills into submission?"
With the overbuild bug, what's set at 10 might end up 11 or as much as 20! 8 might end up 9 to 15. By the time the ai's in FULL OVERRUN-YOUR-BASE MODE, it probably still won't have reached that pathetically low limit. And if it's not overruning your base by the time it reaches those limits, it's my opinion the ai can better spend its resources elsewhere! It's quite possible that the player has DT'ed his/her base so that the ai cannot reach it, in which case making more ground forces is pretty futile except for extra AA defense.

As you already know, the resources required for a nonmetal map nonhacked ai to reach the level-1 unit maximums is FAR greater than Queller can do within an hour. (At least if you want it to do much else, that is!) Heck, even MH (when left alone)- which has very low unit limits compaired to Queller - cannot get enough resources to reach its level-1 unit maximums in less than 30 minutes despite building resources at a suicidal pace. More level-1 factories wouldn't help it either, because it's almost always draining resources faster than it can make more. If the ai cannot get the resources, it won't be able to turn that into something nasty to throw at the player. Likewise, if the ai cannot KEEP the resources it won't keep fight effective -- which is MH's problem with early raids.

"10 Cans?  I'd rather it had more Cans than Sumo's, smaller, faster."
Ai Cans are very easy to kill with micromanagement -- you can easily avoid their short-range laser (FAR shorter than Sumo's!) and pick it apart with missiles and rockets. With a Sumo, you pretty much have to kill the thing from the air or avoid it -- thank God it's so danged slow! The reason why I emphasize Cans less than you is I fight the ai by trying to fight from a distance with lots of missile units. If I prefered in close fights, I'd feel differently.

"6 Fusions...........no :)  I think it needs spares, and enough to fire guns, but 6?  That's overkill in my opinion."
Each Moho Metal Maker needs most of a fusion for itself, (Arm's more so than Core's) and the overbuild bug is very common on Moho Metal Makers. Each bertha certainly needs a piece of a fusion. Each factory and c-unit ties up some resources as well -- advanced air plants need more than most, and their limits are the highest. HLT's and Annihilator lasers soak up a lot of energy while they are firing. (You DO want to increase the limit on the number of Arm Shooters, don't you? :p) Ai-only nuke silos instantaneously drain 18,000 energy per shot when they fire -- it takes a while to fill that back up with only a +100 energy surplus! Plus, if Vulcans and Buzzsaws are not limited to 0, they take 8 fusions and 4 fusions respectfully EACH (MORE once they're at extreme vetran status) to sustain full firing rate. I'd say, 6 fusions may not be enough!

"9 Guadians but only 2 LLT's?  This is Queller not MH, it can't afford the Guardians, this is because it is more laser orientated than MH."
Other than the rare DT'ed LLT, I don't believe I've ever seen a LLT get vetran status -- not when I'm fighting an ai, nor when I'm fighting versis other players. And if it doesn't get vetran status, it quite likely doesn't even pay for itself. If I find a dt'ed LLT, I ussually go around it and/or hit it with ballistic weapons or Merls. If it's not DT'ed, I just gather my force in a tight clump and charge it -- if it's by itself, it'll be lucky to get more than 1 kill. 1 Flash tank for a dead LLT sounds like a fair trade to me! There's also another reason why I don't like the ai building them -- the commander can build them. The ai shoots itself in the foot wasting enough time between build projects as it is, especially with the VERY valuable fast-building commander (which all too soon goes into his stupid wander-in-circles mode.) I'd rather the ai get another factory, HLT, or just more attack units than to have its commander spend 1/2 of a critical resource-gathering minute building a LLT. Another reason why I don't like LLT's: if the LLT is in range to fire, the enemy is too close!

HLT's aren't nearly as bad for the ai as LLT's (especially Arm's HLT which has greater range than Core's HLT) because they can cover a larger area and are harder to kill. Unlike LLT's, by the time your ground troops SEE an ai HLT they are probably taking hits from it unless terrain is blocking. And 1 HLT can kill units slightly faster than 1 LLT as well. :) The Core Viper is sort of a HLT/LLT in that it has less range and damage as a HLT (GAAT GUN!) but it costs less energy per shot to fire.

Guardians are a disaster if built too early and a double disaster if not built later on for the ai. They eat a huge amount of metal, but if placed near other defenses (or manage to get partially protected by the ai's attack force, which is common when the ai's base is under attack) it can rip up huge numbers of enemy attackers. These base guns have such long range that even placed randomly they can often offer fire-support to units 1 screen away. I weight them low -- if not very low. To me, they're as important as berthas to the ai. It doesn't surprise me to see the ai's guardian have more kills than its bertha -- guardians don't MISS nearly as badly! Once Krogoths or massive ground armies show up, it's my opinion you can NEVER have too many guardians.

"Snipe 2?  What's it gonna beat a Krogoth with?  Force of will?  I gave it 5 because....well they have a crummy reload time, but are excellent for picking things off, and they get hit by players quick, it pays to have spares.  Building another after you lost it is too late."
The Arm Shooter doesn't have a ballistic weapon like the elegant Core Morty. Also, the original ai Shooter never cloaks. If it did either, I'd give it a MUCH higher rating. In the event the ai has them to stop a krogoth, they will most likely only get 1 shot before they die. They only have 320 armor and cost over 900 metal and take a long time to build compaired to other adv. kbots -- the Krogoth's longest-ranged weapon is 50% longer ranged than their laser and can kill them with 3 shots. Plus, if the ai's shooters are facing another ai's krogoth -- the krogoth will have no problem "seeing" them and killing them. (Ai has permanent radar.) On an open field, the ai Krogoth will waste them. in hilly terrain, the ai's krogoth's rockets can shoot over low hills and proabably still kill them. Against a player, the shooter will shoot at the first available target it sees which will probably be something fast like a level 1 unit... (You did send escorts with your krogoth?) Plus, the reload time on their laser is 20 game-seconds -- which makes hitting with their first shot even more important, it may be the only shot they get! Since even level 1 missile-carrying ground units have the same range as the shooter's weapon, the ai shooter will possibly die after killing 1 level 1 unit. Maybe the ai shooter got lucky and killed a sumo with its one shot before it died, but even a sumo costs less than a shooter.

The only things I know that can beat a krogoth short of base defenses is LOTS and LOTS of units. Spiders don't hurt, if they can get lucky. Merls in huge numbers can waste an ai's krogoth because the krogoth will quit moving probabaly after each kill -- just enough time for a volley of merl rockets to arrive. Also, the metal value of a krogoth-stopping force need only be about 1/4 to 1/2 the cost of a krogoth -- even in the ai's crappy hands! Also, if the krogoth is quickly killing the ai's ENTIRE attack force it is still quite likely the ai has berthas (or intimidators) pounding it. I've seen krogoths die from ai bertha fire without getting a single kill! By the time a krogoth shows up, it's not unusual for the ai to have at least 1 extreme-vetran (40+ kills) bertha in its base that has aquired the annoying habit of killing metal extractors on the first shot. Krogoths are slightly bigger targets. :)

Play Counterstrike ai -- 1 Arm ai (your ally!) versis 1 Core ai on Painted Desert, and get them on opposite corners of the map (with fixed starts) so neither side can use berthas on each other's base. Now remove ai-only nukes (it tends to mess up the results :p) AND remove Ai-only aircraft (the new file :p) so you don't get a rude interruption. Set the gamespeed to +10 and hide your commander in the Arm ai's base. Krogoths will PROBABLY show up in such a game within as little as 40 minutes gametime. And they WILL die too, unless Arm's already losing. And this is an OPEN map -- where Mr. Kroggy is at his best! All you have to do is figure out what it is that's doing most of the krogoth-killing, because even I'm not sure how to fight krogoths -- I have virtually no experience!

"Oh, and I still hate your layout :)"
As I've said, weights seem more effective if high ones are on top and low ones are on bottom -- at least relative to each other. The limits section is sorted alphabetically, first by type then by unitname. The reason why I didn't separate the base defense weapons from resource buildings is because they're on the same build menus for the c-units -- and there's so few of each of them. I presume you like Cavedog's layout better?

"I will admit I have a weakness in that for some reason I always round to the nearest 5......hmmmmmm.  I think some limits could be changed and have planned it, though I wouldn't do what you have."
I tailor ai's I modify to play at least vaguely similar to how I play -- that is if the exe cooperates. (MH is supposed to make far more missile towers far more often than it does! After all, only solars and geos have a higher weight.)

"2 Ambushers?  The player will feel fear now if he tries to Flea rush <g>"
That could be any number you want, up to 40 -- so long as their weight is very low. Personally, I'd rather have the ai building a fusion, air repair pad, or a bertha than an ambusher -- but that's just me. Ambushers aren't nearly as tough when closed as Core Toasters are. They also only have about 3/4 the range of a guardian, even though they cost more and get built at a faster rate becaise adv. c-units have roughly double the buildrate of regular c-units. The faster buildrate means the build costs get drained from the ai's resources EVEN QUICKER than a guardian does. (This is not good for an ai which is prone to nano-stall to begin with, such as Cavedog v3.1 ai...) Other than occassionally shooting down Brawlers and barely surviving nukes that hit it (the nuke probably won't leave much nearby for the Ambusher to defend though ;), I do not know of any particular ability that makes Ambushers considerably more valuable than Guardians.

"Ah the wonder of arguing AI :)"
Might as well be arguing politics considering almost every bit of it is strictly opinion and half-assed ideas. Neither one of us has ever seen TA's ai source code -- and I doubt we could understand it if we did!

"I thought you'd halved the damage so as to balance the LOS, so I doubled it on mine :)  When those Rapiers first ripped though my MT's with impunity, I knew something was wrong :)"
When making changes to TA with ufo files, you're pretty much stuck with however AA units already treats aircaft. I may not LIKE the limitations this puts on me as a designer of Counterstrike (ai-aircraft are paper to flakkers), but it's either that or redesign TA Bugfix's weapons to handle ALL ai-only units. (Which would only slow the game down, whether the ai-only units were installed or not!)

"But isn't the biggest audience of the bugfix the skirmish one?  If they cheat.......well they're entitled to because it isn't really cheating if they do it against an AI (as long as they don't claim they won fiar and square).  Also, at LAN parties people usually play fair or monitor one another (at least at mine :)).  OK online it's harder, but you did rename
him so that the decoy couldn't be spotted, but then you leave in this so that he can be spotted?"
If you're busy pressing F1 on every "commander" you see, the decoy has already gone a little ways toward paying for itself. Plus, my biggest audience for TA Bugfix is me! If me decides they're sick of TA Bugfix, I'll more than likely quit working on it even if everyone else begs me not to. :p Like you said though, if they want to cheat in skirmish they can and will. But I don't want to make it cheap to build commanders because I am still wanting to finish the Arm campaign I've started that's included with TA Bugfix. In it, at one point or another, at least 1 side *DOES* start manufacturing Commanders again. For the campaign to work, even as an idea, these commanders need to be expensive in every way. I could make decoy commanders as expensive as real commanders, but then I doubt anyone would build them. :p

> "Hammer script bugs (?)"
> "There is a model bug for the Thunder,
I'll be looking into how UH fixes unit scripts QUITE CLOSELY!

"What about cliffs?  Is that a TA bug?  If you drop bombs against something near the base of a cliff or hill, the bombers will start bombing too late and miss everytime (if they fly OVER the hill on the way in)."
Yes, that's a TA bug -- I believe it's a problem with the way the physics engine inside TOTALA.EXE decides when units are to shoot. That can't be fixed with a script change to my knowledge.

> "CORE adv con Kbot has a script bug that means it
> loses build and repair orders."

"Yes, THUG keeps whinging about it ;)."
Whinging? don't you mean whine-ing? That's a big problem for CORE too, in unfixed TA. Couple that with other bugs such as the Pelican bug, and it seems little doubt which side CAVEDOG favors. :p

> YES! PLEASE put my email address linked to my
> nickname.
"I'll get it done tonight then."
Don't hurt yourself in your hurry to do so. :p

"Though no where near as big, unless I keep the Queller AI Update file constantly up to date, I forget what I've done.  I have a sneaking suspicion that some versions are almost the same because I go in, make radical changes, undo them, then think that I've been working on it long enough for a new test release :)"
It's easy to do. I try to keep notes on all the changes I make to TA Bugfix, but with the way I brainstorm and how fast I work, I sometimes forget what I'm doing (or TRYING to do!) mid-stream.

>Your recordings didn't go very far did they :)  Every
>time the game was staring *FOOM!* your line went.
And you can imagine how mad I got each time that happened...
Ah well, at least the demos are small as a result. :p

The Counterstrike ai will NOT get a description from me after all. Fnordia wants me to update it so he can add recognition for its units in TA Demo Recorder. (IE: such-and-such as Counterstrike ai installed!) Also, there are so many ways that I consider the version of Counterstrike ai on your website a failure. Suffice to say, when the NEXT version of Counterstrike comes out I *WILL* give a long and short description of it.

David
--------
Ah the failsafes. Hmmmmm. Well speaking as a layman of such matters, as the safeguards kick in around about <100 M and E, is it possible to make TA think that the avlue 100 is 0?
--------
That is only ONE of the many failsafes.
And only by gettting the ai's resources there will that failsafe activate. HOWEVER, in the case of Moho Metal Makers, if the ai has vast energy storage it generally won't let energy drop to 0 before turning it off. This is only a rule-of-thumb and not a hard rule, but for some reason the ai is a little more careful with a MMMK when it has lots of energy storage.
--------
Sorry I meant to say *most* of the safeguards, not to imply all of them. The AI switches on the MMM long enough though so that each time it does it loses more energy than it makes, so eventually it stalls itself.
--------
Subject: Re: Bugfix..........some of it anyway
Date: Thu, 18 Nov 1999 12:32:07 +0000 (GMT)


> Subject: AI_Air.ufo
> Date: Mon, 15 Nov 1999 22:50:27 +0000 (GMT)
>
> "I also have to violently disagree with your limits!
>  2 floating heavy
> lasers???  Fat chance!  A lot of default maps have
> water, and the AI needs
> these because it doesn't waste time on ships
> (generally because the water
> doesn't go anywhere).  However these lasers also
> cover land."
> I need to correct that mistake... I probably
> should've set floating laser
> max to 0. Those lasers cannot be built by anything
> other than c-hovers,
> c-seaplanes, and c-ships. Unless I'm mistaken, your
> ai doesn't make any of
> these on a land map. :p

HA HA!  WRONG! (Sorry I get like that :)).  C-Planes can build them (and do).

> "99 Geos?  Sure encourage it, but up the weight
> don't give it an insanse
> limit that will lead to kamikaze units."
> Ok, insane though it might sound -- the ai still
> will NOT regularly build
> geos and will NOT look more than a screen or 2 out
> normally (except
> occassionaly with c-airs) to find them. It doesn't
> seem that the ai hunts
> down thermal vents the same way it does metal spots,
> as a matter of fact the
> ai tends to ignore them -- even with me giving it
> stupid orders to
> suicidally build on ALL of them! If a thermal vent
> was in or near your base,
> presumeably you'd build a geo on it yourself because
> geothermal plants have
> the highest energy-to-metal-cost ratio of any
> high-output resource. For Wind
> generators to even TIE the Geothermal plant, they
> would have to make +24
> energy/second *CONSTANTLY*! Solars don't even RATE
> on this scale. Even
> fusions are only 2/5ths as good, metal-wise. You are
> a complete TA munchkin
> if you can think of ANY energy-provider that ever
> gives nearly as much
> energy/sec per point of metal cost that geos do. (I
> only know of 2 myself!)
> This includes anything you would have to turn off to
> use as an energy
> source, like radar towers. Plus, if there's a
> geothermal ALREADY built on
> the vent, (or anything covering the vent) the ai
> will *NOT* send c-units out
> to build on it. With that all said, in an EXTREMELY
> long game where the ai
> has done EVERYTHING else it possibly can with its
> basic c-units, yes the ai
> might send kamakazi c-airs out to build on far-away
> thermal vents. But
> chances are if it gets to that point you are already
> losing. :p

I knocked them up, but not to 99 (it's tempting though)

> I will admit this value on geos might have
> interesting results on
> Pincushion, considering the number of thermal vents
> there, but that's the
> breaks! And who knows, the ai might get 10+ moho
> metal makers on that map as
> well -- and have the energy (thanks to all the geos)
> to run them. You can
> probably only guess how many attack units per minute
> it can make with that
> amount of resources. :)

Heh, wait for the metal safeguards to kick in though and send that 200 strong attack force home :)

> "10 Instigators and 8 Raiders?  I think not.  That
> would get the AI pounded
> on maps like Greenhaven, what's it gonna do with
> those Slashers?  Smash
> hills into submission?"
> With the overbuild bug, what's set at 10 might end
> up 11 or as much as 20! 8
> might end up 9 to 15. By the time the ai's in FULL
> OVERRUN-YOUR-BASE MODE,
> it probably still won't have reached that
> pathetically low limit. And if
> it's not overruning your base by the time it reaches
> those limits, it's my
> opinion the ai can better spend its resources
> elsewhere! It's quite possible
> that the player has DT'ed his/her base so that the
> ai cannot reach it, in
> which case making more ground forces is pretty
> futile except for extra AA
> defense.

What can I say, after defying you on thse issues, and went and changed most of them anyway :)

> As you already know, the resources required for a
> nonmetal map nonhacked ai
> to reach the level-1 unit maximums is FAR greater
> than Queller can do within
> an hour. (At least if you want it to do much else,
> that is!) Heck, even MH
> (when left alone)- which has very low unit limits
> compaired to Queller -
> cannot get enough resources to reach its level-1
> unit maximums in less than
> 30 minutes despite building resources at a suicidal
> pace. More level-1
> factories wouldn't help it either, because it's
> almost always draining
> resources faster than it can make more. If the ai
> cannot get the resources,
> it won't be able to turn that into something nasty
> to throw at the player.
> Likewise, if the ai cannot KEEP the resources it
> won't keep fight effective
> -- which is MH's problem with early raids.

Yes you see MH at 5 minutes, then look away, look back 5 min later and wondered if there's any terrain behind that flood of MH dots.

> "10 Cans?  I'd rather it had more Cans than Sumo's,
> smaller, faster."
> Ai Cans are very easy to kill with micromanagement
> -- you can easily avoid
> their short-range laser (FAR shorter than Sumo's!)
> and pick it apart with
> missiles and rockets. With a Sumo, you pretty much
> have to kill the thing
> from the air or avoid it -- thank God it's so danged
> slow! The reason why I
> emphasize Cans less than you is I fight the ai by
> trying to fight from a
> distance with lots of missile units. If I prefered
> in close fights, I'd feel
> differently.
>
I'm sticking with Cans.  Because of the sleep bug, the only time a Sumo gets near the front is when the AI is retreating (so maybe it's the back still? :)).

> "6 Fusions...........no :)  I think it needs spares,
> and enough to fire
> guns, but 6?  That's overkill in my opinion."
> Each Moho Metal Maker needs most of a fusion for
> itself, (Arm's more so than
> Core's) and the overbuild bug is very common on Moho
> Metal Makers. Each
> bertha certainly needs a piece of a fusion. Each
> factory and c-unit ties up
> some resources as well -- advanced air plants need
> more than most, and their
> limits are the highest. HLT's and Annihilator lasers
> soak up a lot of energy
> while they are firing. (You DO want to increase the
> limit on the number of
> Arm Shooters, don't you? :p) Ai-only nuke silos
> instantaneously drain 18,000
> energy per shot when they fire -- it takes a while
> to fill that back up with
> only a +100 energy surplus! Plus, if Vulcans and
> Buzzsaws are not limited to
> 0, they take 8 fusions and 4 fusions respectfully
> EACH (MORE once they're at
> extreme vetran status) to sustain full firing rate.
> I'd say, 6 fusions may
> not be enough!

8!!  ARG!  Maybe more fusions..........(currently at 5)

> "9 Guadians but only 2 LLT's?  This is Queller not
> MH, it can't afford the
> Guardians, this is because it is more laser
> orientated than MH."
> Other than the rare DT'ed LLT, I don't believe I've
> ever seen a LLT get
> vetran status -- not when I'm fighting an ai, nor
> when I'm fighting versis
> other players. And if it doesn't get vetran status,
> it quite likely doesn't
> even pay for itself. If I find a dt'ed LLT, I
> ussually go around it and/or
> hit it with ballistic weapons or Merls. If it's not
> DT'ed, I just gather my
> force in a tight clump and charge it -- if it's by
> itself, it'll be lucky to
> get more than 1 kill. 1 Flash tank for a dead LLT
> sounds like a fair trade
> to me! There's also another reason why I don't like
> the ai building them --
> the commander can build them. The ai shoots itself
> in the foot wasting
> enough time between build projects as it is,
> especially with the VERY
> valuable fast-building commander (which all too soon
> goes into his stupid
> wander-in-circles mode.) I'd rather the ai get
> another factory, HLT, or just
> more attack units than to have its commander spend
> 1/2 of a critical
> resource-gathering minute building a LLT. Another
> reason why I don't like
> LLT's: if the LLT is in range to fire, the enemy is
> too close!

I just weight them real low.  Sometimes having these units is useful because you can destract the AI construction units with a cheap task and let your resources recharge.

> HLT's aren't nearly as bad for the ai as LLT's
> (especially Arm's HLT which
> has greater range than Core's HLT) because they can
> cover a larger area and
> are harder to kill. Unlike LLT's, by the time your
> ground troops SEE an ai
> HLT they are probably taking hits from it unless
> terrain is blocking. And 1
> HLT can kill units slightly faster than 1 LLT as
> well. :) The Core Viper is
> sort of a HLT/LLT in that it has less range and
> damage as a HLT (GAAT GUN!)
> but it costs less energy per shot to fire.

I give HLT's the edge over Vipers, firstly because they fire slightly further (useful in random placing) and also hold 300 energy, which the Viper does not.

> Guardians are a disaster if built too early and a
> double disaster if not
> built later on for the ai. They eat a huge amount of
> metal, but if placed
> near other defenses (or manage to get partially
> protected by the ai's attack
> force, which is common when the ai's base is under
> attack) it can rip up
> huge numbers of enemy attackers. These base guns
> have such long range that
> even placed randomly they can often offer
> fire-support to units 1 screen
> away. I weight them low -- if not very low. To me,
> they're as important as
> berthas to the ai. It doesn't surprise me to see the
> ai's guardian have more
> kills than its bertha -- guardians don't MISS nearly
> as badly! Once Krogoths
> or massive ground armies show up, it's my opinion
> you can NEVER have too
> many guardians.

I'm still working on this one.  My Guardian weights are pretty much OK for Metal, but on normal maps......

> "Snipe 2?  What's it gonna beat a Krogoth with?
> Force of will?  I gave it 5
> because....well they have a crummy reload time, but
> are excellent for
> picking things off, and they get hit by players
> quick, it pays to have
> spares.  Building another after you lost it is too
> late."
> The Arm Shooter doesn't have a ballistic weapon like
> the elegant Core Morty.
> Also, the original ai Shooter never cloaks. If it
> did either, I'd give it a
> MUCH higher rating. In the event the ai has them to
> stop a krogoth, they
> will most likely only get 1 shot before they die.
> They only have 320 armor
> and cost over 900 metal and take a long time to
> build compaired to other
> adv. kbots -- the Krogoth's longest-ranged weapon is
> 50% longer ranged than
> their laser and can kill them with 3 shots. Plus, if
> the ai's shooters are
> facing another ai's krogoth -- the krogoth will have
> no problem "seeing"
> them and killing them. (Ai has permanent radar.) On
> an open field, the ai
> Krogoth will waste them. in hilly terrain, the ai's
> krogoth's rockets can
> shoot over low hills and proabably still kill them.
> Against a player, the
> shooter will shoot at the first available target it
> sees which will probably
> be something fast like a level 1 unit... (You did
> send escorts with your
> krogoth?) Plus, the reload time on their laser is 20
> game-seconds -- which
> makes hitting with their first shot even more
> important, it may be the only
> shot they get! Since even level 1 missile-carrying
> ground units have the
> same range as the shooter's weapon, the ai shooter
> will possibly die after
> killing 1 level 1 unit. Maybe the ai shooter got
> lucky and killed a sumo
> with its one shot before it died, but even a sumo
> costs less than a shooter.

Already done this, the Mav swarms are slightly bigger, Mavs and Panthers......Merls too.

> The only things I know that can beat a krogoth short
> of base defenses is
> LOTS and LOTS of units. Spiders don't hurt, if they
> can get lucky. Merls in
> huge numbers can waste an ai's krogoth because the
> krogoth will quit moving
> probabaly after each kill -- just enough time for a
> volley of merl rockets
> to arrive. Also, the metal value of a
> krogoth-stopping force need only be
> about 1/4 to 1/2 the cost of a krogoth -- even in
> the ai's crappy hands!
> Also, if the krogoth is quickly killing the ai's
> ENTIRE attack force it is
> still quite likely the ai has berthas (or
> intimidators) pounding it. I've
> seen krogoths die from ai bertha fire without
> getting a single kill! By the
> time a krogoth shows up, it's not unusual for the ai
> to have at least 1
> extreme-vetran (40+ kills) bertha in its base that
> has aquired the annoying
> habit of killing metal extractors on the first shot.
> Krogoths are slightly
> bigger targets. :)

You know what they say, big Krogoth, small.........

> Play Counterstrike ai -- 1 Arm ai (your ally!)
> versis 1 Core ai on Painted
> Desert, and get them on opposite corners of the map
> (with fixed starts) so
> neither side can use berthas on each other's base.
> Now remove ai-only nukes
> (it tends to mess up the results :p) AND remove
> Ai-only aircraft (the new
> file :p) so you don't get a rude interruption. Set
> the gamespeed to +10 and
> hide your commander in the Arm ai's base. Krogoths
> will PROBABLY show up in
> such a game within as little as 40 minutes gametime.
> And they WILL die too,
> unless Arm's already losing. And this is an OPEN map
> -- where Mr. Kroggy is
> at his best! All you have to do is figure out what
> it is that's doing most
> of the krogoth-killing, because even I'm not sure
> how to fight krogoths -- I
> have virtually no experience!

That's the sign of a good ARM player :)

> "Oh, and I still hate your layout :)"
> As I've said, weights seem more effective if high
> ones are on top and low
> ones are on bottom -- at least relative to each
> other. The limits section is
> sorted alphabetically, first by type then by
> unitname. The reason why I
> didn't separate the base defense weapons from
> resource buildings is because
> they're on the same build menus for the c-units --
> and there's so few of
> each of them. I presume you like Cavedog's layout
> better?

I like mine :)  You though sorting mine out was hell? Look at the MoonAI, he alternates the sides LINE BY LINE!!!!

> "I will admit I have a weakness in that for some
> reason I always round to
> the nearest 5......hmmmmmm.  I think some limits
> could be changed and have
> planned it, though I wouldn't do what you have."
> I tailor ai's I modify to play at least vaguely
> similar to how I play --
> that is if the exe cooperates. (MH is supposed to
> make far more missile
> towers far more often than it does! After all, only
> solars and geos have a
> higher weight.)

I just went the whole hog and weighted MT's to 100 <g>

> "2 Ambushers?  The player will feel fear now if he
> tries to Flea rush <g>"
> That could be any number you want, up to 40 -- so
> long as their weight is
> very low. Personally, I'd rather have the ai
> building a fusion, air repair
> pad, or a bertha than an ambusher -- but that's just
> me. Ambushers aren't
> nearly as tough when closed as Core Toasters are.
> They also only have about
> 3/4 the range of a guardian, even though they cost
> more and get built at a
> faster rate becaise adv. c-units have roughly double
> the buildrate of
> regular c-units. The faster buildrate means the
> build costs get drained from
> the ai's resources EVEN QUICKER than a guardian
> does. (This is not good for
> an ai which is prone to nano-stall to begin with,
> such as Cavedog v3.1
> ai...) Other than occassionally shooting down
> Brawlers and barely surviving
> nukes that hit it (the nuke probably won't leave
> much nearby for the
> Ambusher to defend though ;), I do not know of any
> particular ability that
> makes Ambushers considerably more valuable than
> Guardians.

It's more likely to hit it's intended target because it uses a Guass weapon.

> "Ah the wonder of arguing AI :)"
> Might as well be arguing politics considering almost
> every bit of it is
> strictly opinion and half-assed ideas. Neither one
> of us has ever seen TA's
> ai source code -- and I doubt we could understand it
> if we did!

I argue politics too <g>

> "I thought you'd halved the damage so as to balance
> the LOS, so I doubled it
> on mine :)  When those Rapiers first ripped though
> my MT's with impunity, I
> knew something was wrong :)"
> When making changes to TA with ufo files, you're
> pretty much stuck with
> however AA units already treats aircaft. I may not
> LIKE the limitations this
> puts on me as a designer of Counterstrike
> (ai-aircraft are paper to
> flakkers), but it's either that or redesign TA
> Bugfix's weapons to handle
> ALL ai-only units. (Which would only slow the game
> down, whether the ai-only
> units were installed or not!)

As long as the Comm wiggle didn't come back, eek!

> "But isn't the biggest audience of the bugfix the
> skirmish one?  If they
> cheat.......well they're entitled to because it
> isn't really cheating if
> they do it against an AI (as long as they don't
> claim they won fiar and
> square).  Also, at LAN parties people usually play
> fair or monitor one
> another (at least at mine :)).  OK online it's
> harder, but you did rename
> him so that the decoy couldn't be spotted, but then
> you leave in this so
> that he can be spotted?"
> If you're busy pressing F1 on every "commander" you
> see, the decoy has
> already gone a little ways toward paying for itself.
> Plus, my biggest
> audience for TA Bugfix is me! If me decides they're
> sick of TA Bugfix, I'll
> more than likely quit working on it even if everyone
> else begs me not to. :p
> Like you said though, if they want to cheat in
> skirmish they can and will.
> But I don't want to make it cheap to build
> commanders because I am still
> wanting to finish the Arm campaign I've started
> that's included with TA
> Bugfix. In it, at one point or another, at least 1
> side *DOES* start
> manufacturing Commanders again. For the campaign to
> work, even as an idea,
> these commanders need to be expensive in every way.
> I could make decoy
> commanders as expensive as real commanders, but then
> I doubt anyone would
> build them. :p
>
> >"Hammer script bugs (?)"
> >"There is a model bug for the Thunder,
> I'll be looking into how UH fixes unit scripts QUITE
> CLOSELY!
>
> "What about cliffs?  Is that a TA bug?  If you drop
> bombs against something
> near the base of a cliff or hill, the bombers will
> start bombing too late
> and miss everytime (if they fly OVER the hill on the
> way in)."
> Yes, that's a TA bug -- I believe it's a problem
> with the way the physics
> engine inside TOTALA.EXE decides when units are to
> shoot. That can't be
> fixed with a script change to my knowledge.
>
> >"CORE adv con Kbot has a script bug that means it
> >loses build and repair orders."
>
> "Yes, THUG keeps whinging about it ;)."
> Whinging? don't you mean whine-ing? That's a big
> problem for CORE too, in
> unfixed TA. Couple that with other bugs such as the
> Pelican bug, and it
> seems little doubt which side CAVEDOG favors. :p

I don't know, their AI was always better with CORE.

> >YES! PLEASE put my email address linked to my
> >nickname.
> "I'll get it done tonight then."
> Don't hurt yourself in your hurry to do so. :p

Well if you want to pay for the PEAK TIME RATE call that anything before the evening would cost.  Feel free.

> "Though no where near as big, unless I keep the
> Queller AI Update file
> constantly up to date, I forget what I've done.  I
> have a sneaking suspicion
> that some versions are almost the same because I go
> in, make radical
> changes, undo them, then think that I've been
> working on it long enough for
> a new test release :)"
> It's easy to do. I try to keep notes on all the
> changes I make to TA Bugfix,
> but with the way I brainstorm and how fast I work, I
> sometimes forget what
> I'm doing (or TRYING to do!) mid-stream.

What were we talking about? :)

> >Your recordings didn't go very far did they :)
> Every
> >time the game was staring *FOOM!* your line went.
> And you can imagine how mad I got each time that
> happened...
> Ah well, at least the demos are small as a result.
> :p

Yes that's one way of putting it.........

> The Counterstrike ai will NOT get a description from
> me after all. Fnordia
> wants me to update it so he can add recognition for
> its units in TA Demo
> Recorder. (IE: such-and-such as Counterstrike ai
> installed!) Also, there are
> so many ways that I consider the version of
> Counterstrike ai on your website
> a failure. Suffice to say, when the NEXT version of
> Counterstrike comes out
> I *WILL* give a long and short description of it.
>
> David
Failure eh, 40min Krog's isn't good enough? :)
--------
Please use HPI Viewer and TA to verify what you claim -- I do! Don't accept what others say as "fact", ESPECIALLY TA strategies. I've heard a lot of seeming-sound strategies that should work but generally don't (such as many methods for how to counter original Arm Pelicans) and I've also heard a couple hair-brained strategies that sorta work. (You ever heard of using a ship transport to counter Pelicans? -- Supposedly you can grab ENOUGH Pelicans with its loader arm, and it can hold enough pelicans internally, that it pays for itself. Combine that with taking a FULL transport back and capturing them individually, and it might be a doable strategy! Or load one crawling bomb in the ship first, and self-d it when it's about to die -- that'll kill any pelicans nearby that it missed!)

"HA HA!  WRONG! (Sorry I get like that :)).  C-Planes can build them (and do)."
Unless your ai is a hacked, cheating ai like BAI, C-Planes cannot build Floating HLT's. It could be you added such features to your sidedata.tdf -- but it is not a feature of the unmodified game. At least not for the players... :)

"I knocked them up, but not to 99 (it's tempting though)"
Just for the record, I *HAVE* seen c-airs wandering all over the map to cover every last Geo. But this was only about every other C-air, and the ai had almost maxed out on everything else the C-airs could build. An interesting side-effect of this was mini-bases would sometimes pop up around a geo. This meant that ground forces would join the main attack force from ODD directions, and would sometimes cut through MY mini-bases to do it!

"Heh, wait for the metal safeguards to kick in though and send that 200 strong attack force home :)"
If the ai has 10+ MMM's and enough Geos to run them, the safeguards won't kick in. However, if the ai had too few Geos, its metal AND energy would still be occassionally bouncing above 100 -- if not over 1,000 for the energy. Either way, the ai's attack force would probably still go out raiding, and wouldn't return home until beaten. And they'd be resupplied with NEW troops too, just not as often as they would if they had a steady resource income. +100metal/sec with some 0 energy halts and stops looks like ~+50Metal/sec effective outcome.

"> "10 Instigators and 8 Raiders?  I think not.  That"
"What can I say, after defying you on thse issues, and went and changed most of them anyway :)"
Our ai's could end up playing nearly identical, yet be made with a different philosophy behind them. What's more, they may PLAY identical but still be considerably different in their ai profiles! The numbers I use aren't set in stone -- I change them often: sometimes up, sometimes down, sometimes sideways. :p

ME: "Likewise, if the ai cannot KEEP the resources it won't keep fight effective -- which is MH's problem with early raids."
"Yes you see MH at 5 minutes, then look away, look back 5 min later and wondered if there's any terrain behind that flood of MH dots."
MH isn't *THAT* good, except maybe on a Metal map. I regularly outproduce it in resources on just about any map I play it on, however since it devotes so much of its resources for attack units I have a HARD time (as the demo/s showed) beating back its attack force because I cannot be strong enough to beat it back everywhere.

"I'm sticking with Cans.  Because of the sleep bug, the only time a Sumo gets near the front is when the AI is retreating (so maybe it's the back still? :))."
I've had good luck with Sumos, so I put a reasonably high limit on them -- but a LOW weight on them so they're not constantly dragging down the ai's metal. They're expensive Metal/Second wise! But Goliaths are even MORE expensive Metal/Second wise! (~18m/sec!) Maybe I need to emphasize Diplomats more? :)

"8!!  ARG!  Maybe more fusions..........(currently at 5)"
Overbuild bug applies to them TOO, fortunately -- it just seldom kicks in early. But if your ai ever makes a lot of adv. c-units, it can ALWAYS build more than 1 fusion at a time -- which is generally disasterous resource-wise! Mine has this problem already.

"I just weight them real low.  Sometimes having these units is useful because you can destract the AI construction units with a cheap task and let your resources recharge."
LLT's though are a high energy/sec drain even though they do build quickly, especially if the commander builds one! It'd be better if the ai's c-units spent a lot of time making missile towers (MT) instead. MT would be useful at any point in the game, poor LLT's are only good if the ai is probably losing anyway. :(

"I give HLT's the edge over Vipers, firstly because they fire slightly further (useful in random placing) and also hold 300 energy, which the Viper does not."
It is a bug on Cavedog's part that the Viper does not, IMHO. Even the LLT holds 100 energy. I'll change the Viper and Core Immolator (CORPLAS) to store 100 energy each in TA Bugfix. (I doubt anyone will give a shit about such a miniscule change one way or another...) However, the Core HLT only holds 200 energy - enough for ~2.6 shots - not 300 energy which would be enough for 4 shots. Also, a Core Viper only requires 25 energy per shot (1/3 that of the HLT!) so it's less of a drain when enemies come knocking.

"I'm still working on this one.  My Guardian weights are pretty much OK for Metal, but on normal maps..."
MH gets serious metal shortages from guardians, but it ussually only builds 1 or 2 at a time. Just enough for a VERY rude surprise for an attack force who thinks raiding will be easy...

"Already done this, the Mav swarms are slightly bigger, Mavs and Panthers......Merls too."
I think Mavs cost lots of energy/sec but are low on metal/sec. Dunno about the others. MH's "trick" to getting lots of units is to have factories producing some items that are a low drain on resource income. But once it gets advanced factories, it ussually goes to shit on metal.

"You know what they say, big Krogoth, small..."
In multiplayer, the general rule is someone who TRIES to build krogoths in a laggy online game will die if their opponent/s has a similar resource income and spend it in ANY other way... Even nukes, which are notoriously inefficient, are a better deal when used wisely -- I've had to resort to using my nuke silo on enemy attackers and outlying enemy defenses because my opponents had too many antinukes.

"That's the sign of a good ARM player :)"
What? Not knowing how to counter Krogoths?

"I like mine :)  You though sorting mine out was hell? Look at the MoonAI, he alternates the sides LINE BY LINE!!!"
I didn't say I hated yours, only that I didn't understand it all. Most of it is in order, but you still have the occassional unit that's not in alphabetical order.

"I just went the whole hog and weighted MT's to 100 <g>"
Because of the nasty metal and energy safeguards (failsafes, I also call them), the reason why MH makes so FEW MT's before the 15 minute mark is it is ussually out of either metal or energy -- thus it is constantly forced to build solars or metal extractors instead.

"It's more likely to hit it's intended target because it uses a Guass weapon."
Ambusher's weapon is VERY similar to the Guardian's weapon. It is not a gauss weapon. It has the same shot velocity, thus making it no more or less likely to hit fast-moving targets directly than the Arm Guardian's weapon. However, it does have a larger area of effect (100 vs 64) to make near-misses hurt more. It has a slightly higher reload rate, and does less damage per shot. In short, it is a LESS effective weapon!

"I argue politics too <g>"
I vote where I live, but that doesn't mean I like it. :p
"Cthulu (a demon, supposedly) for president! -- Let's stop voting for the LESSER of 2 evils for a change!!!"

"As long as the Comm wiggle didn't come back, eek!"
Presumeably somewhere between OTA and TA Bugfix, that bug would kick in again. TA Bugfix may have only "barely" fixed it, so even installing lots of maps or 3rd party units can set it off.

> seems little doubt which side CAVEDOG favors. :p
"I don't know, their AI was always better with CORE."
Well, that's because their ai coders apparently weren't listening to what the REST of the company was sasying. :p
In online games, as far back as v1.0, ARM simply swamped Core because people became more familiar with Arm than Core. And despite Core having some VERY NASTY level 2 units, Core never got to use them because of the dreaded (lag-inducing) flash rush. Even without lag, such as on a LAN, Flashes beat Instigators -- but not by much!

"Well if you want to pay for the PEAK TIME RATE call that anything before the evening would cost.  Feel free."
I'm using a crappy, but free internet provider now. And all my local calls are free, so I can call my ISP for free anytime of day -- assuming I don't get a busy signal. :p

"What were we talking about? :)"
The main thing that causes me to forget was like when I was editing categories and bad targetting sections in units -- with the 200+ units out there, it was easy to forget which unit I was on, which category types it should have, and which targets its weapons should generally ignore unless there are no other targets available.

"Yes that's one way of putting it..."
I'd prefer that the demos were just under 1,000kb compressed -- that's the largest file attachment HOTMAIL will allow me.

"Failure eh, 40min Krog's isn't good enough? :)"

Counterstrike is a failure because it is not reliable. While it may get a krogoth in one game in as little as 30 minutes, it may take an hour in the next regardless of whether you're pounding on it. As you've well noticed, sometimes it builds only 1 fluxor at start, sometimes 2, and sometimes 3 or more! These units are EXTREME in the cheating department, I want the ai to have 2, no more than 2, no less than 2 on the hard setting -- and I just cannot get that kind of reliability out of it.

Another way it is not reliable is nuke silos. I've seen the first nuke fired in just under 15 minutes. And other times, the ai starts building its first nuke silo after the 45 minute mark. 15 is too early, 45+ is too late.

Counterstrike has already diverged away from BAI. It did this a while back too, when I decided that making an ai STRICTLY to beat BAI was pointless and wasn't fun to play against. So I watered down the original Counterstrike to be something barely beatable by above-average players who used some but not all anti-ai tactics. A really GOOD player might be able to beat Counterstrike without using any anti-ai tactics at all. It is worth noting that just like BAI, CS ai is EXTREMELY hard to beat on certain maptypes - namely nonmetal maps with no metal spots to build on. Those Fluxors go far in handing CS ai a win!

TA Bugfix does *NOT* fix the invisible DT bug which occurs infrequently in multiplayer TA games. (where DT is only visible to the person who built them)

Packet loss is the cause of this bug -- the moment the DT is built, the completed DT packet gets lost in transit to the other computers.

TA Demo Recorder has a partial fix for this bug, which is at least 80% effective.

I don't know WHERE you read that it did -- or did you test it and see no missing DT in LAN games? If you've tested this repeatedly over the internet I might believe it, but without support, I cannot.
--------
Subject: Re: Bugfix..........some of it anyway
Date: Sat, 20 Nov 1999 01:35:57 +0000 (GMT)

P by P.

> Please use HPI Viewer and TA to verify what you claim -- I do!

Do you mean that DT thing?  If so, I got that from your e-mail I recalled you saying something about an INFIITE loop, and invisible DT's, so I thought it was one and the same.

>Don't accept what others say as "fact",

I run a clan and fight with the MSN Gaming Zone over technicalities, I never trust anyone online :)

> ESPECIALLY TA strategies. I've heard a lot of
> seeming-sound strategies that should work but
> generally don't (such as many methods for how to
> counter original Arm Pelicans) and I've also heard a
> couple hair-brained strategies that sorta work. (You
> ever heard of using a ship transport to counter
> Pelicans? -- Supposedly you can grab ENOUGH
> Pelicans with its loader arm, and it can hold enough
> pelicans internally, that it pays for itself.
> Combine that with taking a FULL transport back and
> capturing them individually, and it might be a
> doable strategy! Or load one crawling bomb in the
> ship first, and self-d it when it's about to die --
> that'll kill any pelicans nearby that it missed!)
>
> "HA HA!  WRONG! (Sorry I get like that :)).
> C-Planes can build them (and do)."
> Unless your ai is a hacked, cheating ai like BAI,
> C-Planes cannot build Floating HLT's. It could be
> you added such features to your sidedata.tdf --
> but it is not a feature of the unmodified game. At
> least not for the players... :)

I'm sure they were building them........sidedata again?

> "I knocked them up, but not to 99 (it's tempting
> though)"
> Just for the record, I *HAVE* seen c-airs wandering
> all over the map to cover every last Geo. But this
> was only about every other C-air, and the ai had
> almost maxed out on everything else the C-airs
> could build. An interesting side-effect of this was
> mini-bases would sometimes pop up around a geo. This
> meant that ground forces would join the main attack
> force from ODD directions, and would sometimes cut
> through MY mini-bases to do it!

Or more often your mini base would destory them cutting off the flow of reinforcements :)  Nothing like a Brawler meeting a Goliath in the middle of no where.

> "Heh, wait for the metal safeguards to kick in
> though and send that 200 strong attack force home :)"
> If the ai has 10+ MMM's and enough Geos to run them,
> the safeguards won't kick in. However, if the ai had
> too few Geos, its metal AND energy would still be
> occassionally bouncing above 100 -- if not over
> 1,000 for the energy. Either way, the ai's attack
> force would probably still go out raiding, and
> wouldn't return home until beaten. And they'd be
> resupplied with NEW troops too, just not as often as
> they would if they had a steady resource income.
> +100metal/sec with some 0 energy halts and stops
> looks like ~+50Metal/sec effective outcome.

If only it would build the damn things, you can never rely on those safeguards to do anything!

> "> "10 Instigators and 8 Raiders?  I think not. That"
> "What can I say, after defying you on thse issues,
> and went and changed most of them anyway :)"
> Our ai's could end up playing nearly identical, yet
> be made with a different philosophy behind them.
> What's more, they may PLAY identical but still be
> considerably different in their ai profiles! The
> numbers I use aren't set in stone -- I change them
> often: sometimes up, sometimes down, sometimes
> sideways. :p

Heh, I had to redo a lot of it once because my values were all starting to sneak into double figures :) Sometimes I make changes and don't realise the ratios have stayed the same :)  I doubt our AI's will become the same (till you see sense and bow to Queller <g>), because your AI will always expand like mad immediately, where as mine will attack ASAP, and play
to defeat raids too.

> ME: "Likewise, if the ai cannot KEEP the resources
> it won't keep fight effective -- which is MH's
> problem with early raids."
> "Yes you see MH at 5 minutes, then look away, look
> back 5 min later and wondered if there's any terrain
> behind that flood of MH dots."
> MH isn't *THAT* good, except maybe on a Metal map. I
> regularly outproduce it in resources on just about
> any map I play it on, however since it devotes so
> much of its resources for attack units I have a HARD
> time (as the demo/s showed) beating back its attack
> force because I cannot be strong enough to beat it
> back everywhere.

What I like to do is grab a main group, 1, and send it to fight the enemy army.  Then I get 5 Flash tanks and make them 2, and anything resource or con wise that the first group spots I send in number 2 to bag it. With the enemy army held up it works quite well (till I realise I've left my entire construction operation doing nothing :)).

> "I'm sticking with Cans.  Because of the sleep bug,
> the only time a Sumo gets near the front is when the
> AI is retreating (so maybe it's the back still? :))."
> I've had good luck with Sumos, so I put a reasonably
> high limit on them -- but a LOW weight on them so
> they're not constantly dragging down the ai's
> metal. They're expensive Metal/Second wise! But
> Goliaths are even MORE
> expensive Metal/Second wise! (~18m/sec!) Maybe I
> need to emphasize Diplomats more? :)

Don't bother, in OTA Goliaths are trundling death machines.  They barely need any support as is, only against the skies do they fail.  I always think of Sumos as more defensive, if the AI is winning they are always stuck at the back because of the AI sleep bug.

> "8!!  ARG!  Maybe more fusions..........(currently at 5)"
> Overbuild bug applies to them TOO, fortunately -- it
> just seldom kicks in early. But if your ai ever makes
> a lot of adv. c-units, it can ALWAYS build more than
> 1 fusion at a time -- which is generally disasterous
> resource-wise! Mine has this problem already.
>
> "I just weight them real low.  Sometimes having
> these units is useful because you can destract the
> AI construction units with a cheap task and let
> your resources recharge."
> LLT's though are a high energy/sec drain even though
> they do build quickly, especially if the commander
> builds one! It'd be better if the ai's c-units
> spent a lot of time making missile towers (MT)
> instead. MT would be useful
> at any point in the game, poor LLT's are only good
> if the ai is probably losing anyway. :(

Well in general it's probably a good thing for non-cheating AI's to build them then :)

> "I give HLT's the edge over Vipers, firstly because
> they fire slightly further (useful in random placing)
> and also hold 300 energy, which the Viper
> does not."
> It is a bug on Cavedog's part that the Viper does
> not, IMHO. Even the LLT holds 100 energy. I'll
> change the Viper and Core Immolator (CORPLAS) to
> store 100 energy each in TA Bugfix. (I doubt anyone
> will give a shit about such a miniscule change one
> way or another...) However, the Core HLT only
> holds 200 energy - enough for ~2.6 shots - not 300
> energy which would be enough for 4 shots. Also, a
> Core Viper only requires 25 energy per shot (1/3
> that of the HLT!) so it's less of a drain when
> enemies come knocking.

D'OH, me bad.

> "I'm still working on this one.  My Guardian weights
> are pretty much OK for Metal, but on normal maps..."
> MH gets serious metal shortages from guardians, but
> it ussually only builds
> 1 or 2 at a time. Just enough for a VERY rude
> surprise for an attack force
> who thinks raiding will be easy...

I think I may have stumbled onto the weight in a UH test.  Now if only the damn AI Commander would stop trying to "help".

> "Already done this, the Mav swarms are slightly
> bigger, Mavs and Panthers......Merls too."
> I think Mavs cost lots of energy/sec but are low on
> metal/sec. Dunno about the others. MH's "trick" to
> getting lots of units is to have factories
> producing some items that are a low drain on
> resource income. But once it gets advanced
> factories, it ussually goes to shit on metal.

I try to get Queller to get an Adv Kbot lab up ASAP, because Mavs are so dirt cheap (and take so damn long to build).

I've noticed something that I'm sure didn't happen in MH 11.7, the MH factories always seem to close up after about 2 units.  Less nano-stall but a lot of the time the MH factories are in safelimit territory.  I think it might benefit from one less at the beginning, or prehaps a slightly larger time spacing between them.

> "You know what they say, big Krogoth, small..."
> In multiplayer, the general rule is someone who
> TRIES to build krogoths in a laggy online game
> will die if their opponent/s has a similar resource
> income and spend it in ANY other way... Even nukes,
> which are notoriously inefficient, are a better
> deal when used wisely -- I've had to resort to
> using my nuke silo on enemy attackers and outlying
> enemy defenses because my
> opponents had too many antinukes.

I never get that far, I win with my first 5 Flashes (happens more than you'd think), or it's game over :)

> "That's the sign of a good ARM player :)"
> What? Not knowing how to counter Krogoths?

Not having met enough to know how.

> "I like mine :)  You though sorting mine out was
> hell? Look at the MoonAI,
> he alternates the sides LINE BY LINE!!!"
> I didn't say I hated yours, only that I didn't
> understand it all. Most of it
> is in order, but you still have the occassional unit
> that's not in
> alphabetical order.

I shove them in mmmmmmmmmmmmmmmyyyyyyyyy waaaaaayyyyyyyy!

> "I just went the whole hog and weighted MT's to 100
> <g>"
> Because of the nasty metal and energy safeguards
> (failsafes, I also call them), the reason why MH
> makes so FEW MT's before the 15 minute mark is it
> is ussually out of either metal or energy -- thus it
> is constantly forced to
> build solars or metal extractors instead.

Metal is such a pain, usually my problem is that Queller hits it's limit, but knock up the limit and it screws everything.  Change the weights and each MEX built knocks less off the total weight, and so the factories take longer in coming.  Sigh.

> "It's more likely to hit it's intended target
> because it uses a Guass weapon."
> Ambusher's weapon is VERY similar to the Guardian's
> weapon. It is not a gauss weapon. It has the same
> shot velocity, thus making it no more or less
> likely to hit fast-moving targets directly than the
> Arm Guardian's weapon. However, it does have a
> larger area of effect (100 vs 64) to make
> near-misses hurt more. It has a slightly higher
> reload rate, and does less damage per shot. In
> short, it is a LESS effective weapon!

They certinally cost a bomb in metal.

> "I argue politics too <g>"
> I vote where I live, but that doesn't mean I like
> it. :p
> "Cthulu (a demon, supposedly) for president! --
> Let's stop voting for the
> LESSER of 2 evils for a change!!!"

Ummmmmm.

> "As long as the Comm wiggle didn't come back, eek!"
> Presumeably somewhere between OTA and TA Bugfix,
> that bug would kick in
> again. TA Bugfix may have only "barely" fixed it, so
> even installing lots of
> maps or 3rd party units can set it off.

Luckily I'm a purist.

> >seems little doubt which side CAVEDOG favors. :p
> "I don't know, their AI was always better with
> CORE."
> Well, that's because their ai coders apparently
> weren't listening to what
> the REST of the company was sasying. :p

Or anyone else for that matter.

> In online games, as far back as v1.0, ARM simply
> swamped Core because people
> became more familiar with Arm than Core. And despite
> Core having some VERY
> NASTY level 2 units, Core never got to use them
> because of the dreaded
> (lag-inducing) flash rush. Even without lag, such as
> on a LAN, Flashes beat
> Instigators -- but not by much!

I remember that, I was always told not to use the Flash because of lag, but the Flash was ARM's main weapon!!!  I gave up in the end, and went back to good old lag free skirmishes.

> "Well if you want to pay for the PEAK TIME RATE call
> that anything before
> the evening would cost.  Feel free."
> I'm using a crappy, but free internet provider now.
> And all my local calls
> are free, so I can call my ISP for free anytime of
> day -- assuming I don't
> get a busy signal. :p

Well living in the UK we don't get free local calls :(

> "What were we talking about? :)"
> The main thing that causes me to forget was like
> when I was editing
> categories and bad targetting sections in units --
> with the 200+ units out
> there, it was easy to forget which unit I was on,
> which category types it
> should have, and which targets its weapons should
> generally ignore unless
> there are no other targets available.

Heh, you start out thinking you'll do it all then have no idea where you've got to.  Damn big projects.

> "Yes that's one way of putting it..."
> I'd prefer that the demos were just under 1,000kb
> compressed -- that's the
> largest file attachment HOTMAIL will allow me.

5min games then :P

> "Failure eh, 40min Krog's isn't good enough? :)"
>
> Counterstrike is a failure because it is not
> reliable. While it may get a
> krogoth in one game in as little as 30 minutes, it
> may take an hour in the
> next regardless of whether you're pounding on it. As
> you've well noticed,
> sometimes it builds only 1 fluxor at start,
> sometimes 2, and sometimes 3 or
> more! These units are EXTREME in the cheating
> department, I want the ai to
> have 2, no more than 2, no less than 2 on the hard
> setting -- and I just
> cannot get that kind of reliability out of it.

I always noted that the Counterstrike performance seemed to hinge entirely on how many resource things it built at the beginning, and that it would never seem to build them again later.  If you want 2, why not just weight them to 255, and set the limit to 2? The con units don't seem to suffer from overbuild unless they start while another one is doing something else, but you'd only have 1 con, the Commander.  If they always overbuild then you switch the limit to 1.

> Another way it is not reliable is nuke silos. I've
> seen the first nuke fired
> in just under 15 minutes. And other times, the ai
> starts building its first
> nuke silo after the 45 minute mark. 15 is too early,
> 45+ is too late.

Probably depends on the factor above.

> Counterstrike has already diverged away from BAI. It
> did this a while back
> too, when I decided that making an ai STRICTLY to
> beat BAI was pointless and
> wasn't fun to play against. So I watered down the
> original Counterstrike to
> be something barely beatable by above-average
> players who used some but not
> all anti-ai tactics. A really GOOD player might be
> able to beat
> Counterstrike without using any anti-ai tactics at
> all. It is worth noting
> that just like BAI, CS ai is EXTREMELY hard to beat
> on certain maptypes -
> namely nonmetal maps with no metal spots to build
> on. Those Fluxors go far in handing CS ai a win!

Heh, BAI and CS performing best on totally different map types.  God, BAI 5 sounds like it's gonna be invincible.

That's all my replies.
No really.
You can stop reading now.
Fine carry on reading but there's nothing left.
--------
This is what I sent you in an earlier email:

The BIG change in v1.3 was removing all the replicating "damaged terrain" scar features -- these features had a corpse -- themselves -- which made their creation and destruction an INFINITE loop. This bug is actually closely related to the missing DT (holes in DT walls) which sometimes occur in multiplayer games.

I said the bug is closely related to missing DT, I should've said what CAUSES the bug is closely related to what causes missing DT. In other words, what causes the missing DT to occur (packet loss) also causes this as well. However, all I said is I've removed the replicating "damaged terrain" scar features. Although unstated, the similar missing DT problem is not fixed, because I can't prevent packet loss.

"I run a clan and fight with the MSN Gaming Zone over technicalities, I never trust anyone online :)"
Interesting you only just now mention this... because in general I *HATE* clans.
Too competitive, too secretive, and/or just too dammed selfish/arrogant to help TA newbies.
A couple clans offered me membership, but I have refused to join them after I found out more about their "policies" towards non-clan members.
As a cynical request, I *HAVE* asked to have my ai admitted into a couple clans - if not as a member, then as a method to TRAIN clan members and as a test of would-be clan members. I was turned down on that every time, despite (in some cases) that same ai beating the clan members I was asking.
Please tell me why I wouldn't hate yours.

"I'm sure they were building them........sidedata again?"
Either that or an edited download *.tdf file somewhere.

"Or more often your mini base would destory them cutting off the flow of reinforcements :)  Nothing like a Brawler meeting a Goliath in the middle of no where."
I don't build Brawlers much anymore -- and certainly not to kill Goliaths, it takes too long. Brawlers do make good metal extractor raiders though!

"If only it would build the damn things, you can never rely on those safeguards to do anything!"
I've experimenting with adding geos to the minelayer's and decoy commander's build menus. So far it seems to guarentee a geo before the 15 minute mark, at least if a thermal vent is NEAR these c-units when they are built.

"I doubt our AI's will become the same (till you see sense and bow to Queller <g>), because your AI will always expand like mad immediately, where as mine will attack ASAP, and play to defeat raids too."
My ai should need higher limits than yours were it able to keep feeding its attack force as quickly as it expands. Yet the limits are very low and it doesn't seem to matter. BOTH our ai's have roughly the same number of attack units at roughly the same time, the ai EXE engine alone can explain any differences in numbers I see early on before 15 minutes. Our ai's already play VERY similarly:
1.they both spread out with c-units, just yours uses the commander a little longer
2.they both attack early for nonhacked ai's
3.they both go light on base defenses early on, compaired to BSRAI or McnTerra ai
4.they even have almost identical limits on the MOST KEY units needed by the ai
(these are metal extractors, solars, c-units, and factories.)

The major differance is how we use weights and orginizational schemes of the limits in the ai profile -- which doesn't much seem to matter to the ai EXE engine! Your weighting method looks different, but ends up with a similar result as mine. As I commented when I first saw Queller in action:
(the ai profiles) "It looks like BSRAI, but it plays like mine!" (meaning my MH ai)

"What I like to do is grab a main group, 1, and send it to fight the enemy army.  Then I get 5 Flash tanks and make them 2, and anything resource or con wise that the first group spots I send in number 2 to bag it. With the enemy army held up it works quite well (till I realise I've left my entire construction operation doing nothing :))."
I use my first attack force along with whatever base defenses I have to fight a pitched battle with the ai's first attack force. Then I send in c-units to clean up -- IF I win, that is... It's a Napoleonic-style fighting, and it's not very modern, but it's effective enough for me because it allows me to concentrate immense firepower on an attack force that lacks base defense support. Unless I'm either out-resourced and/or outfought, it works well on people too - because their attack force is generally no larger than my defensive mobile force - which has base defenses to back them up! I do some raiding, and the occassional rushing too, but consider it unsportsmanly to build lots of flashes and simply go for the throat. Doesn't stop me from doing it sometimes tho, as CyberDrgn found out in our brief recorded game. :)

"Don't bother, in OTA Goliaths are trundling death machines."
And it's probably death to an ai if it's trying to build 2 or more of these -18metal/sec monsters at once, if it doesn't already have a vast number of metal extractors running constantly somewhere. If you figure on each metal extractor making a nominal +1metal/sec, (which isn't uncommon as an average on a metal-poor map like Painted Desert) you only need 36 metal extractors devoted strictly for goliath production in 2 plants. 36 mex's take a total of -108 energy/sec to run and Goliaths also take about -110 energy/sec each to make. So that's about -320 energy/sec for the sustained self-contained production of 2 Goliaths at a time -- not counting brief factory downtimes when the Goliaths are moving out of the factory. It takes 16 solars to make that much energy, or a geo and 3.5 solars -- but the ai's seldom lucky or smart enough to make geos! Assuming your ai profiles has reached its unit max of 40 on solars and mex's and has no other resource income, it would only be producing +4 metal/sec and +480 energy/sec MORE than is needed (by solars and mex's alone) for just these 2 Goliath factories. In short, Goliaths better be dammed good, because that's about all your ai will have to beat its opponent/s with!

"I always think of Sumos as more defensive,"
If Sumos are BARELY worth building because they're little more than a defensive weapon -- what about the slower moving LLT? :p

(LLT'S)
"Well in general it's probably a good thing for non-cheating AI's to build them then :)"
LLT's slow down the ai's resource production considerably while offering little base coverage (and the ai's poor placement!) and limited effectiveness against enemy attackers. (Even a small number of level 1 units will waste them.) 
They'd be worth it if the ai DT'ed them and placed them well, but it doesn't. My ai is building expensive stuff like Sumos and Goliaths. As pointed out in my calculations above, my ai is literally starving for resources as it is. Is it feasible to have c-units spend time building LLT's when they could be building mex's, solars, factories, HLT's, defenders, and guardians instead?

(Vipers)
"D'OH, me bad."
Vipers are also more costly than HLT's, so I never weight or limit them very high.

(Guardians)
"I think I may have stumbled onto the weight in a UH test.  Now if only the damn AI Commander would stop trying to "help"."
The commander drains a staggering -43.6 metal/second and -172 energy/second when helping build a guardian. Even BAI on hard can nano-stall if it builds them with its first c-unit!

"I try to get Queller to get an Adv Kbot lab up ASAP, because Mavs are so dirt cheap (and take so damn long to build)."
Despite being powerful when used by players, I see Mavericks as improved Flash tanks to the ai. The sleep bug keeps them out of range and unable to fire most of the time and killed quickly even by slow-moving ballistic weapons that can't normally hit moving targets. On the other hand, ai Arm Pelicans on land or sea try to get in their laser's range so their missiles (which are IDENTICAL to MT's missiles) are almost always in range. Although individually the missiles are wimpy, they are harder to avoid than the hard-hitting but slower (on reload and speed) Rocko rockets.

"Metal is such a pain, usually my problem is that Queller hits it's limit, but knock up the limit and it screws everything.  Change the weights and each MEX built knocks less off the total weight, and so the factories take longer in coming.  Sigh."
(AND)
"I've noticed something that I'm sure didn't happen in MH 11.7, the MH factories always seem to close up after about 2 units.  Less nano-stall but a lot of the time the MH factories are in safelimit territory.  I think it might benefit from one less at the beginning, or prehaps a slightly larger time spacing between them."
Unfortunately, it's not so simple as lowering (relatively) weights on factories FURTHER to prevent the excessive number of factories it makes. MH 11.7 was more metal-hungry. MH 12.6 is TOO stingy with metal, so much so that build factory failsafes kick in and the commander often builds 2 or more factories in a row because the ai has more than 500 metal left in storage after it finishes the first. Because MH 12.6 ends up with more factories quicker, it also uses more energy because of this and ends up in an energy stall. The energy stall causes a FURTHER metal stall when mex's don't have energy to run. So, factories shut down. Later, when c-units first appear they are stuck constantly trying to build mex's (to satisfiy the metal failsafe) or solars (to satisfy the energy failsafe) and don't ever stop to smell the roses. Or build any base defenses till LONG after a player can raid the ai to death. In the rare event that the ai avoids most of these failsafe problems, it is harder to fight! I am working on the problem, but the solution seems incredibly counter-intuitive -- if one even exists!

"I never get that far, I win with my first 5 Flashes (happens more than you'd think), or it's game over :)"
If you mean against players, then surely you're playing against the wrong caliber for your abilities. I often make the game last a long time and victory painfully difficult for whichever side gets it. I've had players quit because even though they were winning (and KNEW it) they just couldn't seem to finish me off.

"They certinally cost a bomb in metal."
In short, Arm Ambushers suck compaired to Core Toasters.

"Luckily I'm a purist."
I use some 3rd party units (CS ai is one big 3rd party unit mod really :p) so I have to consider 3rd party when I'm making TA Bugfix -- but they're SECONDARY to purists.

ME:"ai coders apparently weren't listening to what the REST of the company was saying."
"Or anyone else for that matter."
The reason why failsafes exist in the first place was either there WAS no ai profiles to control the ai *OR* more likely ai profiles just weren't working well enough. The fact the ai profilers obviously didn't communicate well with the exe coders shows too -- it seems like the ai profiles (even mine) are in a constant tug-of-war with the ai code in the EXE.

"I remember that, I was always told not to use the Flash because of lag, but the Flash was ARM's main weapon!!!  I gave up in the end, and went back to good old lag free skirmishes."
I believe the major complaint now is how Samsons swarms murder Flash rushes, it's just not fair! BSR even went so far as to prevent that in UH by making Samsons and most other missile units anti-air only units. 

"Heh, you start out thinking you'll do it all then have no idea where you've got to.  Damn big projects."
Double-checking my work is harder than doing it.

"5min games then :P"
I figure a 1 vs 1 demo that zips up to 1,000 kb might be 1 hour long.

"I always noted that the Counterstrike performance seemed to hinge entirely on how many resource things it built at the beginning, and that it would never seem to build them again later.  If you want 2, why not just weight them to 255, and set the limit to 2? The con units don't seem to suffer from overbuild unless they start while another one is doing something else, but you'd only have 1 con, the Commander.  If they always overbuild then you switch the limit to 1."
They're not always overbuilt, in fact it's quite rare they're overbuilt. They are underbuilt however painfully often. Weights won't help much, because of the dreaded build factory failsafes -- once the ai has more than 500 metal in storage and a positive metal income, that's about all it wants to do - build factories.

(early nuke silos)
"Probably depends on the factor above."
This one depends actually on the numbers of nearby c-units that are willing to assist in the building of the nuke silo -- I've seen 1 nuke silo suck up ever c-air that 2 nearby aircraft plants were making, as well as getting c-vehicles helping as well. Other times, I've seen the nuke silo get built FAR away from any other units and takes a VERY long time to finish. Resources don't seem to be the issue, because the nuke silo's cost in per-second terms is so low.

"Heh, BAI and CS performing best on totally different map types.  God, BAI 5 sounds like it's gonna be invincible."
I have already helped Creeperman a lot on BAI 5. A LOT of what is in CS ai CAME from ideas he, Mephisto, and I were creating. You can bet that whatever is in CS ai will be in BAI 5 in SPADES. All I can say is I'm not interested in playing against it. BAI, CS, and Alpha ai can beat me well enough as it is!
--------
Subject: Re: Bugfix Q/A abridged version
Date: Sat, 20 Nov 1999 12:37:53 +0000 (GMT)

"OK, it *sounded* like the problem people were having."
I posted on Cavedog's message forum (which I almost never post at) in contradiction to your post stating that TA Bugfix does not fix the missing DT problem. I really didn't want to say "No you're wrong!" in an open forum like that, but didn't see that I had a choice.

"Well firstly, it's not a Total Annihilation clan :) We play Outwars and Tribes, and I've been running it for about a year (I was Second in Command June 1998, and took over in December 1998, though I was always running the clan really)."
So, it's not a TA clan. But what purpose is your clan?

"As I haven't changed C-Planes or Sea Lasers I doubt that."
But you said you saw c-airs building FHLT's?

"I still like watching AI ones shread whole columns if they find loads of Morties."
I've seen Morties kill Brawlers trying just such a trick.

"I wonder how the ratios compare though?  I think out level 1 Kbot ones are still fairly different :)"
As I say in Ai Tweaking Guide, getting the resources is what matters -- how the ai uses it is strictly minor details.

"It looks a lot less like BSRAI nowadays, it's....well......Queller :)."
I only compaired them superficially. IE: separate level 1 and 2 sections.

"Mind you the order can make a hell of a difference, I remember moving plants above defences and resources.  Eek!"
Limit orders don't seem to matter, but weights are like night and day if moved from top to bottom.

"You gotta love it :)"
"It's gonna getcha!"
If it wasn't for lag, I have VERY little problems stopping a flash-rush on a medium to huge map. Flash RAIDING on the other hand still kills me though.

"Two Goliaths can turn the game."
Not against a swarmer, a raider, or a porcer... :p The swarmer will smash Goliaths with sheer weight of numbers. A raider will keep a goliath-builder in permanent metal nano-stall. :p And a Porcer will make a defensive line capable of stopping Goliaths *IN DROVES* -- ai's like BAI and CS are a bitch to base-crack with Goliaths. Against a rusher however, Goliaths will quite likely keep the enemy attackers beat back -- but that's the ONLY tactic Goliaths are truely effective against. If an opposing attack force to Goliaths combines 2-pronged (or more!) attacks versis Goliaths in the field, they will quite likely lose far less in metal than the Goliath users will.

(Sumos)
"Slow moving HLT's.......with a ton of armour and impervious to everything."
So, they are slightly more worthwhile than LLT's.

"I've been working LLT's down simply to stop my Commander EVER building the damn things."
Having the commander building anything other than factories and resources before he goes into patrol mode seems an invitation for defeat on anything but the smallest of maps. However, on the small maps (such as Sherewood) a LLT is far more likely to be useful. In short, the small maps probably need a different ai profile than default.txt -- big maps and small maps should be played with some distinction. EMG Man is *RIGHT* to say MH is incredibly easy to raid on Sherewood as a result of this.

(Vipers)
"Yes but they cost less per tick"
And take longer to build, keeping c-units away from metal patches, thermal vents, and building solars for longer. :p

"A lot of AI's like to give Plasma Batteries a very high weight.......of course they nano-stall immediately."
Maybe they're intended to be played on the 10k/10k setting? At the least, they are considerably more effective there. MH on the other hand expands SLOWER resource-wise on 10k/10k than on 1k/1k. At 10k/10k, it gets stuck BIGTIME by the build factory failsafe.

"Yeah but a human doesn't fight like an AI.  Get a human with Flash tanks and unlike an AI he'll often charge them slap bang into the middle of your force. Often an AI doesn't have to move up, because the human comes to them."
That's how you may fight, but it is not how I fight. And I'm sure others prefer flanking manuevers that are even MORE effective against/with flash tanks than either tactic. If you send a horde of flash tanks into the middle of a swarm of mavericks the flashes will die very quickly. After one such disasterous encounter, either you'd better change said tactics or be willing to surrender. :p

"Anyway, Mavs are the best weapon an AI has, it needs them to stop CORE heavy armour rolling over it."
Merls are the best weapon an ARM AI has, IMHO. Level 1 units are its support, as are Mavericks. Close Support, nothing more. Wreckage defeats the Mavericks even more effectively than it does Flash tanks because of their slower speed.

"I wouldn't bother with pelicans, just use Samsons."
Samsons don't have as much armor as Pelicans, or that nasty laser in case some fool decides to charge the ai's attack force with Flash tanks. :p

"I'm currently only having late game metal problems. Still working on MMM's appearing at the right time."
MMM's are a BANE to the ai's resource model -- they are TOO sudden and TOO great of a change for the ai EXE engine to handle. That is UNLESS it gets fusions first!

"I have to win fast because I can't handle a large construction operation.  I'd much rather be handling multiple groups of troops, rather than a single con. Of course if Mapped mode is on, then I can play a long game perfectly well."
This is where the ai and I are similar, I'm perfectly at home with having 30+ c-units running around doing my bidding. However, on offensive manuevers I seldom have more than 1 ground attack force (but it's going to be BIG) to kill my enemy with. If I get into a stalemate, THEN I start resorting to small raiding groups.

"Well I'll have to let them know I'm employable to do profiles :).  Heh."
I believe they'll never hire anyone for that purpose -- as they have already written off TA's ai as "hopeless"...

"I enjoyed showing people that missile units could be eaten alive by Flashes."
Greenhaven favors Flashes, Painted Desert favors Samsons -- but it is wise to have both in either place. Paper, Rock, Sissors wars are for attack forces with little diversity. An attack force that is MOSTLY one thing, but with a little others for extra "flavor" are almost always at an advantage over 1-unittype attack forces.

"Bugfixing takes longer than anything else."
TA Bugfix sure has: over a year, and counting!

"Yes but that would only happen if the Commander froze for a while after the first Fluxor.  because after building it he wouldn't have over 500 Metal or energy, so he'd build another, but then he would and should go for a factory."
The commander has NO CHOICE but to build factories after build fluxors, that's the beautiful simplicity of CS. Other than getting stuck with 1 or 3 fluxors when I only want 2, the only other "risk" the ai runs is going air first. And even that's not SO bad, because the commander will build 1 or 2 other factores in short concession as well. Sometimes the ai goes completely boneheaded and builds 2 or even 3 air plants in a row, then the commander goes into patrol mode. It's only then that the ai is truely screwed versis raiding/rushing forces.

"I like to get them started ASAP so by the time the resources exist to fire it, it's been made."
I like ai nuke silos because it keeps the c-units from doing something that will cripple its tenuous resource economy.

"I don't enjoy playing cheating AI's, it just doesn't do anything for me."
I intended CS ai to play more like a human player than other cheating ai's and not just cheat via hacked resource units because the ai cannot win playing fair. If you remove AI_Cheat.ufo and use ALL the other ai-only components with MH ai, you should get an ai hacked in much the same manner as Queller ai -- harder because of improvements in unit abilities more than blatant resource-cheats.
--------
Other than the fluxors, what is there *IN* CS ai that allows the ai to cheat? There's the Arm ai vehicle plant, to be sure, but what about Core? The only thing really cheating about the ai-only krogoth gantry is it supplies its own resources. Any player bothering to build krogoths will generally guard the gantry up to the limits of their economy -- which may be FAR beyond the assistance of 2 Adv. C-vehicles. 11 minutes for a Krogoth is not what I'd call "fast" for a player's Gantry.

CS's cloaking fusions add considerably more resources than the player's do, but these aren't exactly eary-game winners like fluxors are meant to be. And the Arm Shooter's vastly reduced cloak cost is to cover for the fact that SMART players would only have their Shooters cloaked for the brief moments as they get in range, the poor ai has to spend -200 e/sec constantly if its shooters are on the move -- whether it's of any "value" or not! 5 such shooters moving about while cloaked drains as much energy as BOTH of its 2 fluxors produce. This is not good, considering that the ai has far fewer supplimentary energy sources than other ai's.

This is why I generally don't make posts about what Queller can and cannot do -- I might be wrong because I didn't make it.
--------
Subject: Re: Bugfix Q/A abridged version
Date: Sun, 21 Nov 1999 11:16:48 +0000 (GMT)

> "OK, it *sounded* like the problem people were
> having."
> I posted on Cavedog's message forum (which I almost
> never post at) in
> contradiction to your post stating that TA Bugfix
> does not fix the missing
> DT problem. I really didn't want to say "No you're
> wrong!" in an open forum
> like that, but didn't see that I had a choice.

Don't worry, I'd have done the same for you :)

> So, it's not a TA clan. But what purpose is your
> clan?

Well it started off just as a social thing.  That worked for Outwars because it was a small game, a sinlge lobby on the MSN gaming Zone and everyone knew everyone else.  Now we're trying to get into a Tribes league.

> "As I haven't changed C-Planes or Sea Lasers I doubt
> that."
> But you said you saw c-airs building FHLT's?

Yes.

> "I still like watching AI ones shread whole columns
> if they find loads of
> Morties."
> I've seen Morties kill Brawlers trying just such a
> trick.

They obviously weren't using enough then :P

> As I say in Ai Tweaking Guide, getting the resources
> is what matters -- how
> the ai uses it is strictly minor details.

You mean if it builds metal so that it can gain metal to build more metal, that's good?  Sounds like 3.1 to me....... :)

> "You gotta love it :)"
> "It's gonna getcha!"
> If it wasn't for lag, I have VERY little problems
> stopping a flash-rush on a
> medium to huge map. Flash RAIDING on the other hand
> still kills me though.

I prefer PeeWees for raiding because they're easier to direct.  I never use them though because MT's whip them, they don't have enough armour.

> "Two Goliaths can turn the game."
> Not against a swarmer, a raider, or a porcer... :p
> opposing attack force to
> Goliaths combines 2-pronged (or more!) attacks
> versis Goliaths in the field,
> they will quite likely lose far less in metal than
> the Goliath users will.

That's why you have an escort.

> (Sumos)
> "Slow moving HLT's.......with a ton of armour and
> impervious to everything."
> So, they are slightly more worthwhile than LLT's.

LLT's usually see more action :)

> In short, the small maps probably need
> a different ai profile than default.txt -- big maps
> and small maps should be
> played with some distinction. EMG Man is *RIGHT* to
> say MH is incredibly
> easy to raid on Sherewood as a result of this.

I have thought of doing one, but for the moment I'd rather get the exisiting (and used) profiles finished before starting new totally new ones.

> (Vipers)
> "Yes but they cost less per tick"
> And take longer to build, keeping c-units away from
> metal patches, thermal
> vents, and building solars for longer. :p

And from starting on that Punisher it's itching to build :)

> Maybe they're intended to be played on the 10k/10k
> setting? At the least,
> they are considerably more effective there. MH on
> the other hand expands
> SLOWER resource-wise on 10k/10k than on 1k/1k. At
> 10k/10k, it gets stuck
> BIGTIME by the build factory failsafe.

Yes I said Queller is hard at 10K in the readme, then I watched it.  For the first 5 min it's hard, then it nano-stalls hard and stays there.  Eek!   Might take that paragraph out :)

> That's how you may fight, but it is not how I fight.
> And I'm sure others
> prefer flanking manuevers that are even MORE
> effective against/with flash
> tanks than either tactic. If you send a horde of
> flash tanks into the middle
> of a swarm of mavericks the flashes will die very
> quickly. After one such
> disasterous encounter, either you'd better change
> said tactics or be willing
> to surrender. :p

I don't usually do that myself :)  I prefer to use a small group of Flashes as spotters for Samsons and to delay groups of enemies.  What I love to do is get a big group and peel them off one by one (it looks stylish :)).  Then when I've got the base surrounded (this is early on) select their group and they come in from different directions.

> "Anyway, Mavs are the best weapon an AI has, it
> needs them to stop CORE
> heavy armour rolling over it."
> Merls are the best weapon an ARM AI has, IMHO. Level
> 1 units are its
> support, as are Mavericks. Close Support, nothing
> more. Wreckage defeats the
> Mavericks even more effectively than it does Flash
> tanks because of their
> slower speed.

No, PeeWee's are! :P

> "I wouldn't bother with pelicans, just use Samsons."
> Samsons don't have as much armor as Pelicans, or
> that nasty laser in case
> some fool decides to charge the ai's attack force
> with Flash tanks. :p

On the other hand Samsons only use a Lv1 factory, where as pelicans are going to tie up a Lv2.

Oi!  Nothing like a good Flash charge to put your fate in the hands of the Gods :)

> MMM's are a BANE to the ai's resource model -- they
> are TOO sudden and TOO
> great of a change for the ai EXE engine to handle.
> That is UNLESS it gets
> fusions first!

The problem is the AI is getting about three or four fusions before building an MMM...........

> "I have to win fast because I can't handle a large
> construction operation.
> I'd much rather be handling multiple groups of
> troops, rather than a single
> con. Of course if Mapped mode is on, then I can play
> a long game perfectly
> well."
> This is where the ai and I are similar, I'm
> perfectly at home with having
> 30+ c-units running around doing my bidding.
> However, on offensive manuevers
> I seldom have more than 1 ground attack force (but
> it's going to be BIG) to
> kill my enemy with. If I get into a stalemate, THEN
> I start resorting to small raiding groups.

As I said I like to operate with 2 groups.  Though I do like to try and win with 5 Flashes ASAP :)

I still can't believe I used 5 Flashes to hold off 2 human foes on Lava Highground, while I uses a Con unit to HLT the ramps up to the centre!  Hah!

> I believe they'll never hire anyone for that purpose
> -- as they have already
> written off TA's ai as "hopeless"...

As has everyone else.

> "I enjoyed showing people that missile units could
> be eaten alive by
> Flashes."
> Greenhaven favors Flashes, Painted Desert favors
> Samsons -- but it is wise
> to have both in either place. Paper, Rock, Sissors
> wars are for attack
> forces with little diversity. An attack force that
> is MOSTLY one thing, but
> with a little others for extra "flavor" are almost
> always at an advantage
> over 1-unittype attack forces.

Mmmmmmmmm, crunchy PeeWee's.

> I like ai nuke silos because it keeps the c-units
> from doing something that
> will cripple its tenuous resource economy.

Or anything that could save it as well.
--------
Subject: Re: Bugfix Q/A abridged version
Date: Sun, 21 Nov 1999 11:16:48 +0000 (GMT)

"Don't worry, I'd have done the same for you :)"
I don't know if I should thank you for this comment or want to smack you.

"They obviously weren't using enough then :P"
It takes a pretty huge number of Brawlers to kill just 10 Morties and if there's any AA around, it takes even more.

"You mean if it builds metal so that it can gain metal to build more metal, that's good?  Sounds like 3.1 to me."
Cavedog's ai sucks SHIT for getting metal. It makes a LOT of c-units that spend most of their time trying to build base defenses they can't afford. The default weights seem to be the major cause of this.

"I prefer PeeWees for raiding because they're easier to direct.  I never use them though because MT's whip them, they don't have enough armour."
Building 5 Peewees then going Rockos on small maps can certainly surprise people if they're used to being Jeffy-rushed. What shocked me once was when I saw 10+ Jeffys attacking at once on radar. (I thought they were flash tanks, and would probably have been dead if they were!) My ally in that 2 vs 2 game on Great Divide 2 lost all his metal extractors and most of his solars in a matter of seconds. I on the other hand had built 3 Samsons first from my first vehicle plant, and ended up with one getting vetran status at a price of losing only 1 metal extractor and getting the Samsons slightly damaged.

"That's why you have an escort."
If your ai can afford massed Goliaths with escorts before the 30 minute mark, I honestly believe yours is far better than mine!

"LLT's usually see more action :)" (than sumos)
I agree, if your ai builds LLT's with its commander it will be too far enough behind in the resource AND war race to prevent its opponent/s from attacking its base with ground attackers. It may well be that the LLT's is the ONLY base defense that ai will have, since its ground force growth could be stunted. So in short, if the ai builds LLT's it will probably need them. I on the other hand don't like my opponents to get that close! Sumos are a different kind of "insurance", I've seen them used in many ai vs ai fights where Core's attack force gets stripped of its frontline units (Instigators, Storms, and Thuds) due to vast amounts of concentrated firepower which leaves a lone Sumo behind while everything else retreats. Ai Flashes quickly swarm it, Samsons pelt it, Merls hammer it -- and yet it ussually leaves a huge number of kills behind. Sometimes it even survives. :p So sumos are the sacrificial unit to prevent a retreat from turning into a total defeat -- they CANNOT be overrated as such because their armor is *SO* high and their weapon is so deadly to anything DUMB enough to get close!

Unfortunately, the same cannot be said for the lowly ai LLT. By the time the LLT "saves the day", the ai's attack force is almost certainly destroyed, it probably has already lost significant numbers of c-units and metal extractors out in the field, its base probably already taken serious damage, and quite possibly its commander is dead as well. This is not a situation that the ai can comeback from with any degree of reliability -- even if LLTs destroy the enemy's entire attack force. If the ai knew to place its LLTs in choke-points on the map and DT them, I'd have the ai build them in a heartbeat.

"I have thought of doing one, but for the moment I'd rather get the exisiting (and used) profiles finished before starting new totally new ones."
Queller's default.txt is slightly closer to being a small-map land ai than MH is. For me, I'd have to start almost from scratch!

"And from starting on that Punisher it's itching to build :)"
At some point, I *WANT* the ai to scratch that itch -- once. :)
For Vipers on the other hand, I'm not so sure I want the ai to build them...

"Yes I said Queller is hard at 10K in the readme, then I watched it.  For the first 5 min it's hard, then it nano-stalls hard and stays there.  Eek!   Might take that paragraph out :)"
Mine ussually energy-stalls with 10k/10k, because it does occassionally build metal extractors it doesn't need immediately -- but it forgets solars/wind to go with it!

"I don't usually do that myself :)  I prefer to use a small group of Flashes as spotters for Samsons and to delay groups of enemies."
Is there anything that can offer better LOS than Flashes if all you want them for is spotters and mobile DT?

"What I love to do is get a big group and peel them off one by one (it looks stylish :)).  Then when I've got the base surrounded (this is early on) select their group and they come in from different directions."
Actually surrounding someone's base suggests a HUGE overproduction of units compaired to your opponent. What that translates to me is: "After I outproduce them, I kill them." It *IS* a good strategy, but if it fails what do you do? How do you fight battles where you're seeing 2x as many enemy units than you can produce in return?

In the TA demo of me vs CyberDrgn, I pushed him to the desperate point (and he pushed me to the same!) of resorting to using an Atlas to grab my raiding Flash tanks -- then using his commander to reclaim them for METAL! I saw his Atlas and thought he was intending to try to com-nap me -- which considering my painful lack of AA near my commander, was a viable tactic. What other uses might that Atlas had?

In chat, sometime after the battle, I suggested if he was going to "nab" my tanks he might have been better off capturing them than reclaiming them -- he said at the time all he could think about was getting metal.

"No, PeeWee's are! :P"
PeeWees are scouts and raiders, they'll never be decent close support.

"On the other hand Samsons only use a Lv1 factory, where as pelicans are going to tie up a Lv2."
OMG! You're correct! Pelicans cost 2 times as much too! So let's not build them since they are of no value at all! My bad... Mavericks cost slightly more even. :p I'm just saying a Maverick-only strategy from the Arm Adv. Kbot lab is a little limited. Better to have both - Mavericks are good if players try to overrun the ai's attack force, Pelicans are good if they don't.

Among online players, it's ussually only newbies that complain about Mavericks -- but most non-newbies all the way up to elite players say that the Pelican *IS* overpowered, at least while on the water. Many of them say the Pelican is overpowered even without that bug, due to their good armor, good weapons, cheap cost, and good mobility over land.

"Oi!  Nothing like a good Flash charge to put your fate in the hands of the Gods :)"
Or at least offer a distraction from the MAIN attack.

"The problem is the AI is getting about three or four fusions before building an MMM."
Mine ussually has 2-3 MMM's before the FIRST fusion! This is partly why I work so hard for it to get Geos. :p

"As I said I like to operate with 2 groups.  Though I do like to try and win with 5 Flashes ASAP :)"
I prefer long games on LANs and medium-length games online. Team games are my favorite, sometimes half the challenge is helping someone who isn't as good as yourself survive. Good team communication skills is rare even among expert TA players.

"I still can't believe I used 5 Flashes to hold off 2 human foes on Lava Highground, while I uses a Con unit to HLT the ramps up to the centre!  Hah!"
Think that's bad, I killed both of my opponent's commanders in a 2 vs 2 on Lava Highground with 5 Flash tanks. They were both in the center, and one was too busy building at first to notice. The other was either out of energy or afraid to d-gun because I tried very hard to keep my flashes directly between the 2 commanders. I still lost the game though, it was commander continues and my ally was a newbie.

"As has everyone else."
I haven't, I consider it mostly harmless. :)

I like ai nuke silos because it keeps the c-units... busy
"Or anything that could save it as well."
You misunderstand! (Or are you attempting to insult my methods?) That's not by far ALL I want them to do, but I'd prefer for 4+ adv. c-units to NOT all be building 4+ separate Fusions or even Big Berthas at the same time. Better to have 2 of them doing "busywork" making "useless" ai nuke silos than putting them into NEGATIVE metal nano-stall from multiple high per-second cost build projects.

Slowhandd thinks an unbeatable ai (or at least a hard-to-beat ai) can be made by not setting limits on c-units, metal extractors, and solars -- as he claims Queller lacks. Surely you will at least grudgingly agree with me that the ai having near-unlimited numbers of c-units wandering around looking for something to do (and some will certainly end up building guardians!) is counter-productive for the ai's surivival. Just because a few is good, doesn't mean a whole lot is better.
--------
Subject: Re: Bugfix Q/A abridged version part 2 :p
Date: Mon, 22 Nov 1999 17:19:50 +0000 (GMT)

> "Don't worry, I'd have done the same for you :)"
> I don't know if I should thank you for this comment
> or want to smack you.

And I'm not sure which I'd prefer :)

> "They obviously weren't using enough then :P"
> It takes a pretty huge number of Brawlers to kill
> just 10 Morties and if
> there's any AA around, it takes even more.

Yeah, but there's no better way to round off a game than with a nice Brawler rush and watching the minimap cloud over with EMG.  Of course you're rounding off the game because you just wasted all your metal on them instead of Hawks :)

Though incidently (have I mentioned this?) I let Queller build Brawlers more than Hawks simply because I don't believe it uses the Hawks anywhere near as well.  They constantly change target, and I rarely see them even take out a MEX because of it.

> "You mean if it builds metal so that it can gain
> metal to build more metal,
> that's good?  Sounds like 3.1 to me."
> Cavedog's ai sucks SHIT for getting metal. It makes
> a LOT of c-units that
> spend most of their time trying to build base
> defenses they can't afford.
> The default weights seem to be the major cause of
> this.

I still love the Comm wiggle (beware those 2 Fleas :)).  Actually, I hate sleep time more.

> "I prefer PeeWees for raiding because they're easier
> to direct.  I never use
> them though because MT's whip them, they don't have
> enough armour."
> Building 5 Peewees then going Rockos on small maps
> can certainly surprise
> people if they're used to being Jeffy-rushed. What
> shocked me once was when
> I saw 10+ Jeffys attacking at once on radar. (I
> thought they were flash
> tanks, and would probably have been dead if they
> were!) My ally in that 2 vs
> 2 game on Great Divide 2 lost all his metal
> extractors and most of his
> solars in a matter of seconds. I on the other hand
> had built 3 Samsons first
> from my first vehicle plant, and ended up with one
> getting vetran status at
> a price of losing only 1 metal extractor and getting
> the Samsons slightly damaged.

I suck at Jeffy raiding, but then after PeeWee raiding I suck with Rocko's, ain't life a pain :)

> "That's why you have an escort."
> If your ai can afford massed Goliaths with escorts
> before the 30 minute mark, I honestly believe yours
> is far better than mine!

I think v9 could do this, but only if it wasn't going to be harrased, simply because it didn't do much in the way of level 1 (which meant it generally wouldn't survive long enough).  Massed of course, well 7 (roughly) more like, but that's not bad :)

> "LLT's usually see more action :)" (than sumos)
> I agree, if your ai builds LLT's with its commander
> it will be too far
> enough behind in the resource AND war race to
> prevent its opponent/s from
> attacking its base with ground attackers. It may
> well be that the LLT's is
> the ONLY base defense that ai will have, since its
> ground force growth could
> be stunted. So in short, if the ai builds LLT's it
> will probably need them.
> I on the other hand don't like my opponents to get
> that close! Sumos are a
> different kind of "insurance", I've seen them used
> in many ai vs ai fights
> where Core's attack force gets stripped of its
> frontline units (Instigators,
> Storms, and Thuds) due to vast amounts of
> concentrated firepower which
> leaves a lone Sumo behind while everything else
> retreats. Ai Flashes quickly
> swarm it, Samsons pelt it, Merls hammer it -- and
> yet it ussually leaves a
> huge number of kills behind. Sometimes it even
> survives. :p So sumos are the
> sacrificial unit to prevent a retreat from turning
> into a total defeat --
> they CANNOT be overrated as such because their armor
> is *SO* high and their weapon is so deadly to
> anything DUMB enough to get close!

It's not as if I don't use Sumos, but I consider the front action where Cans are, more important than coverinmg your ass.  Anyway, you don't need as many Sumos due to armour and range considerations.

> Unfortunately, the same cannot be said for the lowly
> ai LLT. By the time the
> LLT "saves the day", the ai's attack force is almost
> certainly destroyed, it
> probably has already lost significant numbers of
> c-units and metal
> extractors out in the field, its base probably
> already taken serious damage,
> and quite possibly its commander is dead as well.
> This is not a situation
> that the ai can comeback from with any degree of
> reliability -- even if LLTs
> destroy the enemy's entire attack force. If the ai
> knew to place its LLTs in
> choke-points on the map and DT them, I'd have the ai
> build them in a
> heartbeat.

I only have it sprinkle them for a nice bit of variety (which my testers demanded stay in).

> "I have thought of doing one, but for the moment I'd
> rather get the
> exisiting (and used) profiles finished before
> starting new totally new ones."
> Queller's default.txt is slightly closer to being a
> small-map land ai than MH is. For me, I'd have to
> start almost from scratch!

Small!!!  Pah! :)  Queller can handle big maps, if you let it start to win it gets construction out in the background very quickly while it's forces advance.

> "And from starting on that Punisher it's itching to build :)"
> At some point, I *WANT* the ai to scratch that itch
> -- once. :)
> For Vipers on the other hand, I'm not so sure I want
> the ai to build them...

I find they help against the smug players who Merl all day (they hold them up, while not taking a lot longer to build then an HLT).  I consider them a worthwhile investment, though I favour HLT's.

> "Yes I said Queller is hard at 10K in the readme,
> then I watched it.  For
> the first 5 min it's hard, then it nano-stalls hard
> and stays there.  Eek!
> Might take that paragraph out :)"
> Mine ussually energy-stalls with 10k/10k, because it
> does occassionally
> build metal extractors it doesn't need immediately
> -- but it forgets solars/wind to go with it!

While you're on wind.  I have noticed that the AI doesn't seem to treat wind the same as Solar.  In a game on metal it would run out of energy while using wind when with solar it would be just fine.  I figure that prehaps soem of the inbuilt code that forces solars isn't there for wind.  Really though, I'm not sure why this happens (I don't think it's all in my
mind, it's quite a noticable difference).

> "I don't usually do that myself :)  I prefer to use
> a small group of Flashes
> as spotters for Samsons and to delay groups of
> enemies."
> Is there anything that can offer better LOS than
> Flashes if all you want
> them for is spotters and mobile DT?

Peepers, which mean you have to invest in a very expensive, very slow factory, when you could get another VP for more Samsons, and Flashes fire back too.  Flashes are also fairly flexible units if you decide you want them for something else.

> "What I love to do is get a big group and peel them
> off one by one (it looks
> stylish :)).  Then when I've got the base surrounded
> (this is early on)
> select their group and they come in from different
> directions."
> Actually surrounding someone's base suggests a HUGE
> overproduction of units
> compaired to your opponent. What that translates to
> me is: "After I
> outproduce them, I kill them." It *IS* a good
> strategy, but if it fails what
> do you do? How do you fight battles where you're
> seeing 2x as many enemy
> units than you can produce in return?

Sorry, surrounded doesn't mean constant line, I merely meant to show the coming from different directions.  I usually only do this with 5 or 6 Flash tanks.

> In the TA demo of me vs CyberDrgn, I pushed him to
> the desperate point (and
> he pushed me to the same!) of resorting to using an
> Atlas to grab my raiding
> Flash tanks -- then using his commander to reclaim
> them for METAL! I saw his
> Atlas and thought he was intending to try to com-nap
> me -- which considering
> my painful lack of AA near my commander, was a
> viable tactic. What other
> uses might that Atlas had?

Get Adv Kbot and find out :)  Personally I think he'd have been better off building a different plant, that for a start would have saved him metal.  Surprised that he didn't build a Con plane rather than that Atlas, that would have been a vastly superior investment for metal.

> In chat, sometime after the battle, I suggested if
> he was going to "nab" my
> tanks he might have been better off capturing them
> than reclaiming them --
> he said at the time all he could think about was
> getting metal.

Why capturing first?  What does that gain (excuse my ignorance I never bother since people just Self-D the things).

> "No, PeeWee's are! :P"
> PeeWees are scouts and raiders, they'll never be
> decent close support.

Yeah but they're so cute :)

> "On the other hand Samsons only use a Lv1 factory,
> where as pelicans are
> going to tie up a Lv2."
> OMG! You're correct! Pelicans cost 2 times as much
> too! So let's not build
> them since they are of no value at all! My bad...
> Mavericks cost slightly
> more even. :p I'm just saying a Maverick-only
> strategy from the Arm Adv.
> Kbot lab is a little limited. Better to have both -
> Mavericks are good if
> players try to overrun the ai's attack force,
> Pelicans are good if they
> don't.

Pelicans blow, give me Mavs anyday.  Nothing like a rational line of argument.  Basically I'd rather have Mavs than Pels, and you like Pels.

I'm right :)

> Among online players, it's ussually only newbies
> that complain about
> Mavericks -- but most non-newbies all the way up to
> elite players say that
> the Pelican *IS* overpowered, at least while on the
> water. Many of them say
> the Pelican is overpowered even without that bug,
> due to their good armor, good weapons, cheap cost,
> and good mobility over land.

What is there to complain about?  Guass lag? :).  Well I don't need to worry about the Pel being overpowered since I'm an average player, we average players see much more exciting and varied battles because we don't go for statistically right forces :)

I still can't believe I errected 2 HLT's by each Lava highground ramp, the bozos :)

> "Oi!  Nothing like a good Flash charge to put your
> fate in the hands of the
> Gods :)"
> Or at least offer a distraction from the MAIN
> attack.

Damn, that was the main attack :)

> "The problem is the AI is getting about three or
> four fusions before
> building an MMM."
> Mine ussually has 2-3 MMM's before the FIRST fusion!
> This is partly why I
> work so hard for it to get Geos. :p

Good, I'll play it on Plains and Passes :P

> "As I said I like to operate with 2 groups.  Though
> I do like to try and win
> with 5 Flashes ASAP :)"
> I prefer long games on LANs and medium-length games
> online. Team games are
> my favorite, sometimes half the challenge is helping
> someone who isn't as
> good as yourself survive. Good team communication
> skills is rare even among
> expert TA players.

I'm still trying to convince my LAN friends what a great game this is.  Philistines don't like the music, I'll tell em where to put Red Alert. :)

My team communication is OK, except when I play against AI's (I have this whole new lazy skill reserved especially for them).  I really should stop Queller crushing EMG on the large maps :)

> "I still can't believe I used 5 Flashes to hold off
> 2 human foes on Lava
> Highground, while I uses a Con unit to HLT the ramps
> up to the centre! Hah!"
> Think that's bad, I killed both of my opponent's
> commanders in a 2 vs 2 on
> Lava Highground with 5 Flash tanks. They were both
> in the center, and one
> was too busy building at first to notice. The other
> was either out of energy
> or afraid to d-gun because I tried very hard to keep
> my flashes directly
> between the 2 commanders. I still lost the game
> though, it was commander
> continues and my ally was a newbie.

I lost too, one enemy slipped 4 Flashes through, I destroyed 2 of them, but then he was in my base.  I had called for my ally to help defend the ramps while I got the defence up, but he ignored me.  Later he dropped.  Sigh!

> "As has everyone else."
> I haven't, I consider it mostly harmless. :)

Well I ain't got the rest of that sentance, but usually I disagree with you, so consider yourself disagreed with :).

> I like ai nuke silos because it keeps the c-units...
> busy
> "Or anything that could save it as well."
> You misunderstand! (Or are you attempting to insult
> my methods?) That's not
> by far ALL I want them to do, but I'd prefer for 4+
> adv. c-units to NOT all
> be building 4+ separate Fusions or even Big Berthas
> at the same time. Better
> to have 2 of them doing "busywork" making "useless"
> ai nuke silos than
> putting them into NEGATIVE metal nano-stall from
> multiple high per-second
> cost build projects.

Mmmmmmmm, multiple high per-second cost build projects.  Say that 10 times fast :).  Well I just keep the high level construction vehicles coming out slowly so that they generally aren't starting 6 Fusions at once, though not too slowly that my AI falls behind in the Level 3 race (which was a problem in v10 fixed for v11).

> Slowhandd thinks an unbeatable ai (or at least a
> hard-to-beat ai) can be
> made by not setting limits on c-units, metal
> extractors, and solars -- as he
> claims Queller lacks. Surely you will at least
> grudgingly agree with me that
> the ai having near-unlimited numbers of c-units
> wandering around looking for
> something to do (and some will certainly end up
> building guardians!) is
> counter-productive for the ai's surivival. Just
> because a few is good,
> doesn't mean a whole lot is better.

Queller lacks no such thing, Queller has limits on all of those (he obviously hasn't bothered to look).  I would agree with you that he is 100% wrong.  Do you really want an Ai that everytime it loses a attacking unit replaces it with a solar panel?  The only threat you'd be under was claustrophobia as those deadly panels sneaked every closer to you.  At least your BB wouldn't be able to miss!
--------
Subject: Re: Bugfix Q/A abridged version part 2 :p
Date: Mon, 22 Nov 1999 17:19:50 +0000 (GMT)

"And I'm not sure which I'd prefer :)"
I receive somewhere between 2 and 20+ emails a week about 3rd party work I have done on TA or MW3. People almost never email just to say they like it. So that makes many/most of the emails something like:
"I like TA Bugfix mostly, but I think this part sucks..."
And that's the friendly posts!

I get the occassional post that calls me nothing more than a filthy cheater/hacker who's intent to make trainers to ruin the various games.

It gets to you after a while, as you may have already seen on Queller. Please try not to disagree with me in an attempt at humor, as I try not to do the same to you.

My nerves are often shot as it is. And my health is nowhere where it should be.

"Yeah, but there's no better way to round off a game than with a nice Brawler rush and watching the minimap cloud over with EMG.  Of course you're rounding off the game because you just wasted all your metal on them instead of Hawks :)"
I've seen 5-10 ai Brawlers wipe out every outlying metal extractor the enemy has, for that they are excellent.

"Though incidently (have I mentioned this?) I let Queller build Brawlers more than Hawks simply because I don't believe it uses the Hawks anywhere near as well.  They constantly change target, and I rarely see them even take out a MEX because of it."
It's the 1+K LOS for the ai Hawks -- they're ALWAYS spotting something they'd RATHER be shooting at, especially if it isn't in their badtargetting section. That and they are ADD (Attention Deficit Disorder) to begin with. :p

"I still love the Comm wiggle (beware those 2 Fleas :)).  Actually, I hate sleep time more."
From discussions I've had with Creeperman, Onslaught, CyberKewl, and Mephisto, it seems that sleep time is NOT something that is programmed into the game to make the ai easier to beat. Instead, it is an unfortunate side effect of the ai's cyclic control structure. Only about once a second (at most) can the ai actually give new move orders for its ground troops. The cycle time is increased and decreased based on the computer speed and how much is already happening in the game. If the CPU is struggling with pathfinding problems, the cycle time can be huge. If move orders occurs right after the ai's ground troops just killed an enemy, they immediately begin the new order almost seamlessly. But there could be a long wait if they killed an enemy unit right after the new order "tick." It's all theory tho, but my programming "instincts" seem to support this theory.

"I suck at Jeffy raiding, but then after PeeWee raiding I suck with Rocko's, ain't life a pain :)"
Rockos are easy to raid with, although they're a tad slow. 2 Rocko rockets = 1 dead mex. If they get attacked by flashes, retreat! If they get attacked by Samsons/Slashers in equal or lesser numbers, charge! After big fights, move the most damaged Rockos to the back or middle of the pack -- this way they'll live longer and still offer support. I've even seen a huge Rocko swarms (>40 in number) take down Krogoths. This is in LAN games with my friends -- the ai's were supplying the Krogoths.

"I think v9 could do this, but only if it wasn't going to be harrased, simply because it didn't do much in the way of level 1 (which meant it generally wouldn't survive long enough).  Massed of course, well 7 (roughly) more like, but that's not bad :)"
7 is a HUGE number of Goliaths for the ai to have. MH's Arm ai might have a SLIGHT chance of fighting that many Goliaths in the field at once if it's 40 minutes or later in the game. In your opinion, about how many Mavericks (plus a small amount of level 1 escort units) is enough to stop 7 Goliaths if the Goliaths also have escorts?

"It's not as if I don't use Sumos, but I consider the front action where Cans are, more important than coverinmg your ass.  Anyway, you don't need as many Sumos due to armour and range considerations."
In MH ai Cans have higher limits AND weights than sumos. However, Morties get a FAR bigger nod... Dominators are my ai in quantity because they offer instantaneous damage like Merls do -- by the time your outlying defenses/attack force takes its first hit from them, something's dead.

"I only have it sprinkle them for a nice bit of variety (which my testers demanded stay in)."
It sounds like MOST of your testers are raiders/rushers. Unfortunately, almost none of them seem to be ai designers as well. LLT's aren't useless for the ai, I just see them as a too little, too late problem. If you can "trick" the ai to not build them until later in the game (where they'd more likely be guarding a mex or 2) it'd be good. I ussually don't expect laser towers in the middle of Painted Desert at the 15 min mark. The problem with building them early is resource aquisition is exponential: it takes resources to quickly gain more resources -- the ai who gets the most metal extractors on metal spots the fastest can have a shitty defense of them because it only needs to keep them about 1 minute to turn a profit from them. And that tiny profit, compounded over 20+ metal extractors, may mean an additional 1,000 metal (or more) a minute. And when I see THAT amount of additional metal, I tend to think how many more Flash tanks I could get with that.

"Small!!!  Pah! :)  Queller can handle big maps, if you let it start to win it gets construction out in the background very quickly while it's forces advance."
Queller is GOOD on big to even huge maps, but unlike MH it is not SHITTY on small maps. It is only so-so there. Which is pretty good for a nonhacked ai!

"I find they help against the smug players who Merl all day (they hold them up, while not taking a lot longer to build then an HLT).  I consider them a worthwhile investment, though I favour HLT's."
The problem there is the ai doesn't know to keep its Viper/s shut when under Merl fire. And while open, a Viper has even LESS armor than the HLT. (988 vs 1,200) I say keep them for variety, but favor HLT's by a considerable margin.

"While you're on wind.  I have noticed that the AI doesn't seem to treat wind the same as Solar.  In a game on metal it would run out of energy while using wind when with solar it would be just fine.  I figure that perhaps some of the inbuilt code that forces solars isn't there for wind.  Really though, I'm not sure why this happens (I don't think it's all in my
mind, it's quite a noticable difference)."
Solars actually CONSUME -20 energy while on, which results in a +20 energy production (2 negatives cancel out). Wind however produces positive energy. Someone thoeried that the ai makes solars to REDUCE its energy consumption, since solars actually cancel out energy consumption via their negative consumption. Also, there doesn't seem to be as much inbuilt code for wind to me either. One thing's for sure, the ai sure doesn't feel the same way about fusions, geothermals, and light carriers as it does about solars! I can barely STOP it from building solars, but I have to FORCE it to do the others!

"Peepers, which mean you have to invest in a very expensive, very slow factory, when you could get another VP for more Samsons, and Flashes fire back too.  Flashes are also fairly flexible units if you decide you want them for something else."
Actually I was hinting at Stumpies, many people treat them as nearly worthless because flash tanks seem "so" much better. Stumpies have 50% more armor (and roughly 50% more cost) and hav 250 LOS as compaired to 225 LOS for Flashes. MH ai probably should build some, but I like Flashes too much. :p

"Get Adv Kbot and find out :)"
EVERYONE (except complete newbies) knows about flying bombs, but CyberDrgn didn't even have a basic Kbot lab. But what about the value of airlifting a c-vehicle (which has 3x the buildrate of a c-air) to a far corner? Maybe making multiple runs to airlift multiple C-vehicles and Samsons as well.

"Personally I think he'd have been better off building a different plant, that for a start would have saved him metal.  Surprised that he didn't build a Con plane rather than that Atlas, that would have been a vastly superior investment for metal."
He was possibly intending to go air early, thinking I wasn't going to be doing much raiding. And he possibly built the Atlas to airlift already-built C-vehicles between distant metal patches. When the raids first appeared, his already-built Atlas had to be put to use as a flash nabber, to make up for the lack of other options.

"Why capturing first?  What does that gain (excuse my ignorance I never bother since people just Self-D the things)."
You cannot give orders to a unit that is being transported. So, if he captures the flash while still under the transport you can do nothing but watch. Plus, I was UNAWARE in the battle that he *WAS* nabbing my flash tanks. I was too busy looking elsewhere. The flash tanks were meant as a distraction for him more than me. :p

"Yeah but they're so cute :)"
Against non-missile units, PeeWees do ok dodging fire. I remember playing the 1st Core mission in TA and wishing I had PeeWees instead of AK's. What's really sad is watching a whole army of PeeWees try to attack a HLT or even a LLT.

"Pelicans blow, give me Mavs anyday."
Blow?
They have the same missile weapon as the missile tower. When the ai is using them, they fire their missiles from max range thereby outranging almost all other mobile units. They can deal damage slightly faster than a flash and nearly 4 times faster than a Samson or Slasher!

Compaired to Mavericks:
Mavericks can deal damage 6.6 times faster than Pelicans.
And they BETTER TOO, because Pelicans build 4.1 times faster.
Mavericks have 870 armor, Pelicans only 800.
Pelicans move faster, Mavericks have a higher turnrate.
Mavericks cost 1.9 times as much metal as Pelicans and 4.4 times as much energy.

In the time the Core ai can make 7 Goliaths from one Adv Vehicle Plant an Arm ai's Adv Kbot Lab can build 2 Mavericks. Might this be partially why Queller's Arm ai gets stomped by Queller's Core ai?

"What is there to complain about?  Gauss lag? :)."
The Maverick's Gauss weapon does more damage per second than any mobile unit weapon other than the Krogoth's laser. Having a couple Mav's in an attack force equals evaporation of any level 1 units they run across and quick death for advanced units as well. Complaints about the Mav are because of its weapon -- it doesn't take many to kill a commander no matter HOW quick the player can d-gun! Mav's are also great for killing Krogoths for the same reason. Unfortunately, the ai manages to fuckup this awesome unit with its dammed sleep bug! This is not to say ai Mavericks aren't worth building from the Adv. Kbot Lab, just that they probably shouldn't be the vast majority.

"Nothing like a rational line of argument.  Basically I'd rather have Mavs than Pels, and you like Pels."
In player versis player games on land maps, I personally build Mavericks and seldom build Pelicans. That's not what I'm trying to argue tho. :)

"Well I don't need to worry about the Pel being overpowered since I'm an average player"
I didn't think you wanted your ai to be only average?

"Damn, that was the main attack :)"
A Flash attack almost has a "freshness date" on it. If it's not used early, your opponent can have up ample means (DT'ed HLT's anyone?) to stop it.

"Good, I'll play it on Plains and Passes :P"
MH ai is pretty good at reclaiming trees, so it should do good there -- if it can find a place to build its factories without blocking them.

"I'm still trying to convince my LAN friends what a great game this is.  Philistines don't like the music, I'll tell em where to put Red Alert. :)"
Red Alert is good, if you don't know about TA. Most people don't have the mental stamina to play TA open-ended for an hour. After 1 hour, I'm ussually either repairing what I have or working on eliminating my opponent. Were I to continue building at the same rate as I was ~20 minutes earlier, my base would grow far faster than any opponent could destroy it. Red Alert on the other hand has much stricter limits on just how MUCH players can create through the limits in resources on the map. Aircraft are pretty weak in Red Alert, and base defenses are pitiful with exception of Tesla Coils, which is excellent against infantry...

"My team communication is OK, except when I play against AI's (I have this whole new lazy skill reserved especially for them).  I really should stop Queller crushing EMG on the large maps :)"
Team communication does not neccessarily mean playing better. KNOWING that EMG is getting crushed is team communication, saving his butt is teamwork. (or from the sound of it, YOUR work.) I also play conservative against ai's, other than Counterstrike and BAI -- I play dirty versis them.

"I had called for my ally to help defend the ramps while I got the defence up, but he ignored me.  Later he dropped."
He may have ignored you because he never got the message packet. Ussually, players are disconnected from the game before you see the "so-and-so is not responding" message.

"Well I just keep the high level construction vehicles coming out slowly so that they generally aren't starting 6 Fusions at once, though not too slowly that my AI falls behind in the Level 3 race (which was a problem in v10 fixed for v11)."
Very early adv. c-units are almost as hard on resources as early guardians.

"Queller lacks no such thing, Queller has limits on all of those (he obviously hasn't bothered to look).  I would agree with you that he is 100% wrong."
It shoots full of holes most of his accusations about you, because he doesn't know what he's talking about when he mentions Queller.

"Do you really want an Ai that everytime it loses an attacking unit replaces it with a solar panel?"
BAI v4.0 on hard has this exact problem! One big complaint about BAI v4.0 was its attack force becomes smaller and smaller late in the game. When it was first revealed, BSR was clueless what the problem was. :p

It's pretty sad when the ai defeats itself.
--------
Subject: Re: Bugfix part 3
Date: Wed, 24 Nov 1999 12:38:04 +0000 (GMT)

> I receive somewhere between 2 and 20+ emails a week
> about 3rd party work I
> have done on TA or MW3. People almost never email
> just to say they like it.
> So that makes many/most of the emails something
> like:
> "I like TA Bugfix mostly, but I think this part
> sucks..."
> And that's the friendly posts!

I (as I'm sure many people do) find it easier to point out flaws.  I don't know why, it just doesn't natural to say, Ooooo I like this bit.  I just say I like the whole thing.  I think it's because when people have got something that's perfect they don't feel the will to write because they've already got everything they want, and writing would force them to take time off. However when people see something wrong, they let you know so you can fix it and move the product towards pefection.

I get about 10 e-mails a day :).  I used to get 20-25.

> I get the occassional post that calls me nothing
> more than a filthy
> cheater/hacker who's intent to make trainers to ruin
> the various games.

Probably the same people who became the TAK zealots :)

> It gets to you after a while, as you may have
> already seen on Queller.
> Please try not to disagree with me in an attempt at
> humor, as I try not to
> do the same to you.

I just remember that I'm so right it's unbelievable :).

> My nerves are often shot as it is. And my health is
> nowhere where it should be.

Nerves?  I have nerves of cotton :).

> "Yeah, but there's no better way to round off a game
> than with a nice
> Brawler rush and watching the minimap cloud over
> with EMG.  Of course you're
> rounding off the game because you just wasted all
> your metal on them instead
> of Hawks :)"
> I've seen 5-10 ai Brawlers wipe out every outlying
> metal extractor the enemy
> has, for that they are excellent.

I've gone back to Hawks.  By having a thick patrol route the AI builds up it's planes.  When I saw 12 hawks fly in an decimate an entire mini base in just 2 sweeps I knew I'd made the right choice :)

> "Though incidently (have I mentioned this?) I let
> Queller build Brawlers
> more than Hawks simply because I don't believe it
> uses the Hawks anywhere
> near as well.  They constantly change target, and I
> rarely see them even
> take out a MEX because of it."
> It's the 1+K LOS for the ai Hawks -- they're ALWAYS
> spotting something
> they'd RATHER be shooting at, especially if it isn't
> in their badtargetting
> section. That and they are ADD (Attention Deficit
> Disorder) to begin with. :p

After all that I've swapped back :)

> "I still love the Comm wiggle (beware those 2 Fleas
> :)).  Actually, I hate
> sleep time more."
> From discussions I've had with Creeperman,
> Onslaught, CyberKewl, and
> Mephisto, it seems that sleep time is NOT something
> that is programmed into
> the game to make the ai easier to beat. Instead, it
> is an unfortunate side
> effect of the ai's cyclic control structure. Only
> about once a second (at
> most) can the ai actually give new move orders for
> its ground troops. The
> cycle time is increased and decreased based on the
> computer speed and how
> much is already happening in the game. If the CPU is
> struggling with
> pathfinding problems, the cycle time can be huge. If
> move orders occurs
> right after the ai's ground troops just killed an
> enemy, they immediately
> begin the new order almost seamlessly. But there
> could be a long wait if
> they killed an enemy unit right after the new order
> "tick." It's all theory
> tho, but my programming "instincts" seem to support
> this theory.

As far as I could tell TAK doesn't seem to suffer from this.  Guess they fixed at least 1 thing.

> "I suck at Jeffy raiding, but then after PeeWee
> raiding I suck with Rocko's,
> ain't life a pain :)"
> Rockos are easy to raid with, although they're a tad
> slow. 2 Rocko rockets =
> 1 dead mex. If they get attacked by flashes,
> retreat! If they get attacked
> by Samsons/Slashers in equal or lesser numbers,
> charge! After big fights,
> move the most damaged Rockos to the back or middle
> of the pack -- this way
> they'll live longer and still offer support. I've
> even seen a huge Rocko
> swarms (>40 in number) take down Krogoths. This is
> in LAN games with my
> friends -- the ai's were supplying the Krogoths.

Rocko's are brilliant base decimators.  Especially for the numerous level 1 structures.  Send in a demo squad of Rocko's and BOOM, bye bye base.

> "I think v9 could do this, but only if it wasn't
> going to be harrased,
> simply because it didn't do much in the way of level
> 1 (which meant it
> generally wouldn't survive long enough).  Massed of
> course, well 7 (roughly)
> more like, but that's not bad :)"
> 7 is a HUGE number of Goliaths for the ai to have.
> MH's Arm ai might have a
> SLIGHT chance of fighting that many Goliaths in the
> field at once if it's 40
> minutes or later in the game. In your opinion, about
> how many Mavericks
> (plus a small amount of level 1 escort units) is
> enough to stop 7 Goliaths
> if the Goliaths also have escorts?

Ah, well Queller can most definately pull off 7 Golaiths by the 30th minute with escort (depending on casulties suffered).  I watched it yesterday, and it must have cycled through several Golaiths by that time.  It all depends on how well it does, and how early it goes level 2.

> "It's not as if I don't use Sumos, but I consider
> the front action where
> Cans are, more important than coverinmg your ass.
> Anyway, you don't need as
> many Sumos due to armour and range considerations."
> In MH ai Cans have higher limits AND weights than
> sumos. However, Morties
> get a FAR bigger nod... Dominators are my ai in
> quantity because they offer
> instantaneous damage like Merls do -- by the time
> your outlying defenses/attack force takes its
> first hit from them, something's dead.

I go for Morties mainly because their range means they aren't so effected by the sleep bug, and also they're fast, finally they aren't going to be beat by a player who makes sure the nearest target is 10 DT'd HLT's.

> "I only have it sprinkle them for a nice bit of
> variety (which my testers
> demanded stay in)."
> It sounds like MOST of your testers are
> raiders/rushers. Unfortunately,
> almost none of them seem to be ai designers as well.
> LLT's aren't useless
> for the ai, I just see them as a too little, too
> late problem. If you can
> "trick" the ai to not build them until later in the
> game (where they'd more
> likely be guarding a mex or 2) it'd be good. I
> ussually don't expect laser
> towers in the middle of Painted Desert at the 15 min
> mark. The problem with
> building them early is resource aquisition is
> exponential: it takes
> resources to quickly gain more resources -- the ai
> who gets the most metal
> extractors on metal spots the fastest can have a
> shitty defense of them
> because it only needs to keep them about 1 minute to
> turn a profit from
> them. And that tiny profit, compounded over 20+
> metal extractors, may mean
> an additional 1,000 metal (or more) a minute. And
> when I see THAT amount of
> additional metal, I tend to think how many more
> Flash tanks I could get with that.

I put the LLT's down quite a way though (before this discussion), just because it CAN build them, doesn't mean it will.

> "Small!!!  Pah! :)  Queller can handle big maps, if
> you let it start to win
> it gets construction out in the background very
> quickly while it's forces
> advance."
> Queller is GOOD on big to even huge maps, but unlike
> MH it is not SHITTY on
> small maps. It is only so-so there. Which is pretty
> good for a nonhacked ai!

Well if it's gonna be a big pat on the back session I'd say MH is probably better on larger maps because it goes construction sooner :)

I have still to find an AI that can play Ashap Plateau, that metal spacing is anti-AI from the word go.

> "I find they help against the smug players who Merl
> all day (they hold them
> up, while not taking a lot longer to build then an
> HLT).  I consider them a
> worthwhile investment, though I favour HLT's."
> The problem there is the ai doesn't know to keep its
> Viper/s shut when under
> Merl fire. And while open, a Viper has even LESS
> armor than the HLT. (988 vs
> 1,200) I say keep them for variety, but favor HLT's
> by a considerable margin.

Same here.  I keep them for the times when the player hits solar first.  The good old AI always seems to put it's lasers BEHIND solars, so that by the time the player reaches them, they don't have anything to power them :)

> "While you're on wind.  I have noticed that the AI
> doesn't seem to treat
> wind the same as Solar.  In a game on metal it would
> run out of energy while
> using wind when with solar it would be just fine.  I
> figure that perhaps
> some of the inbuilt code that forces solars isn't
> there for wind.  Really
> though, I'm not sure why this happens (I don't think
> it's all in my
> mind, it's quite a noticable difference)."
> Solars actually CONSUME -20 energy while on, which
> results in a +20 energy
> production (2 negatives cancel out). Wind however
> produces positive energy.
> Someone thoeried that the ai makes solars to REDUCE
> its energy consumption,
> since solars actually cancel out energy consumption
> via their negative
> consumption. Also, there doesn't seem to be as much
> inbuilt code for wind to
> me either. One thing's for sure, the ai sure doesn't
> feel the same way about
> fusions, geothermals, and light carriers as it does
> about solars! I can
> barely STOP it from building solars, but I have to
> FORCE it to do the others!

I just dispair as the solar code is great, but it builds 2 winds then decides prehaps the next best thing is 40 MEX's.

> "Peepers, which mean you have to invest in a very
> expensive, very slow
> factory, when you could get another VP for more
> Samsons, and Flashes fire
> back too.  Flashes are also fairly flexible units if
> you decide you want
> them for something else."
> Actually I was hinting at Stumpies, many people
> treat them as nearly
> worthless because flash tanks seem "so" much better.
> Stumpies have 50% more
> armor (and roughly 50% more cost) and hav 250 LOS as
> compaired to 225 LOS
> for Flashes. MH ai probably should build some, but I
> like Flashes too much.
> :p

I get Queller to build them so it doesn't take a beating from early DT's (though it's not as if Stumpies are kings of anti-DT), also on maps like Yerrot Mountains they're invaluable for nailing MEX's on top of pillars.  Finally in wreckage areas they can come in useful.

> "Get Adv Kbot and find out :)"
> EVERYONE (except complete newbies) knows about
> flying bombs, but CyberDrgn
> didn't even have a basic Kbot lab. But what about
> the value of airlifting a
> c-vehicle (which has 3x the buildrate of a c-air) to
> a far corner? Maybe
> making multiple runs to airlift multiple C-vehicles
> and Samsons as well.

I wasn't impling that you didn't know, it just sounded better than saying flying BOOM :).

I still think capturing is generally a waste of time.

> "Personally I think he'd have been better off
> building a different plant,
> that for a start would have saved him metal.
> Surprised that he didn't build
> a Con plane rather than that Atlas, that would have
> been a vastly superior
> investment for metal."
> He was possibly intending to go air early, thinking
> I wasn't going to be
> doing much raiding. And he possibly built the Atlas
> to airlift already-built
> C-vehicles between distant metal patches. When the
> raids first appeared, his
> already-built Atlas had to be put to use as a flash
> nabber, to make up for the lack of other options.

Air first on Greenhaven, Ooooooooo I wanna crush him :)

> "Why capturing first?  What does that gain (excuse
> my ignorance I never
> bother since people just Self-D the things)."
> You cannot give orders to a unit that is being
> transported. So, if he
> captures the flash while still under the transport
> you can do nothing but
> watch. Plus, I was UNAWARE in the battle that he
> *WAS* nabbing my flash
> tanks. I was too busy looking elsewhere. The flash
> tanks were meant as a
> distraction for him more than me. :p

Heh, can't keep up eh :P  For me he could have nabbed my Commander and I'd never have noticed, but a Flash I'd spot in an instance.  Think my priorities are off? :)

> "Yeah but they're so cute :)"
> Against non-missile units, PeeWees do ok dodging
> fire. I remember playing
> the 1st Core mission in TA and wishing I had PeeWees
> instead of AK's. What's
> really sad is watching a whole army of PeeWees try
> to attack a HLT or even a LLT.

HAH!  Good old demo mission 2, send in the PeeWees! Even as early as then I suspected that PeeWees were better than A.K.'s.

> "Pelicans blow, give me Mavs anyday."
> Blow?
> They have the same missile weapon as the missile
> tower. When the ai is using
> them, they fire their missiles from max range
> thereby outranging almost all
> other mobile units. They can deal damage slightly
> faster than a flash and
> nearly 4 times faster than a Samson or Slasher!

There I go using Brit terms again, suck.  I would prfer a Mav because by that point you're gonna have a lot of structures.  You could be approaching the unit limit and you need units that operate well in small numbers.  Pelicans need to be in fairly large packs, and this IMHO is a bad idea ta level 2 because of the unit limit.

> Compaired to Mavericks:
> Mavericks can deal damage 6.6 times faster than
> Pelicans.
> And they BETTER TOO, because Pelicans build 4.1
> times faster.
> Mavericks have 870 armor, Pelicans only 800.
> Pelicans move faster, Mavericks have a higher
> turnrate.
> Mavericks cost 1.9 times as much metal as Pelicans
> and 4.4 times as much energy.
>
> In the time the Core ai can make 7 Goliaths from one
> Adv Vehicle Plant an Arm ai's Adv Kbot Lab can build
> 2 Mavericks. Might this be partially why
> Queller's Arm ai gets stomped by Queller's Core ai?

Hell no :)  The ARM can operate just as well at level 2 because the Mavs cost next to nothing to build, so while the CORE may slow down with these expensive units, the ARM does no such thing (and can afford to get Mavs out LONG before the CORE gets Gols).

> "What is there to complain about?  Gauss lag? :)."
> The Maverick's Gauss weapon does more damage per
> second than any mobile unit
> weapon other than the Krogoth's laser. Having a
> couple Mav's in an attack
> force equals evaporation of any level 1 units they
> run across and quick
> death for advanced units as well. Complaints about
> the Mav are because of
> its weapon -- it doesn't take many to kill a
> commander no matter HOW quick
> the player can d-gun! Mav's are also great for
> killing Krogoths for the same
> reason. Unfortunately, the ai manages to fuckup this
> awesome unit with its
> dammed sleep bug! This is not to say ai Mavericks
> aren't worth building from
> the Adv. Kbot Lab, just that they probably shouldn't
> be the vast majority.

Have players missed the very loooooooonnnnnnnnnnng build time of a Mav?  I still disagree and will unit time comes to an end :)

> "Nothing like a rational line of argument.
> Basically I'd rather have Mavs
> than Pels, and you like Pels."
> In player versis player games on land maps, I
> personally build Mavericks and seldom build Pelicans.
> That's not what I'm trying to argue tho. :)

When I said like, that was meant to be like for an AI, oh well :)

> "Well I don't need to worry about the Pel being
> overpowered since I'm an average player"
> I didn't think you wanted your ai to be only average?

I want my AI to be as good as possible, and to do this it should beat average players.  Average players out number pro's by about 4 or 5 to 1 I'd guess, not only that, but if you're a pro you're also going to play BAI or CounterStrike not Queller.

> "Damn, that was the main attack :)"
> A Flash attack almost has a "freshness date" on it.
> If it's not used early,
> your opponent can have up ample means (DT'ed HLT's
> anyone?) to stop it.

I go around them.  If you can be bothered, a patroling Peeper by an HLT draining 75 energy from their economy until they notice and shut it down (then you Stumpy rush it :)) is always fun.  Course he can send in AA, but that's less for my force to deal with.

> "Good, I'll play it on Plains and Passes :P"
> MH ai is pretty good at reclaiming trees, so it
> should do good there -- if
> it can find a place to build its factories without
> blocking them.

Lack of Geo though.  Actually Queller seems to have fun taunting me here (MH has done it too), there is so much metal so close together the AI can go on a resource frenzy with it's Commander.

> "I'm still trying to convince my LAN friends what a
> great game this is.
> Philistines don't like the music, I'll tell em where
> to put Red Alert. :)"
> Red Alert is good, if you don't know about TA. Most
> people don't have the
> mental stamina to play TA open-ended for an hour.
> After 1 hour, I'm ussually
> either repairing what I have or working on
> eliminating my opponent. Were I
> to continue building at the same rate as I was ~20
> minutes earlier, my base
> would grow far faster than any opponent could
> destroy it. Red Alert on the
> other hand has much stricter limits on just how MUCH
> players can create
> through the limits in resources on the map. Aircraft
> are pretty weak in Red
> Alert, and base defenses are pitiful with exception
> of Tesla Coils, which is
> excellent against infantry...

I remember when somone tried to rush me with everything they had, 4 Telsa coils held them off :)
Heh.  The AI beat me later, I was never very good at RA :)

> "My team communication is OK, except when I play
> against AI's (I have this
> whole new lazy skill reserved especially for them).
> I really should stop
> Queller crushing EMG on the large maps :)"
> Team communication does not neccessarily mean
> playing better. KNOWING that
> EMG is getting crushed is team communication, saving
> his butt is teamwork.
> (or from the sound of it, YOUR work.) I also play
> conservative against ai's,
> other than Counterstrike and BAI -- I play dirty
> versis them.

Well we let each other know what's going on, but I don't remember us doing much about it :)  Mind you we exchange troops some times, or send in task forces. The best team work we did was on Greehaven, I was holding my Queller with some heavy defence and strafing Rocko's, EMG was being beaten back, and I sent out a large task force of Flashes to raid his foe.  There must have been 8 or 9 fights going on after that.  I had to quit though, and he eventually lost.  I blame him :)

> "I had called for my ally to help defend the ramps
> while I got the defence
> up, but he ignored me.  Later he dropped."
> He may have ignored you because he never got the
> message packet. Ussually,
> players are disconnected from the game before you
> see the "so-and-so is not responding" message.

He was still expanding at the time, but yes he did have an unreliable connection.  Damn shame, I was humiliating the foe :)

> "Well I just keep the high level construction
> vehicles coming out slowly so
> that they generally aren't starting 6 Fusions at
> once, though not too slowly
> that my AI falls behind in the Level 3 race (which
> was a problem in v10
> fixed for v11)."
> Very early adv. c-units are almost as hard on
> resources as early guardians.

Not THAT early.

> "Queller lacks no such thing, Queller has limits on
> all of those (he
> obviously hasn't bothered to look).  I would agree
> with you that he is 100%
> wrong."
> It shoots full of holes most of his accusations
> about you, because he
> doesn't know what he's talking about when he
> mentions Queller.
>
> "Do you really want an Ai that everytime it loses an
> attacking unit replaces
> it with a solar panel?"
> BAI v4.0 on hard has this exact problem! One big
> complaint about BAI v4.0
> was its attack force becomes smaller and smaller
> late in the game. When it
> was first revealed, BSR was clueless what the
> problem was. :p

I tried to raid it once, damn HLT's.  It builds so many factories and requires so few resources that the whole base is one big lump, and HLT's have very little trouble covering it.  It's also Nuke bait :)

> It's pretty sad when the ai defeats itself.

Good old MH :P j/k
--------
Subject: Re: Bugfix part 3
Date: Wed, 24 Nov 1999 12:38:04 +0000 (GMT)

BTW, my "fair" ai has been reffered to by me as "Mostly Harmless" ai as far back as v3.5 -- it's now v12+. It's ai reviewers and other people who've tried to stick other names on it, not I.

"I get about 10 e-mails a day :).  I used to get 20-25."
A lot of my emails are either rant-mail or asking me to explain how things work. Oftentimes, it's what I've wrote vast amounts about in my readme's for my files -- which is very irritating to have to answer again. I often just cut and paste from those readme's and email that. :p

"Probably the same people who became the TAK zealots :)"
No, mostly it's not even related to TA -- it's on the MechWarrior 3 BUGFIX I'm working on. Back when I got regular similar emails about TA, it was how my ai was a cheat/trainer/whatever because it could be used on one computer in a multiplayer game and be "different" ai-wise than other computer's ai's. The FEW emails that even seem sort of related to cheating/hacking now either are asking me for cheats/hacks or are saying that TA Bugfix is/isn't compatible with the latest TA trainers out and would I "fix" that?

"I just remember that I'm so right it's unbelievable :)."
You can learn more from mistakes than successes.

"Nerves?  I have nerves of cotton :)."
I have had TA webmasters, TA IRCops, Cavedog employees, major 3rd party developers, and even average Joe's in the TA community telling me to go fuck myself at one time or another. Sometimes, they don't put it so mildly... And on at least one occassion they joined forces to really get the message accross. Oftentimes, this is even over stuff I am not responsible for and didn't do! (I've gotten multiple hate-mails for BAI, by people who thought I made it. If BSR only knew...) Othertimes, it's over complete misunderstandings where people took what I said EXACTLY BACKWARDS of both what my intent was and my wording. (Too many double-negatives, I guess.)

My health doesn't allow me the luxury of holding down a full-time job, or even a part-time job if it's strenuous.

"I've gone back to Hawks.  By having a thick patrol route the AI builds up it's planes.  When I saw 12 hawks fly in an decimate an entire mini base in just 2 sweeps I knew I'd made the right choice :)"
But include some Brawlers for diversity -- I do, just on the slim hopes they prove useful on occassion. :)

"After all that I've swapped back :)"
I think I've reversed-stance on MH's use of many units 2, 3, or even more times as it's progressed. Metal Makers have been a love-hate relationship since MH's very beginning.

"As far as I could tell TAK doesn't seem to suffer from this.  Guess they fixed at least 1 thing."
That is also probably why TA:K runs so slowly.

"Rocko's are brilliant base decimators.  Especially for the numerous level 1 structures.  Send in a demo squad of Rocko's and BOOM, bye bye base."
My strategy on The Cold Place is to go for a Rocko rush. I've killed players off in under 15 minutes doing that.

"Ah, well Queller can most definately pull off 7 Golaiths by the 30th minute with escort (depending on casulties suffered).  I watched it yesterday, and it must have cycled through several Golaiths by that time.  It all depends on how well it does, and how early it goes level 2."
Those are some VERY big claims for Queller -- 7 Goliaths is almost 5k metal, and drained at a considerable rate. Are you sure it could do that on Painted Desert?

"I go for Morties mainly because their range means they aren't so effected by the sleep bug, and also they're fast, finally they aren't going to be beat by a player who makes sure the nearest target is 10 DT'd HLT's."
MH's specialty (as I've found out the hard way) is its ability to remove fixed defenses with ease if not PERM LOS. Stuff like Morties, Dominators, Diplomats, and even Storms, Slashers, and Goliaths are the contributing reasons. :)

"Well if it's gonna be a big pat on the back session I'd say MH is probably better on larger maps because it goes construction sooner :)"
If not too soon. Versions to follow shall act slightly differently on c-units...

"I have still to find an AI that can play Ashap Plateau, that metal spacing is anti-AI from the word go."
Ashap Plateau has contour errors in it, and makes it very difficult to build factories anywhere other than the plateaus themselves because of small "hills" in the lower land. All told, Counterstrike ai still might play "decently" there, but even BAI might not.

"Same here.  I keep them for the times when the player hits solar first.  The good old AI always seems to put it's lasers BEHIND solars, so that by the time the player reaches them, they don't have anything to power them :)"
I put MY HLT's behind my solars often too! But I ussually have excess power by the time I do this, and at least a small mobile force to back my HLT's up with.

"I just dispair as the solar code is great, but it builds 2 winds then decides prehaps the next best thing is 40 MEX's."
Nah, it mixes in a few Guardians, factories, and HLT's as well -- for variety. :p

"I get Queller to build them so it doesn't take a beating from early DT's (though it's not as if Stumpies are kings of anti-DT), also on maps like Yerrot Mountains they're invaluable for nailing MEX's on top of pillars.  Finally in wreckage areas they can come in useful."
I'm at odds on adding Stumpies back to MH ai and nixing Hammers...

"I wasn't impling that you didn't know, it just sounded better than saying flying BOOM :)."
CyberDrgn was HURTING bad for metal, and the best idea you can come up for his Atlas is for him to build a Kbot Lab, an Advanced Kbot Lab, then a crawling bomb? :p His 1 Atlas could be used in many interesting and annoying ways... you ever had your c-units air-napped?

"I still think capturing is generally a waste of time."
Ussually it is, but capturing fully-built units is cheaper than reclaiming them and building your own. Especially when you cannot be stopped from doing so!

"Air first on Greenhaven, Ooooooooo I wanna crush him :)"
Come to annihilated's irc chatroom sometime, and you MIGHT get a chance...

"Heh, can't keep up eh :P  For me he could have nabbed my Commander and I'd never have noticed, but a Flash I'd spot in an instance.  Think my priorities are off? :)"
More like my flash tanks were spread out accross the map -- plus I had *HIS* flash tanks to worry about attacking me from the bottom of the map.

"HAH!  Good old demo mission 2, send in the PeeWees! Even as early as then I suspected that PeeWees were better than A.K.'s."
AK's have the same laser as the Weasal, PeeWees have the same weapon as the Flash tank. If AK's had the same weapon as the Instigator, they'd be almost twice as dangerous.

"There I go using Brit terms again, suck."
Blow here means a very extreme form of suck. It'd suck to find you're out of ice cream. It'd blow to find out you're out of food and snowed in for a few days. See the differance? :p

"I would prefer a Mav because by that point you're gonna have a lot of structures.  You could be approaching the unit limit and you need units that operate well in small numbers.  Pelicans need to be in fairly large packs, and this IMHO is a bad idea ta level 2 because of the unit limit."
My ai won't hit 250 unit limit quickly -- and I don't think it ever reaches 500 unit max unless I have some 3rd party units installed that it doesn't have limits on.

"Hell no :)  The ARM can operate just as well at level 2 because the Mavs cost next to nothing to build, so while the CORE may slow down with these expensive units, the ARM does no such thing (and can afford to get Mavs out LONG before the CORE gets Gols)."
Core can struggle to get out Goliaths and get them out at a reduced rate of 1/3 normal speed and STILL produce Goliaths slightly quicker than Arm can produce Mavericks. It's my opinion that Mavericks can kill about an equal number of Goliath tanks, so that's a fair trade in metal but not a fair trade in energy. Also, because of the sleep bug the ai's Mavericks spend very little time in firing range of enemy units (they kill them too quickly :p) -- I've seen pathetic ai Dominators absolutely murder Mavericks because a single Instigator got close to the Mavericks and killed and then the Mavericks took a 5+ second nap.

"Have players missed the very loooooooonnnnnnnnnnng build time of a Mav?  I still disagree and will unit time comes to an end :)"
Long buildtimes is a problem only for those who aren't inventive... See the attached Pic for details.

"I want my AI to be as good as possible, and to do this it should beat average players.  Average players out number pro's by about 4 or 5 to 1 I'd guess, not only that, but if you're a pro you're also going to play BAI or CounterStrike not Queller."
Only rush junkies play BAI and CounterStrike and like them... Even I prefer playing against MH than CounterStrike. CS's dammed near unbeatable on larger maps! And I don't much like playing on small maps versis the ai...

"I go around them.  If you can be bothered, a patroling Peeper by an HLT draining 75 energy from their economy until they notice and shut it down (then you Stumpy rush it :)) is always fun.  Course he can send in AA, but that's less for my force to deal with."
If someone builds an HLT without putting up at least a MT or 2 nearby, they are lousy players! That's like building a bertha without AA around it! A single level 1 bomber can MAKE TIME to kill HLT's that are in inconvenient locations. :)

"Lack of Geo though.  Actually Queller seems to have fun taunting me here (MH has done it too), there is so much metal so close together the AI can go on a resource frenzy with it's Commander."
Plains and Passes has more geo vents on it than Painted Desert does -- they're just all near the center.

"I remember when somone tried to rush me with everything they had, 4 Telsa coils held them off :)
Heh.  The AI beat me later, I was never very good at RA :)"
I got pretty good at RA, but one of my friends got far better at it. On TA, he was better still! I made my first ai for TA as a TRAINER for me to get me the skills I needed to beat him. It didn't work too good tho, but I learned by playing online games so I went from average player to roughly an expert in about 6 months.
His skills at TA have deteriorated because he doesn't play it much (if any) anymore, while mine is about the same so I can probably beat him now... just barely.

"Well we let each other know what's going on, but I don't remember us doing much about it :)  Mind you we exchange troops some times, or send in task forces. The best team work we did was on Greehaven, I was holding my Queller with some heavy defence and strafing Rocko's, EMG was being beaten back, and I sent out a large task force of Flashes to raid his foe.  There must have been 8 or 9 fights going on after that.  I had to quit though, and he eventually lost.  I blame him :)"
Teamwork doesn't guarentee a win, it just ups the odds.
In our LAN games here, I used to give the ai's loaded nuke silos, if we're winning by a large margin. :p
Worse still, I'd give the ai my entire base - and go into watch mode and laugh as the other players got stomped...

"Not THAT early."
Very early can even mean before 15 minutes. Ussually it seems that about the time MH starts building guardians en mass it's also started building advanced factories, berthas, and fusion. I've seen it stay below 500 metal for something like 40 minutes gametime straight. But I can promise you that were I playing against it I wouldn't have been able to notice!

"I tried to raid it once, damn HLT's.  It builds so many factories and requires so few resources that the whole base is one big lump, and HLT's have very little trouble covering it.  It's also Nuke bait :)"
BAI and Counterstrike both make very compact bases. Counterstike, because it uses NORMAL factories, has to build more and thus spreads out more, so it is slightly less vulnerable to nukes.

Most players when playing on large open land maps like Painted Desert seldom ever bother with antinukes, or seldom build more than one because their bases in LONG games can often cover many screens in area.

I remember playing a 2 Core vs 2 Arm players (me) game on The Cold Place. One of my opponents was Vampyre, who as a really good Core player. I was pissed because my ally was lagging really bad and was killed off less than 15 minutes in the game by a combined attack by both opponents. After that point I was fighting the battles as revenge. I got the attitude that they ganged up on my partner, (who seemed to be a newbie) and they were GOING TO kill me too -- and I was going to make them pay for the right to do so!

About 50 minutes into the game one of my opponents threatened to nuke me. I said, "Sure, go ahead -- I'm at unit max, this will just free up more units for me to throw at you." At the time, I think I had all of 5 factories. But they were screens apart, surrounded by the deepest, thickest missile tower forest I've ever made while playing against people. My "base" covered roughly 1/3 the entire map. Nuke anywhere in it and silo was almost guarenteed vetran status or better. But unless you got my commander I would hardly care - he was guarding my adv. vehicle plant (my ONLY advanced factory) to build Panthers. Missile towers, metal extractors, wind generators, and level 1 factories are pretty cheap to replace for a guy who's got 4+ c-units on reclaimation patrols in the middle of multiple multi-screenlength debris fields. :)

I managed to convince 2 diehard Core players firstly that Panthers were better than Goliaths (which I don't believe to be true, except in narrow instances) and lastly that Arm can rush and raid like there's no tommorrow. I ended up lagging out of the game about 1 1/4 hours into it, so technically I lost. There's reason to suggest that Vampyre booted me out of the game because he didn't want to lose.

PS: what map was it again that my ai makes fleas on?

David
--------
Subject: Re: TA ai, again...
Date: Sat, 27 Nov 1999 17:10:01 +0000 (GMT)

> BTW, my "fair" ai has been reffered to by me as
> "Mostly Harmless" ai as far
> back as v3.5 -- it's now v12+. It's ai reviewers and
> other people who've
> tried to stick other names on it, not I.
>
> "I get about 10 e-mails a day :).  I used to get
> 20-25."
> A lot of my emails are either rant-mail or asking me
> to explain how things
> work. Oftentimes, it's what I've wrote vast amounts
> about in my readme's for
> my files -- which is very irritating to have to
> answer again. I often just
> cut and paste from those readme's and email that. :p

Mines usually clan related, though I get some TA mail from the testers (if I'm lucky :)).

> "Probably the same people who became the TAK zealots
> :)"
> No, mostly it's not even related to TA -- it's on
> the MechWarrior 3 BUGFIX
> I'm working on. Back when I got regular similar
> emails about TA, it was how
> my ai was a cheat/trainer/whatever because it could
> be used on one computer
> in a multiplayer game and be "different" ai-wise
> than other computer's ai's.
> The FEW emails that even seem sort of related to
> cheating/hacking now either
> are asking me for cheats/hacks or are saying that TA
> Bugfix is/isn't
> compatible with the latest TA trainers out and would
> I "fix" that?

Mech 3 fix eh, doing anything about that terrible AI?

Not compatible with trainers, heh.  Fix it then break it with something else.

> "I just remember that I'm so right it's unbelievable :)."
> You can learn more from mistakes than successes.
>
> "Nerves?  I have nerves of cotton :)."
> I have had TA webmasters, TA IRCops, Cavedog
> employees, major 3rd party developers, and even
> average Joe's in the TA community telling me to go
> fuck myself at one time or another. Sometimes, they
> don't put it so mildly... And on at least one occassion
> they joined forces to really get the message
> accross. Oftentimes, this is even over stuff I am
> not responsible for and didn't do! (I've gotten multiple
> hate-mails for BAI, by people who thought I
> made it. If BSR only knew...) Othertimes, it's over
> complete misunderstandings where people took what I
> said EXACTLY BACKWARDS of both what my intent was and
> my wording. (Too many double-negatives, I guess.)

People hear what they want to hear.

> My health doesn't allow me the luxury of holding
> down a full-time job, or
> even a part-time job if it's strenuous.

Nerves?

> "I've gone back to Hawks.  By having a thick patrol
> route the AI builds up
> it's planes.  When I saw 12 hawks fly in an decimate
> an entire mini base in
> just 2 sweeps I knew I'd made the right choice :)"
> But include some Brawlers for diversity -- I do,
> just on the slim hopes they
> prove useful on occassion. :)

I use a 10, 3, 2 ratio for Lv 2, Fighters, Gunships, Bombers.

> "After all that I've swapped back :)"
> I think I've reversed-stance on MH's use of many
> units 2, 3, or even more
> times as it's progressed. Metal Makers have been a
> love-hate relationship
> since MH's very beginning.

I change the AI more often than my underwear :)

> "As far as I could tell TAK doesn't seem to suffer
> from this.  Guess they
> fixed at least 1 thing."
> That is also probably why TA:K runs so slowly.

I think curmmy optomization was also a factor.

> "Rocko's are brilliant base decimators.  Especially
> for the numerous level 1
> structures.  Send in a demo squad of Rocko's and
> BOOM, bye bye base."
> My strategy on The Cold Place is to go for a Rocko
> rush. I've killed players
> off in under 15 minutes doing that.
>
> "Ah, well Queller can most definately pull off 7
> Golaiths by the 30th minute
> with escort (depending on casulties suffered).  I
> watched it yesterday, and
> it must have cycled through several Golaiths by that
> time.  It all depends
> on how well it does, and how early it goes level 2."
> Those are some VERY big claims for Queller -- 7
> Goliaths is almost 5k metal,
> and drained at a considerable rate. Are you sure it
> could do that on Painted Desert?

Well taking into account the fact that I had to estimate, I'd say from 4-8, but it can be quite good with getting Goliaths out if it goes lv 2 early enough, though this is NOT a common event.

> "I go for Morties mainly because their range means
> they aren't so effected by the sleep bug, and also
> they're fast, finally they aren't going to be
> beat by a player who makes sure the nearest target
> is 10 DT'd HLT's."
> MH's specialty (as I've found out the hard way) is
> its ability to remove fixed defenses with ease if
> not PERM LOS. Stuff like Morties, Dominators,
> Diplomats, and even Storms, Slashers, and Goliaths
> are the contributing reasons. :)

I looked around, said Oooooo this looks like a good place to porc (I can't help it :)), then came the AI merls :)

> "Well if it's gonna be a big pat on the back session
> I'd say MH is probably
> better on larger maps because it goes construction
> sooner :)"
> If not too soon. Versions to follow shall act
> slightly differently on c-units...

I look forward to seeing a more Quellerised MH <g>

> "I have still to find an AI that can play Ashap
> Plateau, that metal spacing
> is anti-AI from the word go."
> Ashap Plateau has contour errors in it, and makes it
> very difficult to build
> factories anywhere other than the plateaus
> themselves because of small
> "hills" in the lower land. All told, Counterstrike
> ai still might play
> "decently" there, but even BAI might not.
>
> "Same here.  I keep them for the times when the
> player hits solar first. The good old AI always
> seems to put it's lasers BEHIND solars, so that by
> the time the player reaches them, they don't have
> anything to power them :)"
> I put MY HLT's behind my solars often too! But I
> ussually have excess power by the time I do this,
> and at least a small mobile force to back my
> HLT's up with.

Energy is never a problem.

> "I just dispair as the solar code is great, but it
> builds 2 winds then decides prehaps the next best
> thing is 40 MEX's."
> Nah, it mixes in a few Guardians, factories, and
> HLT's as well -- for variety. :p

The laws of AI dictate it to be so.

> "I get Queller to build them so it doesn't take a
> beating from early DT's
> (though it's not as if Stumpies are kings of
> anti-DT), also on maps like
> Yerrot Mountains they're invaluable for nailing
> MEX's on top of pillars.
> Finally in wreckage areas they can come in useful."
> I'm at odds on adding Stumpies back to MH ai and
> nixing Hammers...

I'm sticking with Stumpies since Queller almost always goes vehicle first.

> "I wasn't impling that you didn't know, it just
> sounded better than saying flying BOOM :)."
> CyberDrgn was HURTING bad for metal, and the best
> idea you can come up for
> his Atlas is for him to build a Kbot Lab, an
> Advanced Kbot Lab, then a
> crawling bomb? :p His 1 Atlas could be used in many
> interesting and annoying
> ways... you ever had your c-units air-napped?

Ummmm, please ignore my bout of stupiditity, it was my invisible friends fault, yes that'll do it.

> "I still think capturing is generally a waste of time."
> Ussually it is, but capturing fully-built units is
> cheaper than reclaiming them and building your own.
> Especially when you cannot be stopped from doing so!

But you also have to spend time micromanaging it, and diverting your Commander to capture rather than expand.

> "Air first on Greenhaven, Ooooooooo I wanna crush him :)"
> Come to annihilated's irc chatroom sometime, and you
> MIGHT get a chance...

*sound of running feet*

> "Heh, can't keep up eh :P  For me he could have
> nabbed my Commander and I'd
> never have noticed, but a Flash I'd spot in an
> instance.  Think my priorities are off? :)"
> More like my flash tanks were spread out accross the
> map -- plus I had *HIS* flash tanks to worry about
> attacking me from the bottom of the map.

I spend too much time microing the attack and forget that my base needs more than just a Comm :)

> "HAH!  Good old demo mission 2, send in the PeeWees!
> Even as early as then I  suspected that PeeWees were
> better than A.K.'s."
> AK's have the same laser as the Weasal, PeeWees have
> the same weapon as the Flash tank. If AK's had the
> same weapon as the Instigator, they'd be almost
> twice as dangerous.

They would still have an inferior turn rate and acceleration.

> "There I go using Brit terms again, suck."
> Blow here means a very extreme form of suck. It'd
> suck to find you're out of
> ice cream. It'd blow to find out you're out of food
> and snowed in for a few days. See the differance? :p

You're right, running out of ice cream is a hell of a lot more worrying.

> "I would prefer a Mav because by that point you're
> gonna have a lot of structures.  You could be
> approaching the unit limit and you need units that
> operate well in small numbers.  Pelicans need to be
> in fairly large packs, and this IMHO is a bad idea
> ta level 2 because of the unit limit."
> My ai won't hit 250 unit limit quickly -- and I
> don't think it ever reaches 500 unit max unless I
> have some 3rd party units installed that it doesn't
> have limits on.

Staying with Mavs.  I tired pelicans, but firstly the AI doesn't wait for hordes, it attacks when it has 5 units.  Secondly, their cost per tick stuffs the way Queller tries to play, so I'm back with Mavs.

> "Hell no :)  The ARM can operate just as well at
> level 2 because the Mavs cost next to nothing to
> build, so while the CORE may slow down with these
> expensive units, the ARM does no such thing (and can
> afford to get Mavs out LONG before the CORE gets Gols)."
> Core can struggle to get out Goliaths and get them
> out at a reduced rate of 1/3 normal speed and STILL
> produce Goliaths slightly quicker than Arm can
> produce Mavericks. It's my opinion that Mavericks
> can kill about an equal number of Goliath tanks,
> so that's a fair trade in metal but not a fair
> trade in energy. Also, because of the sleep bug the
> ai's Mavericks spend very little time in firing
> range of enemy units (they kill them too quickly
> :p) -- I've seen pathetic ai Dominators absolutely
> murder Mavericks because a single Instigator got
> close to the Mavericks and killed and then the
> Mavericks took a 5+ second nap.

However often they get lucky and come accross a whole line of troops and shread them all, and usually they are worth the time.

> "Have players missed the very loooooooonnnnnnnnnnng
> build time of a Mav?  I still disagree and will unit
> time comes to an end :)"
> Long buildtimes is a problem only for those who
> aren't inventive... See the attached Pic for details.

So much for that pic, you're losiing your memory :)

> "I want my AI to be as good as possible, and to do
> this it should beat average players.  Average players
> out number pro's by about 4 or 5 to 1 I'd
> guess, not only that, but if you're a pro you're
> also going to play BAI or CounterStrike not Queller."
> Only rush junkies play BAI and CounterStrike and
> like them... Even I prefer playing against MH than
> CounterStrike. CS's dammed near unbeatable on larger
> maps! And I don't much like playing on small maps
> versis the ai...

But if you're good you'll play several AI's (I'd hope).

> "I go around them.  If you can be bothered, a
> patroling Peeper by an HLT draining 75 energy from
> their economy until they notice and shut it down
> (then you Stumpy rush it :)) is always fun.  Course
> he can send in AA, but that's less for my force to
> deal with."
> If someone builds an HLT without putting up at least
> a MT or 2 nearby, they are lousy players! That's
> like building a bertha without AA around it! A
> single level 1 bomber can MAKE TIME to kill HLT's
> that are in inconvenient locations. :)

Call me newbie crusher :)

> "Lack of Geo though.  Actually Queller seems to have
> fun taunting me here (MH has done it too), there is
> so much metal so close together the AI can go
> on a resource frenzy with it's Commander."
> Plains and Passes has more geo vents on it than
> Painted Desert does -- they're just all near the center.

Yeah but PD only has 4 so it's not hard to outnumber it.

> "I remember when somone tried to rush me with
> everything they had, 4 Telsa coils held them off :)
> Heh.  The AI beat me later, I was never very good at
> RA :)"
> I got pretty good at RA, but one of my friends got
> far better at it. On TA, he was better still! I made
> my first ai for TA as a TRAINER for me to get me
> the skills I needed to beat him. It didn't work too
> good tho, but I learned by playing online games so
> I went from average player to roughly an expert
> in about 6 months. His skills at TA have deteriorated
> because he doesn't play it much (if any)
> anymore, while mine is about the same so I can
> probably beat him now... just barely.

I'm worse since I usually watch rather than play.

> "Well we let each other know what's going on, but I
> don't remember us doing much about it :)  Mind you
> we exchange troops some times, or send in task
> forces. The best team work we did was on Greehaven,
> I was holding my Queller with some heavy defence
> and strafing Rocko's, EMG was being beaten back, and
> I sent out a large task force of Flashes to raid his
> foe.  There must have been 8 or 9 fights going on
> after that.  I had to quit though, and he
> eventually lost.  I blame him :)"
> Teamwork doesn't guarentee a win, it just ups the
> odds. In our LAN games here, I used to give the ai's
> loaded nuke silos, if we're winning by a large margin. :p
> Worse still, I'd give the ai my entire base - and go
> into watch mode and laugh as the other players got stomped...

You evil man :)

> "Not THAT early."
> Very early can even mean before 15 minutes. Ussually
> it seems that about the
> time MH starts building guardians en mass it's also
> started building
> advanced factories, berthas, and fusion. I've seen
> it stay below 500 metal
> for something like 40 minutes gametime straight. But
> I can promise you that
> were I playing against it I wouldn't have been able
> to notice!
>
> "I tried to raid it once, damn HLT's.  It builds so
> many factories and
> requires so few resources that the whole base is one
> big lump, and HLT's
> have very little trouble covering it.  It's also
> Nuke bait :)"
> BAI and Counterstrike both make very compact bases.
> Counterstike, because it
> uses NORMAL factories, has to build more and thus
> spreads out more, so it is
> slightly less vulnerable to nukes.
>
> Most players when playing on large open land maps
> like Painted Desert seldom ever bother with antinukes,
> or seldom build more than one because their bases
> in LONG games can often cover many screens in area.

I don't bother with them because all my games end by 30 min :)

> I remember playing a 2 Core vs 2 Arm players (me)
> game on The Cold Place.
> One of my opponents was Vampyre, who as a really
> good Core player. I was
> pissed because my ally was lagging really bad and
> was killed off less than
> 15 minutes in the game by a combined attack by both
> opponents. After that
> point I was fighting the battles as revenge. I got
> the attitude that they
> ganged up on my partner, (who seemed to be a newbie)
> and they were GOING TO
> kill me too -- and I was going to make them pay for
> the right to do so!
>
> About 50 minutes into the game one of my opponents
> threatened to nuke me. I
> said, "Sure, go ahead -- I'm at unit max, this will
> just free up more units
> for me to throw at you." At the time, I think I had
> all of 5 factories. But
> they were screens apart, surrounded by the deepest,
> thickest missile tower
> forest I've ever made while playing against people.
> My "base" covered
> roughly 1/3 the entire map. Nuke anywhere in it and
> silo was almost
> guarenteed vetran status or better. But unless you
> got my commander I would
> hardly care - he was guarding my adv. vehicle plant
> (my ONLY advanced
> factory) to build Panthers. Missile towers, metal
> extractors, wind
> generators, and level 1 factories are pretty cheap
> to replace for a guy
> who's got 4+ c-units on reclaimation patrols in the
> middle of multiple
> multi-screenlength debris fields. :)
>
> I managed to convince 2 diehard Core players firstly
> that Panthers were
> better than Goliaths (which I don't believe to be
> true, except in narrow
> instances) and lastly that Arm can rush and raid
> like there's no tommorrow.
> I ended up lagging out of the game about 1 1/4 hours
> into it, so technically
> I lost. There's reason to suggest that Vampyre
> booted me out of the game
> because he didn't want to lose.

HAH!  I tired to rush a load of Golaiths with Panthers once, not a pretty sight.  Panthers are more Level 1 shredders, let the Mavs take care of the armour.

> PS: what map was it again that my ai makes fleas on?
>
> David
--------
Subject: Re: TA ai, again...
Date: Sat, 27 Nov 1999 17:10:01 +0000 (GMT)

"Mech 3 fix eh, doing anything about that terrible AI?"
Sort of, I'm giving many mechs more heatsinks (typically double) -- this allows them to fire faster.

"Not compatible with trainers, heh.  Fix it then break it with something else."
So far I've told people that I don't try to make TA Bugfix work with trainers, but if it does happen to work with trainers I'm not going to "fix" it. I'm not for trainers -- but making TA incompatible with them is near-impossible!

"Nerves?"
No, I have a parasitic mold (similar to mold that "eats" on bread) that weakens me. Almost everyone has tiny amounts of it in their intestines to help breakdown food, except mine seems to have spread to other parts of my body. Doctors when they did a colonoscopy on me said that my large intestine's walls looked like it'd been hit by a tiny shotgun in places -- it had many hundreds of pinprick holes in it and was irritated there. I believe that's where the mold latched onto the intestinal wall and punched its way out with root-like apendages. Anyhow, I get bad headaches and various aches and pains from mostly mild to hurting far too bad to sleep from it. I've been given LOTS of antibiotics for sinus infections when I was younger, which doesn't do anything to the mold -- except kill one of *ITS* natural enemies. (Penicillin comes from mold, btw!) And since I've had this problem for many years, the remaining mold sporetypes in me are VERY resistant to both my body's immune system AND to many forms of anti-fungal medications. I can sort of starve it by eating absolutely no surcrose (table sugar) for weeks on end and nothing but vegetables, some metal, and very few fruits (fruit sugar is almost as bad...) but that gets OLD quick -- plus, if you haven't noticed, many foods have added sugar in them "for extra flavor"... Because the mold begins its feeding cycle whenever I eat foods that it loves, like breads, sugars, and peanuts, I test allergic to these foods. I get sleepy after eating often because of this, because the mold's digestive juices are toxic depressants. Another sign that I've got this problem is I get sores in my mouth (ulcers) when I overtax myself phsically or mentally (particularly from lack of sleep!) that don't heal for as long as 2 weeks straight -- then they sometimes heal up in less than a day. It just shows my immune system is still working, but is perhaps stretched to its limits at times by this. Since this mold was originally sort-of needed in the body (but only in the digestive tract) for the body to function correctly, my immune system isn't keyed to hunt it down and likewise it doesn't normally attack individual cells in my body (when it does, it generates a powerful immune response) -- it just feeds off whatever is distributed by my bloodstream. All in all, it's enough to make me sick a lot but not by far enough to kill me anytime soon.

Btw, When your nerves go, your mind is soon to follow.

"I use a 10, 3, 2 ratio for Lv 2, Fighters, Gunships, Bombers."
Ai-only bombers work pretty well, as one-shots over AA and murderous over undefended metal extractor farms. The other 2 work pretty well in the same places, just not as quickly.

"I think curmmy optomization was also a factor."
I was told that the amount of animations on individual units was magnitudes more than in TA. Where most TA units had VERY little animation while moving, TA:K units had many discrete parts of their units moving in various ways even when standing still. The COB files were 4 to 100 times more complex as a result than TA's.

"Well taking into account the fact that I had to estimate, I'd say from 4-8, but it can be quite good with getting Goliaths out if it goes lv 2 early enough, though this is NOT a common event."
Oh, I thought you meant 7+ Goliaths at once by the ai before the 30 minute mark WAS common!

"I looked around, said Oooooo this looks like a good place to porc (I can't help it :)), then came the AI merls :)"
I quit making that mistake versis MH and CS ai a while back -- now it's place mobile units out front and build a small defensive wall behind it.

"I look forward to seeing a more Quellerised MH <g>"
They ALREADY play similar, compaired to other ai's. The main differance may be a timechange in availability of units, not much of a ratio change.

"Energy is never a problem."
It is for me for the first 5-10 minutes, and immediately after I get my first MMM. But that's almost never the case when I play with more than 1k E at the start.

"The laws of AI dictate it to be so."
They certainly make anything ELSE difficult!

"I'm sticking with Stumpies since Queller almost always goes vehicle first."
It's my impression that Stumpies early on is like Merls built before the first 5 minutes. VERY few things would be DT'ed so that Stumpies (or Merls) are really needed, but either one will be slaughtered by a player who wants to drive their flash tank force into the middle of the ai's attack force.

"Ummmm, please ignore my bout of stupiditity, it was my invisible friends fault, yes that'll do it."
You're prejudiced for how powerful flying bombs are -- and I agree that they are WAY too powerful in internet games.

"But you also have to spend time micromanaging it, and diverting your Commander to capture rather than expand."
He was low on metal, and he didn't really want to keep his commander building things at the time so his drain wouldn't be increased, instead he kept his commander near an undefended portion of his base while capturing flashes. I *DID* wander over there with 2 flashes a little later, and he d-gunned them. Had his commander been out in the field, I would've had fun with his undefended base. :)

*sound of running feet*
Getting slaughtered really badly by these good players is a good way to keep your ego in check. And because you've never been there much before (if any) they'd think you were a newbie (even if you aren't!) and probably PAIR you with a good player in a 2 vs 2 game.

"I spend too much time microing the attack and forget that my base needs more than just a Comm :)"
I borderline porc, even pretty early on. But all I normally make is MT's, just far enough apart to barely give each other full LOS. It's a nasty web to drive in with anything other than long-ranged units, because you really can only kill 1 MT at a time while 5+ more MT's kill your units. And it's the CHEAPEST way to get good LOS in many places so you seldom ever get blind-sided by a sneak attack. All in all though, I seldom top 100 MT's -- even when covering an area as big as an island on Metal Isles -- that's how THIN my MT's are spread!

"They would still have an inferior turn rate and acceleration."
That can be a plus, if their weapons track consderably slower. I see many units able to turn almost as fast, if not faster, than their turrets/upper torsos can rotate. Makes them almost useless in melee combats. :p

"Staying with Mavs.  I tired pelicans, but firstly the AI doesn't wait for hordes, it attacks when it has 5 units.  Secondly, their cost per tick stuffs the way Queller tries to play, so I'm back with Mavs."
Sounds like lots of Goliaths for Core is disaster for Queller then?

"However often they get lucky and come accross a whole line of troops and shread them all, and usually they are worth the time."

"So much for that pic, you're losiing your memory :)"
Maybe I'll remember this time. I remembered after I hit send last time, and had just shut down my computer. (I needed sleep really badly, and not showing you the picture just yet should just keep you wondering...)

"But if you're good you'll play several AI's (I'd hope)."
Only if I'm allied against them. MH and Queller's concentrated attack force makes it impossible to be strong enough across a vast front to stop them -- either I can handle only one such attack by losing most of my mobile forces, or I have to have massive choke-points so I can fight numerically-superior numbers to a standstill with mobile units backed up by base defenses.

"Call me newbie crusher :)"
Ok, newbie crusher.

"Yeah but PD only has 4 so it's not hard to outnumber it."
I only remember 3. :)

"I'm worse since I usually watch rather than play."
Same, but I still toy around enough to keep my skills semi-good.

"I don't bother with them because all my games end by 30 min :)"
Either you're giving up too quickly, or like you say you're mostly playing against newbies and you're smoking them!

"HAH!  I tired to rush a load of Golaiths with Panthers once, not a pretty sight.  Panthers are more Level 1 shredders, let the Mavs take care of the armour."
I had numerical superiority against the Goliaths of about 2 to 1, and lost less than half my force. I immediately counter-attacked with my remaining forces as well. The Goliaths had wandered into the edges of my base's MT web, so the Panthers kept them busy while everything else killed them. They just thought it was the Panthers doing it. :p

I think both MH ai and Queller ai would play BEST at 1k metal and 10k energy, unless wind generators were the primary energy source. At least for a LONG while there probably wouldn't be an energy stall, and it might not build many factories in a row either. :)

Btw, Queller ai's hacked factories could get a couple minor changes. As it is now, players that capture them will immediately tell something's wrong with them because units are missing on their build menus. (No Levelers, Minelayers, or 3rd party units.) You want me to look into changing that in any way?
--------
Subject: Re: TA ai, again...
Date: Mon, 29 Nov 1999 17:13:40 +0000 (GMT)

> Subject: Re: TA ai, again...
> Date: Sat, 27 Nov 1999 17:10:01 +0000 (GMT)
>
> "Mech 3 fix eh, doing anything about that terrible
> AI?"
> Sort of, I'm giving many mechs more heatsinks
> (typically double) -- this
> allows them to fire faster.

Always more of an Earthsiege series fan myself.  Oh, and the Starsiege music rocks :)

> "Not compatible with trainers, heh.  Fix it then
> break it with something else."
> So far I've told people that I don't try to make TA
> Bugfix work with
> trainers, but if it does happen to work with
> trainers I'm not going to "fix"
> it. I'm not for trainers -- but making TA
> incompatible with them is near-impossible!

Cheats seem to be the only thing that's compatible with EVERYTHING :)  Ironic, no?

> "Nerves?"
> No, I have a parasitic mold (similar to mold that
> "eats" on bread) that
> weakens me. Almost everyone has tiny amounts of it
> in their intestines to
> help breakdown food, except mine seems to have
> spread to other parts of my
> body. Doctors when they did a colonoscopy on me said
> that my large
> intestine's walls looked like it'd been hit by a
> tiny shotgun in places --
> it had many hundreds of pinprick holes in it and was
> irritated there. I
> believe that's where the mold latched onto the
> intestinal wall and punched
> its way out with root-like apendages. Anyhow, I get
> bad headaches and
> various aches and pains from mostly mild to hurting
> far too bad to sleep
> from it. I've been given LOTS of antibiotics for
> sinus infections when I was
> younger, which doesn't do anything to the mold --
> except kill one of *ITS*
> natural enemies. (Penicillin comes from mold, btw!)
> And since I've had this
> problem for many years, the remaining mold
> sporetypes in me are VERY
> resistant to both my body's immune system AND to
> many forms of anti-fungal
> medications. I can sort of starve it by eating
> absolutely no surcrose (table
> sugar) for weeks on end and nothing but vegetables,
> some metal, and very few
> fruits (fruit sugar is almost as bad...) but that
> gets OLD quick -- plus, if
> you haven't noticed, many foods have added sugar in
> them "for extra
> flavor"... Because the mold begins its feeding cycle
> whenever I eat foods
> that it loves, like breads, sugars, and peanuts, I
> test allergic to these
> foods. I get sleepy after eating often because of
> this, because the mold's
> digestive juices are toxic depressants. Another sign
> that I've got this
> problem is I get sores in my mouth (ulcers) when I
> overtax myself phsically
> or mentally (particularly from lack of sleep!) that
> don't heal for as long
> as 2 weeks straight -- then they sometimes heal up
> in less than a day. It
> just shows my immune system is still working, but is
> perhaps stretched to
> its limits at times by this. Since this mold was
> originally sort-of needed
> in the body (but only in the digestive tract) for
> the body to function
> correctly, my immune system isn't keyed to hunt it
> down and likewise it
> doesn't normally attack individual cells in my body
> (when it does, it
> generates a powerful immune response) -- it just
> feeds off whatever is
> distributed by my bloodstream. All in all, it's
> enough to make me sick a lot
> but not by far enough to kill me anytime soon.

Good excuse to be damn tetchy and get away with it :)

> Btw, When your nerves go, your mind is soon to follow.

Too late.

> "I use a 10, 3, 2 ratio for Lv 2, Fighters,
> Gunships, Bombers."
> Ai-only bombers work pretty well, as one-shots over
> AA and murderous over
> undefended metal extractor farms. The other 2 work
> pretty well in the same
> places, just not as quickly.

I try to get it to focus on one thing (so it can swarm with it and keep it together), and have support from other items for interest.

> "I think curmmy optomization was also a factor."
> I was told that the amount of animations on
> individual units was magnitudes
> more than in TA. Where most TA units had VERY little
> animation while moving,
> TA:K units had many discrete parts of their units
> moving in various ways
> even when standing still. The COB files were 4 to
> 100 times more complex as
> a result than TA's.
>
> "Well taking into account the fact that I had to
> estimate, I'd say from 4-8,
> but it can be quite good with getting Goliaths out
> if it goes lv 2 early
> enough, though this is NOT a common event."
> Oh, I thought you meant 7+ Goliaths at once by the
> ai before the 30 minute mark WAS common!

Hell no :)  Don't ask me why, but Queller seems to have a love affair with Painted Desert, I guess it's just because for about the first 4 versions I used PD almost exclusively.  Queller will generally play it's best game on this map.

> "I looked around, said Oooooo this looks like a good
> place to porc (I can't
> help it :)), then came the AI merls :)"
> I quit making that mistake versis MH and CS ai a
> while back -- now it's
> place mobile units out front and build a small
> defensive wall behind it.

Damn CC missions, AI was so crap I'd spend all my time making the biggest and best damn base, using sandbags to allow me to make fences far enough out to stop rocket attacks on my buildings, and then I'd fortify my my tiberium fields, start check posts and have Mammoth tanks by the front gate.  Ah those were the days :)

> "I look forward to seeing a more Quellerised MH <g>"
> They ALREADY play similar, compaired to other ai's.
> The main differance may
> be a timechange in availability of units, not much
> of a ratio change.

Yes, MH must eventually give in to the force of Queller :)

> "Energy is never a problem."
> It is for me for the first 5-10 minutes, and
> immediately after I get my
> first MMM. But that's almost never the case when I
> play with more than 1k E at the start.

Actually with my new start, energy has become a bit scarce, but I reclaim stuff now (never used to bother <g>)

> "The laws of AI dictate it to be so."
> They certainly make anything ELSE difficult!

Like losing? :)

> "I'm sticking with Stumpies since Queller almost
> always goes vehicle first."
> It's my impression that Stumpies early on is like
> Merls built before the first 5 minutes. VERY few
> things would be DT'ed so that Stumpies (or Merls)
> are really needed, but either one will be
> slaughtered by a player who wants to drive their
> flash tank force into the middle of the ai's
> attack force.

I don't let it take them in large numbers early, and it rarely has them in it late, but they make a nice supplement (and if lucky enough are a good HLT distraction)

> "Ummmm, please ignore my bout of stupiditity, it was
> my invisible friends
> fault, yes that'll do it."
> You're prejudiced for how powerful flying bombs are
> -- and I agree that they
> are WAY too powerful in internet games.

Never used them myself, and my games never last long enough for them (I've won or lost).

> "But you also have to spend time micromanaging it,
> and diverting your
> Commander to capture rather than expand."
> He was low on metal, and he didn't really want to
> keep his commander building things at the time so
> his drain wouldn't be increased, instead he
> kept his commander near an undefended portion of his
> base while capturing flashes. I *DID* wander over
> there with 2 flashes a little later, and he
> d-gunned them. Had his commander been out in the
> field, I would've had fun with his undefended base. :)

Only if you'd known that it was.  After all, if he kept D-Gunning your units then you were wasting metal, and maybe you'd have been more cautious?  Well I don't know.

> *sound of running feet*
> Getting slaughtered really badly by these good
> players is a good way to keep your ego in check.
> And because you've never been there much before
> (if any) they'd think you were a newbie (even if
> you aren't!) and probably PAIR you
> with a good player in a 2 vs 2 game.

I know exactly how good I am, distinctly average.  I was probably a better a year ago because I'd play online a lot, now I rarely bother.

> "I spend too much time microing the attack and
> forget that my base needs more than just a Comm :)"
> I borderline porc, even pretty early on. But all I
> normally make is MT's, just far enough apart to
> barely give each other full LOS. It's a nasty web
> to drive in with anything other than long-ranged
> units, because you really can only kill 1 MT at a
> time while 5+ more MT's kill your units. And it's
> the CHEAPEST way to get good LOS in many places so
> you seldom ever get blind-sided by a sneak attack.
> All in all though, I seldom top 100 MT's --
> even when covering an area as big as an island on
> Metal Isles -- that's how THIN my MT's are spread!

I just find having a beautiful base with defences, walls and entrances it brilliant, if 100% impractical. :)

> "They would still have an inferior turn rate and
> acceleration."
> That can be a plus, if their weapons track
> consderably slower. I see many units able to turn
> almost as fast, if not faster, than their turrets/upper
> torsos can rotate. Makes them almost useless in
> melee combats. :p

Stumpies per chance? :)  They seem to be constantly fighting themselves, and this is also what can make a Bulldog useless in the hands of an AI.

> "Staying with Mavs.  I tired pelicans, but firstly
> the AI doesn't wait for hordes, it attacks when it
> has 5 units.  Secondly, their cost per tick
> stuffs the way Queller tries to play, so I'm back
> with Mavs."
> Sounds like lots of Goliaths for Core is disaster
> for Queller then?

No, When it goes Gols it usually only goes Gols, where as ARM goes both Adv Kbots and Vehicles pretty close together.  I find Mavs a safer bet and reduce the chance of nano-stall.

> "However often they get lucky and come accross a
> whole line of troops and shread them all, and usually
> they are worth the time."
>
> "So much for that pic, you're losiing your memory :)"
> Maybe I'll remember this time. I remembered after I
> hit send last time, and had just shut down my computer.
> (I needed sleep really badly, and not showing you the
> picture just yet should just keep you wondering...)

Whoever the foe was, they must have been fairly passive (or dead and you were starting to write your name in DT's :)).

> "But if you're good you'll play several AI's (I'd hope)."
> Only if I'm allied against them. MH and Queller's
> concentrated attack force makes it impossible to be
> strong enough across a vast front to stop them --
> either I can handle only one such attack by losing
> most of my mobile forces, or I have to have massive
> choke-points so I can fight numerically-superior
> numbers to a standstill with mobile units backed up
> by base defenses.

I can beat 2 on the small maps (Quellers), but nothing above 9x9.  Then I usually win 1 on 1's up till 18x18.

> "Call me newbie crusher :)"
> Ok, newbie crusher.

On second though, Lord and Master has a nice ring to it <g>

Newbie Crusher can be my middle name.

> "Yeah but PD only has 4 so it's not hard to
> outnumber it."
> I only remember 3. :)

And that's why this boy plays mapped mode :)

> "I'm worse since I usually watch rather than play."
> Same, but I still toy around enough to keep my
> skills semi-good.

I play with the AI (and sometimes with EMg against it), I'm a man vs machine person at heart.  I have great fun, wonder why I don't play more often, then can't be bothered to play again.  I am incredibly lazy, and find it hard work to drag myself to start something new, and find it hard to stop working on something I'm doing because my perfectionist side kicks in.

> "I don't bother with them because all my games end
> by 30 min :)"
> Either you're giving up too quickly, or like you say
> you're mostly playing
> against newbies and you're smoking them!

I just can't handle too many things going on at once. If 3.1 got a Bertha I used to quit because I got bored trying to stop them (cause I now know to bomb the things).

> "HAH!  I tired to rush a load of Golaiths with
> Panthers once, not a pretty sight.  Panthers are
> more Level 1 shredders, let the Mavs take care of
> the armour."
> I had numerical superiority against the Goliaths of
> about 2 to 1, and lost less than half my force. I
> immediately counter-attacked with my remaining
> forces as well. The Goliaths had wandered into the
> edges of my base's MT web, so the Panthers kept them
> busy while everything else killed them. They
> just thought it was the Panthers doing it. :p

Unfortunately, I tried to do it in a narrow valley. The Charge of The Lightning Brigade :)

> I think both MH ai and Queller ai would play BEST at
> 1k metal and 10k energy, unless wind generators were
> the primary energy source. At least for a LONG while
> there probably wouldn't be an energy stall, and it
> might not build many factories in a row either. :)

I'll have to try that sometime.

> Btw, Queller ai's hacked factories could get a
> couple minor changes. As it is now, players that
> capture them will immediately tell something's wrong
> with them because units are missing on their build
> menus. (No Levelers, Minelayers, or 3rd party units.)
> You want me to look into changing that in any way?

I've been meaning to work on that, but it involves trying to find out how (you told me but I lost the e-mail), so if you don't mind I'd love that!  Another mention of Switeck's name makes the credits <g>
--------
"Always more of an Earthsiege series fan myself.  Oh, and the Starsiege music rocks :)"
Battletech, both as a viable plot and as a game, has been around for over 20 years. It has a more complex storyline and richer universe than even Star Wars. But it is very klunky in places... mechs aren't the "gods" they seem, even 100-tonners can be destroyed from a single lucky shot by a powerful weapon. And all the skills of the mechwarrior pilot won't help a bit...

"Good excuse to be damn tetchy and get away with it :)"
And unfortunately causing me to feel bad and hate life a lot more than is healthy mentally.

"Hell no :)  Don't ask me why, but Queller seems to have a love affair with Painted Desert, I guess it's just because for about the first 4 versions I used PD almost exclusively.  Queller will generally play it's best game on this map."
Mostly Harmless's overtested map is default on Greenhaven. This should explain a lot of it. :)

"Yes, MH must eventually give in to the force of Queller :)"
When Queller beats it consistantly 1 on 1 on Painted Desert, maybe I will make it vastly more Queller-like.

"Actually with my new start, energy has become a bit scarce, but I reclaim stuff now (never used to bother <g>)"
On many of the green maps I typically play on, I reclaim to get all the trees out of the way.

"Never used them myself, and my games never last long enough for them (I've won or lost)."
You must not have ever seen the "Mexican standoff" results I've seen on larger maps -- both me and my opponent/s make bases that it is almost suicidal to attack. Missile towers 3 screens deep is one of the best detterants of a groundforce that I know of. However, I *LOVE* building a jammed guardian outside the missile tower's outer edge, and radar-target it on radar-blips. 1 good guardian hit is overkill on a MT.

"Only if you'd known that it was.  After all, if he kept D-Gunning your units then you were wasting metal, and maybe you'd have been more cautious?  Well I don't know."
If he's d-gunning my flash tanks, HE's wasting "free" metal that I'm giving him.

"I know exactly how good I am, distinctly average.  I was probably a better a year ago because I'd play online a lot, now I rarely bother."
Online games lag really bad, it's so disappointing. And my friends prefer "newer" games and are now bored with TA. The sad part is they're mostly older than I am. Reason why that's sad is their attention span for games is crippled by other obligations... work!

"I just find having a beautiful base with defences, walls and entrances it brilliant, if 100% impractical. :)"
Play the Krogoth Encounter Mission on hard, if you beat it then I *KNOW* you're at least above average if not an expert. It will also FORCE you to make effective AA defenses and possibly killer ground defenses as well.

"Stumpies...seem to be constantly fighting themselves, and this is also what can make a Bulldog useless in the hands of an AI."
OTA Hammers are even worse.

"No, When it goes Gols it usually only goes Gols, where as ARM goes both Adv Kbots and Vehicles pretty close together.  I find Mavs a safer bet and reduce the chance of nano-stall."
That's a major way MH and Queller differ -- Core's choice between Adv Kbots and Vehicles is exactly opposite.

"Whoever the foe was, they must have been fairly passive (or dead and you were starting to write your name in DT's :))."
Hardly! I've been in games where both sides are racking up 100+ kills every 10 minutes. And this is barely 30 minutes into the game! If both side's gets their bases built up early, productions render EVERY unit except a krogoth as just another mark on the killboard 5 seconds after you first see it.

"I can beat 2 on the small maps (Quellers), but nothing above 9x9.  Then I usually win 1 on 1's up till 18x18."
I thought I could still beat 3 MH's on hard on Greenhaven. They beat me soundly, so I changed to 1 Arm ai and 1 Core ai -- they beat me soundly. I changed to 1 Arm ai (the harder of the 2) and barely won.

"On second though, Lord and Master has a nice ring to it <g>"
Maybe Queller calls you that, but I think Newbie Crusher is far more appropriate. I will praise your ability after you beat a Gnug or 2 or some of the Swedish Yankspankers.

"And that's why this boy plays mapped mode :)"
Even many experts plain hate unmapped mode.

"I play with the AI (and sometimes with EMg against it), I'm a man vs machine person at heart."
Co-Op for me. Wanting to knock each other over the head afterwards (not that we don't anyway!) isn't my cup of tea. And 100% competitive generates that.

"I just can't handle too many things going on at once."
I hope you don't have 6+ classes of school a day then.

"Unfortunately, I tried to do it in a narrow valley. The Charge of The Lightning Brigade :)"
Once Panthers start gumming themselves up with wreckage, just turn them around and flee - save them for another battle elsewhere.

"I've been meaning to work on that,"
I'll look into fixing the factories in the next week.
--------
> (I can't remember if I've sent this, please ignore
> if I did!)
>
> "Always more of an Earthsiege series fan myself.
> Oh, and the Starsiege
> music rocks :)"
> Battletech, both as a viable plot and as a game, has
> been around for over 20 years. It has a more complex
> storyline and richer universe than even Star
> Wars. But it is very klunky in places... mechs
> aren't the "gods" they seem, even 100-tonners can
> be destroyed from a single lucky shot by a powerful
> weapon. And all the skills of the mechwarrior pilot
> won't help a bit...

I saw the game in a shop once I think. Always loved board games.

> "Good excuse to be damn tetchy and get away with it :)"
> And unfortunately causing me to feel bad and hate
> life a lot more than is healthy mentally.

Yeah, but you won't suddenly have a big depression dip and enter a mid-life crisis.

> "Hell no :)  Don't ask me why, but Queller seems to
> have a love affair with Painted Desert, I guess it's
> just because for about the first 4 versions I
> used PD almost exclusively.  Queller will generally
> play it's best game on this map."
> Mostly Harmless's overtested map is default on
> Greenhaven. This should explain a lot of it. :)

I've started to use Greenhaven now, probably because I like Green maps, and I can only take so many acres of flat featureless desert.

> "Yes, MH must eventually give in to the force of
> Queller :)"
> When Queller beats it consistantly 1 on 1 on Painted
> Desert, maybe I will make it vastly more Queller-like.

And the day Queller attacks from about the 3 minute mark like any normal average human would, then I'll trust those tests :)

> "Actually with my new start, energy has become a bit
> scarce, but I reclaim stuff now (never used to bother <g>)"
> On many of the green maps I typically play on, I
> reclaim to get all the trees out of the way.

I just demolish them as my large force wades through the trees.  I should be an Orc :)

> "Never used them myself, and my games never last
> long enough for them (I've won or lost)."
> You must not have ever seen the "Mexican standoff"
> results I've seen on larger maps -- both me and my
> opponent/s make bases that it is almost
> suicidal to attack. Missile towers 3 screens deep is
> one of the best detterants of a groundforce that I
> know of. However, I *LOVE* building a jammed
> guardian outside the missile tower's outer edge, and
> radar-target it on radar-blips. 1 good guardian hit
> is overkill on a MT.

The Mexican standoff is the very reason I hate long games, they're so damn boring.  I hate stationary stuff (though I like making nice bases :)), but if I have to use it, urg.  I much prefer micromanaging various Flash attacks, or even better, selecting 50 Panthers and telling the to wade through the enemy.

> "Only if you'd known that it was.  After all, if he
> kept D-Gunning your units then you were wasting metal,
> and maybe you'd have been more cautious?
> Well I don't know."
> If he's d-gunning my flash tanks, HE's wasting
> "free" metal that I'm giving him.

If I was him, I'd have wondered why you kept being so generous :)

> "I know exactly how good I am, distinctly average.
> I was probably a better a year ago because I'd play
> online a lot, now I rarely bother."
> Online games lag really bad, it's so disappointing.
> And my friends prefer "newer" games and are now
> bored with TA. The sad part is they're mostly
> older than I am. Reason why that's sad is their
> attention span for games is crippled by other
> obligations... work!

My work is crippled by my obligations to games.

> "I just find having a beautiful base with defences,
> walls and entrances it brilliant, if 100% impractical. :)"
> Play the Krogoth Encounter Mission on hard, if you
> beat it then I *KNOW* you're at least above average
> if not an expert. It will also FORCE you to
> make effective AA defenses and possibly killer
> ground defenses as well.

Playing that mission would take about half a day, I don't have the patience.

> "Stumpies...seem to be constantly fighting
> themselves, and this is also what can make a
> Bulldog useless in the hands of an AI."
> OTA Hammers are even worse.

Indeed, but they at least usually stand still and don't turn in circles, and when they do, they do it faster.

> "No, When it goes Gols it usually only goes Gols,
> where as ARM goes both Adv Kbots and Vehicles pretty
> close together.  I find Mavs a safer bet and
> reduce the chance of nano-stall."
> That's a major way MH and Queller differ -- Core's
> choice between Adv Kbots
> and Vehicles is exactly opposite.
>
> "Whoever the foe was, they must have been fairly
> passive (or dead and you were starting to write
> your name in DT's :))."
> Hardly! I've been in games where both sides are
> racking up 100+ kills every 10 minutes. And this
> is barely 30 minutes into the game! If both side's gets
> their bases built up early, productions render EVERY
> unit except a krogoth as just another mark on the
> killboard 5 seconds after you first see it.

I never get into that kind of game.

> "I can beat 2 on the small maps (Quellers), but
> nothing above 9x9.  Then I
> usually win 1 on 1's up till 18x18."
> I thought I could still beat 3 MH's on hard on
> Greenhaven. They beat me soundly, so I changed to 1
> Arm ai and 1 Core ai -- they beat me soundly. I
> changed to 1 Arm ai (the harder of the 2) and barely
> won.

I can handle 1 MH on Greenhaven, but I couldn't deal with 2.

> "On second though, Lord and Master has a nice ring to it <g>"
> Maybe Queller calls you that, but I think Newbie
> Crusher is far more appropriate. I will praise your
> ability after you beat a Gnug or 2 or some
> of the Swedish Yankspankers.

Heh, hell the whole TA community would have to call me that too :P

> "And that's why this boy plays mapped mode :)"
> Even many experts plain hate unmapped mode.

Yes, I play PD more than anything and still can't find jack :)

> "I play with the AI (and sometimes with EMg against
> it), I'm a man vs machine person at heart."
> Co-Op for me. Wanting to knock each other over the
> head afterwards (not that we don't anyway!) isn't my cup
> of tea. And 100% competitive generates that.

I just get too annoyed if my teammates mess up, or if I lag, or any other factor that I can't control (control freak), so I make sure I'm on the same team as every other human.  I don't mind losing to the machine (especially not if I built it :)).

> "I just can't handle too many things going on at once."
> I hope you don't have 6+ classes of school a day then.

It's a good time to catch up on my sleep.

> "Unfortunately, I tried to do it in a narrow valley.
> The Charge of The Lightning Brigade :)"
> Once Panthers start gumming themselves up with
> wreckage, just turn them around and flee - save
> them for another battle elsewhere.

Whoops :)

> "I've been meaning to work on that,"
> I'll look into fixing the factories in the next week.

OK, thanks.
--------
"I saw the game in a shop once I think. Always loved board games."
The Battletech novels and tech readout books are what makes the game fascinating to me. Playing it is dammed fun too, but it goes slow if everyone isn't interested. (Sort of like playing Monopoly 8-player when people aren't interested...)

"Yeah, but you won't suddenly have a big depression dip and enter a mid-life crisis."
I wish that were true, but I get depressed a lot.

"I've started to use Greenhaven now, probably because I like Green maps, and I can only take so many acres of flat featureless desert."
For me, Greenhaven is too crowded and Painted Desert is too sparce. I look for 3rd party maps that are in-between.

"And the day Queller attacks from about the 3 minute mark like any normal average human would, then I'll trust those tests :)"
I sometimes don't attack for 15 minutes, even if I'm being raided. The only people that attack at the 3 minute mark are Jeffy-rushers, because that's about all that CAN be made in 3 minutes -- unless the factory is built first and the commander assists-out the first attack unit, which will not happen at 1k/1k.

"I just demolish them as my large force wades through the trees.  I should be an Orc :)"
I've run MMM's almost entirely off reclaiming trees -- at least till my fusion/s get done.

"The Mexican standoff is the very reason I hate long games, they're so damn boring."
It also allows the most innovation for creative players. If your base is relatively invulnerable, you can waste 100+ units trying new strategies without losing the game. "Common" tactics fail in such standoffs, but really odd -- or even dumb things -- might save the day. Having gotten in so many standoffs, I've got a tactic of combined forces that allows me to roll back missile tower forests, unfortunately it's an exensive tactic -- costing me more than the resource value of all the MT's. However it's far more effective than the usual throw-your-units-at-the-MT tactics I've seen in recorded demos from even some of the "great" players.

"I hate stationary stuff (though I like making nice bases :)), but if I have to use it, urg."
I love Big Berthas, or even Guardians -- if they can be used as offensive weapons. :)
MT's are just an annoyance, both to use and to destroy -- but I'm VERY glad I have them!
DT'ed HLT's are what I do when I fear my base is going to be overrun.

"If I was him, I'd have wondered why you kept being so generous :)"
If you can kill a lot of enemy units (or even your own units) near your base, then you get a really nice metal income boost till all the wrecks are gone. A DT'ed HLT and a foolish opponent is a gift that keeps on giving. :)

"My work is crippled by my obligations to games."
What little work I do is on games, and sadly for free because it is illegal for me to profit off it.

"Playing that mission would take about half a day, I don't have the patience."
It took me a few days to a week to beat it, spread over about a month.

"Indeed, but they at least usually stand still and don't turn in circles, and when they do, they do it faster."
TA Bugfix I'm hoping decreases that problem. Now if only I can figure out why ai commanders don't use their lasers much...

"I never get into that kind of game."
Sounds like you either win, lose, or quit before then.
Quitch = Quit?

"I can handle 1 MH on Greenhaven, but I couldn't deal with 2."
I'm sure beating 3 or more is still possible if you raid them to death while building up, but not by me!

"Heh, hell the whole TA community would have to call me that too :P"
I've beaten a Gnug in a 3 way FFA game of Mephisto, Gnug825 and me on Red Triangle. He interrupted Meph's and my fight in the middle, (by attacking both of us with probing raids of samsons) so we both slowed down our attacks on each other (but didn't stop) and annihilated him. After a VERY long fight, I was able to beat Meph. My reputation is that sometimes I play really badly and sometimes extremely good and seldom in between. Part of that is caused by LAG, if the game's lagging really badly -- so is my ability. :p

"I just get too annoyed if my teammates mess up, or if I lag, or any other factor that I can't control (control freak), so I make sure I'm on the same team as every other human."
It could be a self-confidence issue, or a short attention-span problem.

"It's a good time to catch up on my sleep."
My teachers didn't let me sleep, except rarely -- and having to move from class to class every hour made it difficult. Plus, sleeping through lectures in classes like Calculus, Physics, Organic Chemistry, or College-level English would guarentee me being in deep shit come test-time.
--------
Subject: Re: TA, Battletech, School, and other bullshit...
Date: Wed, 1 Dec 1999 14:33:01 +0000 (GMT)
> "I saw the game in a shop once I think. Always loved
> board games."
> The Battletech novels and tech readout books are
> what makes the game fascinating to me. Playing it is
> dammed fun too, but it goes slow if everyone isn't
> interested. (Sort of like playing Monopoly 8-player
> when people aren't interested...)

It's hard to get people into the more in depth games, but I always found the more depth and complexity (to a point) were always the most satisfying to play.

> "Yeah, but you won't suddenly have a big depression
> dip and enter a mid-life crisis."
> I wish that were true, but I get depressed a lot.

I suffer depression sometimes, I swing from depressed to high, and rarely hit inbetweens unless I'm concentrating and then I feel nothing at all.

Though I try to be a happy guy :)

> "I've started to use Greenhaven now, probably
> because I like Green maps, and I can only take so
> many acres of flat featureless desert."
> For me, Greenhaven is too crowded and Painted Desert
> is too sparce. I look for 3rd party maps that are in-between.

I like Greenhaven, the number of hills means you can always try different things.  A couple of Flashes can bog down a big attack force just by getting in the way (well something a bit more metal heavy it probably better but you get the picture).

> "And the day Queller attacks from about the 3 minute
> mark like any normal average human would, then I'll
> trust those tests :)"
> I sometimes don't attack for 15 minutes, even if I'm
> being raided. The only people that attack at the 3
> minute mark are Jeffy-rushers, because that's
> about all that CAN be made in 3 minutes -- unless
> the factory is built first and the commander
> assists-out the first attack unit, which will not
> happen at 1k/1k.

I can launch a 5 Flash attack by the 3 min mark.  I try not to jeffy raid because you must remember, this is the guy who lost two jeffies, one after the other, to the same Stumpy! :)

> "I just demolish them as my large force wades
> through the trees.  I should be an Orc :)"
> I've run MMM's almost entirely off reclaiming trees
> -- at least till my fusion/s get done.

I ummmm, never get that far :)  I much prefer to stick to level 1, and when I do get level 2 I always forget to build construction vehicles.

> "The Mexican standoff is the very reason I hate long
> games, they're so damn boring."
> It also allows the most innovation for creative
> players. If your base is relatively invulnerable,
> you can waste 100+ units trying new strategies
> without losing the game. "Common" tactics fail in
> such standoffs, but really odd -- or even dumb things
> -- might save the day. Having gotten in so many
> standoffs, I've got a tactic of combined forces that
> allows me to roll back missile tower forests,
> unfortunately it's an exensive tactic -- costing me
> more than the resource value of all the MT's.
> However it's far more effective than the usual
> throw-your-units-at-the-MT tactics I've seen in
> recorded demos from even some of the "great"
> players.

The great players insist on reminding us how great they are :)  Oh, and they as ever have a small following of players who remind us as well <g>

> "I hate stationary stuff (though I like making nice
> bases :)), but if I have to use it, urg."
> I love Big Berthas, or even Guardians -- if they can
> be used as offensive weapons. :)
> MT's are just an annoyance, both to use and to
> destroy -- but I'm VERY glad I have them!
> DT'ed HLT's are what I do when I fear my base is
> going to be overrun.

I just regroup my mobile forces, and if I don't have any I kiss my butt goodbye.  Time for the Comms last stand.......  If you're lucky you just have enough time for a good one liner :)

> "If I was him, I'd have wondered why you kept being
> so generous :)"
> If you can kill a lot of enemy units (or even your
> own units) near your base, then you get a really nice
> metal income boost till all the wrecks are
> gone. A DT'ed HLT and a foolish opponent is a gift
> that keeps on giving. :)

Indeedy, you can't beat that kind of foe, can't beat as in the.....well I'm sure you understand

> "My work is crippled by my obligations to games."
> What little work I do is on games, and sadly for
> free because it is illegal for me to profit off it.
>
> "Playing that mission would take about half a day, I
> don't have the patience."
> It took me a few days to a week to beat it, spread
> over about a month.

Eek!  I can't imagine myself lasting that long.  My longest game was when I was new and speant about 3 hours on mission 3 of ARM, take the gate off the CORE. I hit the unit limit, and kept having to sacrifice 100 troops just to remov 5 turrets :)  It took FOREVER to make ground (still won eventually :))

> "Indeed, but they at least usually stand still and don't
> turn in circles, and when they do, they do it faster."
> TA Bugfix I'm hoping decreases that problem. Now if
> only I can figure out why ai commanders don't use their
> lasers much...

Try taking a look at the UH Fidos under AI control, as one of their weapons in fired by the D-Gun action, it may explain more about how it works.

> "I never get into that kind of game."
> Sounds like you either win, lose, or quit before
> then. Quitch = Quit?

Quitch means couche grass, every player walks all over me :)

I never quit, but when I used to play online I had a less than reliable modem.  If I ever quit (and I don't remember ever doing so), I would always concede the game to my opponent first.

> "I can handle 1 MH on Greenhaven, but I couldn't
> deal with 2."
> I'm sure beating 3 or more is still possible if you
> raid them to death while building up, but not by me!

Jeffies take forever to get accross that map, and sometimes don't make it at all and get stuck on a hill, you're better off with PeeWee's.

> "Heh, hell the whole TA community would have to call
> me that too :P"
> I've beaten a Gnug in a 3 way FFA game of Mephisto,
> Gnug825 and me on Red Triangle. He interrupted Meph's
> and my fight in the middle, (by attacking both of us
> with probing raids of samsons) so we both slowed down
> our attacks on each other (but didn't stop) and
> annihilated him. After a VERY long fight, I was able
> to beat Meph. My reputation is that sometimes I play
> really badly and sometimes extremely good and seldom
> in between. Part of that is caused by LAG, if the
> game's lagging really badly -- so is my ability. :p

God how I hate lag.....I hate hangs even more.

> "I just get too annoyed if my teammates mess up, or
> if I lag, or any other factor that I can't control
> (control freak), so I make sure I'm on the same
> team as every other human."
> It could be a self-confidence issue, or a short
> attention-span problem.

Eek, physcologist, get outta ma head boy :)

I'm sure I could make a good guess at the stem (and I have a non-existant concentration span if I lose interest :))

> "It's a good time to catch up on my sleep."
> My teachers didn't let me sleep, except rarely --
> and having to move from class to class every hour
> made it difficult. Plus, sleeping through lectures
> in classes like Calculus, Physics, Organic
> Chemistry, or College-level English would
> guarentee me being in deep shit come test-time.

At least you'd be well rested :)
--------
Subject: Re: TA, Battletech, School, and other bullshit...
Date: Wed, 1 Dec 1999 14:33:01 +0000 (GMT)

"I like Greenhaven, the number of hills means you can always try different things."
The bottom and top right start positions on Greenhaven have to go the farest to both geo spots and metal spots -- plus they are the most cluttered with trees and hills making starting there almost unfair.

"I can launch a 5 Flash attack by the 3 min mark.  I try not to jeffy raid"
Yes, but what I was getting at is a player that *IS* destroying enemy units in the enemy's base at the 3 minute mark. Even com build assisting flash tanks, I REALLY doubt you can have 5 Flash tanks killing your opponent's base in 3 minutes with a 1k/1k start.

"I ummmm, never get that far :)"
I'm surprised you don't try to build an early bertha to get the worm occassionally -- that can be done in less than 15 minutes.

"The great players insist on reminding us how great they are :)  Oh, and they as ever have a small following of players who remind us as well <g>"
The great players are great more often than not for being resource wizards. Being able to win most battles in the field is a rarity even among "experts"...

"Time for the Comms last stand. If you're lucky you just have enough time for a good one liner :)"
IF you hit ENTER before the "Quitch has been eliminated." message, you can send one last message even AFTER you're dead -- if watch mode is on.

"I can't imagine myself lasting that long.  My longest game was when I was new and speant about 3 hours on mission 3 of ARM, take the gate off the CORE. I hit the unit limit, and kept having to sacrifice 100 troops just to remov 5 turrets :)  It took FOREVER to make ground (still won eventually :))"
Either you were limited to semi-useful units, the ai was expanding like a plague, or your tactics stunk -- or all of the above?

"Try taking a look at the UH Fidos under AI control, as one of their weapons in fired by the D-Gun action, it may explain more about how it works."
Doesn't help -- if it's a script problem, I don't know how to fix it -- I barely understand scripts.

"I never quit, but when I used to play online I had a less than reliable modem."
I thought you quit when you got bored, sorry.

"God how I hate lag.....I hate hangs even more."
Lag-induced hangs are the worst of all. :p

"At least you'd be well rested :)"
You must not have strict parents who wanted you to get into a top university. Mine did not, but would get highly upset if I was failing classes in school.
--------
Subject: The wonder of a *Wibble*
Date: Thu, 2 Dec 1999 19:36:40 +0000 (GMT)

> "I like Greenhaven, the number of hills means you
> can always try different things."
> The bottom and top right start positions on
> Greenhaven have to go the farest to both geo spots
> and metal spots -- plus they are the most cluttered with
> trees and hills making starting there almost unfair.

Nothing like trying to convince someone that the top section of the Great Divide does NOT give you a giant advantage :)

> "I can launch a 5 Flash attack by the 3 min mark.  I
> try not to jeffy raid"
> Yes, but what I was getting at is a player that *IS*
> destroying enemy units in the enemy's base at the 3
> minute mark. Even com build assisting flash
> tanks, I REALLY doubt you can have 5 Flash tanks
> killing your opponent's base in 3 minutes with a
> 1k/1k start.

Not not in 3, only 4.  But on maps like Greenhaven a Jeffy takes 4 minutes to get their because it runs into everything.  It's too fast for it's own good sometimes.

> "I ummmm, never get that far :)"
> I'm surprised you don't try to build an early bertha
> to get the worm occassionally -- that can be done in
> less than 15 minutes.

But then I wouldn't be able to build hundreds of Flash tanks :( <g>

I'm from the "Flush" generation, I wasn't playing online when the whole missile thing started.

Slashers Vs Samsons, yawn!  Nothing like catching a Slusher player with his pants down, ambushing his force among some hills :)

> "The great players insist on reminding us how great
> they are :)  Oh, and they as ever have a small
> following of players who remind us as well <g>"
> The great players are great more often than not for
> being resource wizards. Being able to win most battles
> in the field is a rarity even among "experts"...

Hence why I lose a lot :)  I have difficulty handling the Comm alone if unmapped is on.

> "Time for the Comms last stand. If you're lucky you
> just have enough time for a good one liner :)"
> IF you hit ENTER before the "Quitch has been
> eliminated." message, you can send one last message
> even AFTER you're dead -- if watch mode is on.

If I'm losing I prepare my last line :)

It's damn dramatic for your partner simply to hear something shouted about the glory of ARM, then the screen shakes like mad and you drop.  Beautiful, exit stage left Quitch :)

> "I can't imagine myself lasting that long.  My
> longest game was when I was new and speant about
> 3 hours on mission 3 of ARM, take the gate off the
> CORE. I hit the unit limit, and kept having to
> sacrifice 100 troops just to remov 5 turrets :) It
> took FOREVER to make ground (still won eventually :))"
> Either you were limited to semi-useful units, the ai
> was expanding like a plague, or your tactics stunk
> -- or all of the above?

I used CC tactics, I snuck forward and fortified, setting up wonderful turret crossfire zones.  by the time I attacked the middle it was wall to wall with HLT's.

> "I never quit, but when I used to play online I had
> a less than reliable modem."
> I thought you quit when you got bored, sorry.

Only against the AI.  Online I like to play the "honourable" game, to the death and all that.

> "God how I hate lag.....I hate hangs even more."
> Lag-induced hangs are the worst of all. :p

Nothing like doing an all or nothing Comm assault on that heavy forward position, firing your D-Gun, only for the shot to lag out and you to die.  Damnit!

> "At least you'd be well rested :)"
> You must not have strict parents who wanted you to get
> into a top university. Mine did not, but would get
> highly upset if I was failing classes in school.

Oh I did, and I wanted to get here, but now I'm here....

I don't doze through classes though, even if they are boring.
--------
"Nothing like trying to convince someone that the top section of the Great Divide does NOT give you a giant advantage :)"
Ok, Great Divide and some other maps are worse...
Even poor Ashap Plateau has a contour bug on the lower left hill that makes getting down a bit more difficult if there's any debris.

"Not not in 3, only 4.  But on maps like Greenhaven a Jeffy takes 4 minutes to get their because it runs into everything.  It's too fast for it's own good sometimes."
Make a demo (with v0.96 of TA demo recorder, it's cool!) of it sometime versis Queller and email it to me. I gotta see your 4-minute flash attack strategy!

"I'm from the "Flush" generation, I wasn't playing online when the whole missile thing started."
I predate both generations barely. Combined arms was the word of the day then, and it's what I try to use. But sometimes I do get flash-happy.

"Slashers Vs Samsons, yawn!  Nothing like catching a Slusher player with his pants down, ambushing his force among some hills :)"
Core doesn't have much choice between going Samsons or Flashes -- since Instigators are considerably less effective than Flash tanks.

"Hence why I lose a lot :)  I have difficulty handling the Comm alone if unmapped is on."
As do I and most other players -- expert or no.

"If I'm losing I prepare my last line :)"
I only do if I'm pissed. I did manage to type "I'm not dead yet" LONG after I was wiped out in a game on Great Divide because Watching was on and I hit enter JUST as I died. :) My opponents were left wondering for a bit...

"by the time I attacked the middle it was wall to wall with HLT's."
This is why scouting occassionally is so valuable.

"Only against the AI.  Online I like to play the "honourable" game, to the death and all that."
I've quit because lag was too bad -- I couldn't even scroll the screen while the game was unpaused on occassion. Pretty shitty for a P2-400 w/128mb ram, I'd say.

"firing your D-Gun, only for the shot to lag out and you to die."
Most inglamourous lag-induced death I've had was on Painted Desert in a 3 vs 3 game when I spotted enemy dots on radar. So I sent all my mobile units out to "greet" them. Only to have my screen freeze briefly (~1/2 sec) just as I spot the 10+ flash tanks and have the flash tanks teleport 2 screens PAST my attack force, which is now FAR from my base. My undefended base was prompty destroyed.

"Oh I did, and I wanted to get here, but now I'm here...."
Where's here?

PS: What would you think of me taking ALL our messages and editing out lots of it and posting it as an Ai Tweaking Guide, part 2?
--------
Subject: Re: The wonder of a Quibble
Date: Fri, 3 Dec 1999 20:14:27 +0000 (GMT)

> Even poor Ashap Plateau has a contour bug on the
> lower left hill that makes getting down a bit more
> difficult if there's any debris.

And the fact you start further away from your metal than anyone else, have to go back to get it, then find you're further from the rest of the metal now.....

Hence why I say "Hey, we should have FIXED positions" <g>

> Make a demo (with v0.96 of TA demo recorder, it's
> cool!) of it sometime versis Queller and email it to
> me. I gotta see your 4-minute flash attack strategy!

I'll try sometime.  I don't know how fast Flashes make it on Greenhaven, I used to play PD allday, everyday. I Jeffy raid now (poorly) :)

> Combined arms was the word of the day

I just like watching Flea rushes :)

> Flash tanks.

Yes indeed, a rather 1 man army.

> As do I and most other players -- expert or no.

Yet I can handle 7 different Flash attack groups..... :)

> I only do if I'm pissed. I did manage to type "I'm
> not dead yet"...enter JUST as I died. :)
> My opponents were left wondering for a bit...

That's just evil, especially if they decide to send in more troops to find your non-existant spirit :)

> This is why scouting occassionally is so valuable.

Yeah but I was fresh from C&C :)  Scouting was something the losers wished you had to do :)

> I've quit because lag was too bad -- I couldn't even
> scroll the screen while the game was unpaused on
> occassion. Pretty shitty...I'd say.

I remember watching stunned as it brought my PC to it's knees with 50 planes in combat, eek.

> Most inglamourous lag-induced death

Ah that's always a winner.  I had one on Great Divide where my Flash tanks speant about 5 min shooting a radar post, by the time I got out of that it appeared he'd gone in with Flash tanks and blown my Commander to pieces.

> Where's here?

Swansea Unitversity

> PS: What would you think of me taking ALL our messages
> and editing out lots of it and posting it as an Ai
> Tweaking Guide, part 2?

As long as I got to see it and have messages by me cut if I didn't like them, which I doubt I would do at all.
--------
"And the fact you start further away from your metal than anyone else, have to go back to get it, then find you're further from the rest of the metal now..."
I didn't realize that...

"I'll try sometime.  I don't know how fast Flashes make it on Greenhaven, I used to play PD allday, everyday. I Jeffy raid now (poorly) :)"
Successful Jeffy rushing depends on the skill of your opponent/s. If they have built even a single Samson/Slasher, chances are your "rush" will be a Jeffy slaughter.

"Yet I can handle 7 different Flash attack groups..... :)"

I only look like I do, because my flashes converge on a single target from multiple directions.

"That's just evil, especially if they decide to send in more troops to find your non-existant spirit :)"
Because they didn't have allied victory on, their game didn't end with "Victory" so they did look for my "last remaining unit."

"Yeah but I was fresh from C&C :)  Scouting was something the losers wished you had to do :)"
C&C still requires keeping the enemy in check if you have boxes enabled.

"I remember watching stunned as it brought my PC to it's knees with 50 planes in combar, eek."
TA Bugfix online isn't as bad lag-wise as OTA, the improvement there is even greater than the improvements in speed in skirmish. HOWEVER, if your internet connection suck so too will your TA game...

"Ah that's always a winner.  I had one on Great Divide where my Flash tanks speant about 5 min shooting a radar post, by the time I got out of that it appeared he'd gone in with Flash tanks and blown my Commander to pieces."
I hope it wasn't a GW game!

> PS: What would you think of me taking ALL our messages
> and editing out lots of it and posting it as an Ai
> Tweaking Guide, part 2?

"As long as I got to see it and have messages by me cut if I didn't like them, which I doubt I would do at all."
I'm sure its lenth would be at MOST barely 1/5th the length of all our posts combined. I have repeated myself on accident too many times to count. :p
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: Re: Discussion Document
>Date: Mon, 24 Apr 2000 23:02:17 +0100 (BST)
>
>Going off at a tangent (well, moving onto a totally
>different road actually), after playing around with
>it, I think I can safely say that the TAK AI engine is
>a steap backwards from TA.
>
>Though it does nice things like multile groups, it
>never uses enough units (keeping far too many at home
>for defence).  Its biggest problem is though, that it
>cannot handle resources.  Yep, it only has to do mana,
>and it fails.  TA had safeguards that kicked in, but
>in TAK the AI will happily try to produce multiple
>factories and towers with only 1 or 2 lodestones, and
>it never really seems to move deliberately towards a
>sacred stone site like a Con unit would move towards a
>metal patch, it only builds on one if it runs across
>it, or on the odd occasion it actually goes for one on
>purpose.
>
>Having an AI where there seems to be no way to stop it
>mana-stalling kinda stops you from doing anything
>brilliant with the AI.
>
>Personally I don't understand why it is so poor.  The
>AI should have a massive advantage over the player in
>handling resources, it should know exactly how much it
>is spending and on what, and with a very simple
>Factories to Mana table could know exactly what it
>needed.......in advance.
>
>Sigh.
>
>Quitch(y)

Making a good artificial intelligence is more of an art than a science still. Either way, it is probably the most difficult form of programming there is. Most programmers take shortcuts, just because system limitations cannot devote enough "thinking processes" towards the ai.

I'll probably be an old man before I see something truely DESERVING of the name A.I.

David
--------
From: 	Quitch <quitchy@yahoo.co.uk>
Subject: 	Re: Discussion Document	
Date: 	Wed, 26 Apr 2000 20:20:05 +0100 (BST)	

But what I hate is that even basic mechanics are missing.  It isn't that hard to plug in some safeguards for the resources, or even to work out in advance what to do.  Sure, such a system would have holes in it because it wouldn't take into account why the stall was happening, or what would be built next, but it would be a step up.

I think it can be safely said that Cavedog considered AI a looooowwwwwww priority.

Marco Hack AI <-- Ever heard of it? Do you know where I can get it? 

Quitch(y)
----------
"By the way, have you taken a look at Queller TA & CC v13?  I know you normally pick it up and send an e-mail my way explaining in minute detail everything you think is a flaw :)"

I am of the opinion that although I've been very picky on your Queller ai's, I haven't declaired your ai's as massively flawed -- as you seem to imply about the Shonner ai (which in my opinion has earned the distinction of being the DEADLIEST long-game noncheating TA ai... at least for simple land maps.) At worst, Shonner ai's attack force is too generalized. That ai may make weird choices about attack unit composition for its attack force... but it sure does make a LARGE attack force quickly!

I only became aware of Queller TA & CC v13 about a day ago, but will probably use it (at least briefly) while testing TA Bugfix -- if only for a change of scenery. By swapping out the ai's some, I can spot more unit problems -- because my noncheating ai's tend to be VERY limited in the numbers and types of units it uses... to the point of being TOO specialized! Currently, I've been using my Counterstrike ai's for testing - because they pour out so many units so quickly. The samsons-not-firing bug can show up in under 10 minutes, with Cavedog's default ai it might take 30-45 minutes to notice it... (I was also enhancing Counterstrike ai as I edited TA Bugfix, thus creating an even deadlier monster!)
--------
Subject: 	Re: Fwd: Re: Various, probably involing lots of work for little reward :)	
Date: 	Thu, 13 Jul 2000 19:38:23 +0100 (BST)	

> "By the way, have you taken a look at Queller TA &
> CC v13?  I know you
> normally pick it up and send an e-mail my way
> explaining in minute detail
> everything you think is a flaw :)"
>
> I am of the opinion that although I've been very
> picky on your Queller ai's,
> I haven't declaired your ai's as massively flawed --
> as you seem to imply
> about the Shonner ai (which in my opinion has earned
> the distinction of
> being the DEADLIEST long-game noncheating TA ai...
> at least for simple land
> maps.) At worst, Shonner ai's attack force is too
> generalized. That ai may
> make weird choices about attack unit composition for
> its attack force... but
> it sure does make a LARGE attack force quickly!

I did note in my review that a few changes could lead to a big score change.  But it is massively flawed.  A split energy model?  Have you seen what happens when it goes on a popular map like CC?  Never mind the late game, it'd be lucky to reach the mid-game.  A late game AI isn't much use if it can't survive the early game.

Though currently it ranks as 3rd best non-cheating AI, IMHO.  With just a few changes (like no Wind, better early unit choices) he could make it twice as good as it is now.

> I only became aware of Queller TA & CC v13 about a
> day ago, but will

Where did you read/hear about it?  Or did you pop into AIC?

> probably use it (at least briefly) while testing TA
> Bugfix -- if only for a
> change of scenery. By swapping out the ai's some, I
> can spot more unit
> problems -- because my noncheating ai's tend to be
> VERY limited in the
> numbers and types of units it uses... to the point
> of being TOO specialized!

Quellers heading down the specialised route, though I doubt it'll go as far as MH.

> Currently, I've been using my Counterstrike ai's for
> testing - because they
> pour out so many units so quickly. The
> samsons-not-firing bug can show up in
> under 10 minutes, with Cavedog's default ai it might
> take 30-45 minutes to
> notice it... (I was also enhancing Counterstrike ai
> as I edited TA Bugfix,
> thus creating an even deadlier monster!)
>
> David
--------
>Shonner ai
"I did note in my review that a few changes could lead to a big score change.  But it is massively flawed.  A split energy model?  Have you seen what happens when it goes on a popular map like CC?  Never mind the late game, it'd be lucky to reach the mid-game.  A late game AI isn't much use if it can't survive the early game."
I only gave it consideration on Painted Desert, although ideally it's made for higher metal concentration maps than that. (Greenhaven should be good in theory...) True, it'd absolutely suck on low/no wind maps, but then I gave the author the benifit of the doubt and tried to test it where he said it was "best".

"Though currently it ranks as 3rd best non-cheating AI, IMHO.  With just a few changes (like no Wind, better early unit choices) he could make it twice as good as it is now."
I reran some tests on Shonner ai, and didn't get NEARLY the results that the author claims he got on Painted Desert. Queller ate it for breakfast, so did Mostly Harmless. The ONLY game that came close was when on 1million/1million resources -- then it actually beat MH, which is a low-metal ai. (Once out of a couple tries... MH decided that it had to have many solars/mex's before its first factory.)

"Where did you read/hear about it?  Or did you pop into AIC?"
I visit AI Central about 2-3 times a week. Spotted it there, in your email, and in a post at Anni's UH forum on the same day. I was busy though, so I didn't download it till the next.

"Quellers heading down the specialised route, though I doubt it'll go as far as MH."
If it gets too specialized, it could end up like Shonner ai on Comet Catcher -- at least on some maps with poor ai choices.
--------
From: 	Quitch <quitchy@yahoo.co.uk>
Subject: 	Re: Fwd: Re: Various, probably involing lots of work for little reward :)	
Date: 	Sat, 15 Jul 2000 17:06:06 +0100 (BST)	

> >Shonner ai
> "I did note in my review that a few changes could
> lead to a big score
> change.  But it is massively flawed.  A split energy
> model?  Have you seen
> what happens when it goes on a popular map like CC?
> Never mind the late
> game, it'd be lucky to reach the mid-game.  A late
> game AI isn't much use if
> it can't survive the early game."
> I only gave it consideration on Painted Desert,
> although ideally it's made
> for higher metal concentration maps than that.
> (Greenhaven should be good in
> theory...) True, it'd absolutely suck on low/no wind
> maps, but then I gave
> the author the benifit of the doubt and tried to
> test it where he said it was "best".

Actually my review human Vs AI test games were on Greenhaven, it just took too long to get going.

Actually, I'm moving away from Greenhaven now as a test map, I usually use CC now.  GH seems to cause some real unpredictability in AIs early games, MH and Q being no exception.  On CC Queller will virtually always build its first plant between 2 and 2.5 minutes, but on GH this value can range from 2-4, and sometimes it even takes 7-8 minutes.  I've noticed MH often has the same problem.  There's something about that map......I'm guessing lack of flat terrain and number of tress, possibly a limitation of the AIs building technique?

> "Though currently it ranks as 3rd best non-cheating
> AI, IMHO.  With just a
> few changes (like no Wind, better early unit
> choices) he could make it twice
> as good as it is now."
> I reran some tests on Shonner ai, and didn't get
> NEARLY the results that the
> author claims he got on Painted Desert. Queller ate
> it for breakfast, so did
> Mostly Harmless. The ONLY game that came close was
> when on 1million/1million
> resources -- then it actually beat MH, which is a
> low-metal ai. (Once out of
> a couple tries... MH decided that it had to have
> many solars/mex's before its first factory.)

Well on his site he'd changed it (last I saw) to say he hadn't tested it against Queller.  After I read his comments ("Somewhat like BSRAI"), I knew exactly what had happened, and contacted him to let him know that you HAD to use Quellers AI units, otherwise it won't use Vehicles, and Vehicles are the mainstay of Quellers early game.

> "Where did you read/hear about it?  Or did you pop
> into AIC?"
> I visit AI Central about 2-3 times a week. Spotted
> it there, in your email,
> and in a post at Anni's UH forum on the same day. I
> was busy though, so I
> didn't download it till the next.

Heh, I wanted to get it noticed :).  I thought after 4 month wait people might have given up on it.  I wish sites like Annihilated and TA Universe took more interest in AIs as news worthy.  The only site that reported it was TA World.  What's worse is that the news that went on instead were things about Jeremy Soule (ex TA music guy), Taylor (ex TA developer) logo contest by Generals Barracks (graphics).  Not one current TA story.  ARGHHHH! :)

> "Quellers heading down the specialised route, though
> I doubt it'll go as far as MH."
> If it gets too specialized, it could end up like
> Shonner ai on Comet Catcher
> -- at least on some maps with poor ai choices.

I think you've got it the wrong way round.  One of Shonners problems is it's too vague, trying to use Wind and Solar together, when it's usually a case of one or the other is better.

Anyway, Queller always has a bit of this and that, I wouldn't do it any other way :)

Think I might lower the Gator limit a bit again though.
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: switeck@hotmail.com
>Subject: Bugfix bug
>Date: Sun, 16 Jul 2000 15:00:32 +0100 (BST)

>Why have you given that mission on Hard the Krogoth
>profile?  Surely it should use Metal?

Krogoth ai is specifically made as a MISSION ai, and lacks all the cool TA:CC units (or at least can handle the lack...), METAL is made strictly for metal maps -- NOT for missions.
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: switeck@hotmail.com
>Subject: Questions
>Date: Tue, 18 Jul 2000 00:00:37 +0100 (BST)
>
>Shouldn't CORE Mission 2 use METAL as the Hard
>profile, rather than Default?
>
>I've noticed that in your missions you added support
>for ARMAIVP and other AI plants, but not for the much
>more common AAIVP, CAIVP and CAIAVP.  Or is ARMAIVP
>something else?
>
>Quitch(y)

CORE Mission 2 probably should use METAL...or mission/s.

ARMAIVP is the special factory in my counterstrike ai. The others are "more common", but only in BAI-based ai's.
--------
From: 	Quitch <quitchy@yahoo.co.uk>
Subject: 	Re: Bugfix bug	
Date: 	Sun, 16 Jul 2000 22:39:22 +0100 (BST)	

> >Why have you given that mission on Hard the Krogoth
> >profile?  Surely it should use Metal?
>
> Krogoth ai is specifically made as a MISSION ai, and
> lacks all the cool
> TA:CC units (or at least can handle the lack...),
> METAL is made strictly for metal maps -- NOT for missions.

Yes, but the same is true of the Default profile, but it IS used for Missions.  Metal is simply a Metal version of Default, and why should the final mission require anything different?  Krogoth is also a TA:CC profile, however it doesn't fit the map specs.  As you say in your own readme, you're assuming things like the presence of water, but there is no water on the final ARM mission, surely making metal the more logical choice?

> >Also, I had problems with the Bugfix campaign.  It
> >gave me an error involing "No suitable Schema" and it
> >wouldn't let me play mission 1.  In mission 3 several
> >Bulldogs and one of the Con units were stuck in trees
> >on Hard.
>
> Odd... do you have Queller ai loaded with special
> map OTA files as well.
> Because if that's the case, that's the problem -- TA
> Bugfix attempts to
> modify a map OTA, then Queller's map OTA's in the
> maps dir modifies it back!

Seems you're right.

By the way, talking of map OTAs, isn't Tropical Paridise a CC map?  Currently it is WaterWrld, but I think it would be much better off as a SeaBattle map, it has land, there is room to build air plants on it to in places, and there is quite a bit of land metal.
--------
"Yes, but the same is true of the Default profile, but it IS used for Missions.  Metal is simply a Metal version of Default, and why should the final mission require anything different?  Krogoth is also a TA:CC profile, however it doesn't fit the map specs.  As you say in your own readme, you're assuming things like the presence of water, but there is no water on the final ARM mission, surely making metal the more logical choice?"

Krogoth MAY be a water-included ai, but including water support isn't the same as REQUIRING it. If you've played and beaten the Krogoth Encounter, you'd know what I mean. What's more, I believe I used that profile in a CORE mission because KROGOTH ai profile originally controlled the CORE ai only on the Krogoth mission -- but in the CORE missions, it would control the ARM ai, which doesn't have to abide by the same set of restrictions as the Core ai.

Did you know that in missions you cannot FORCE the ai to build TA:CC units, even *IF* they're allowed in unit restrictions for the mission *AND* they're called for in the ai profile *AND* if the ai is given no other choice? The ai will instead do nothing! Tell me why that is the case and we'll both know. Thus, setting a mission to use a Default.txt ai which is made specifically for TA:CC would be like trying to use your Queller TA:CC ai on a TA-only PC.

"By the way, talking of map OTAs, isn't Tropical Paridise a CC map?  Currently it is WaterWrld, but I think it would be much better off as a SeaBattle map, it has land, there is room to build air plants on it to in places, and there is quite a bit of land metal."

Ok, I'll probably change that! I think Tropical Paridise is a DOWNLOAD map, so you COULD use it without TA:CC if it doesn't have any underwater metal spots.

David
--------
From: 	Quitch <quitchy@yahoo.co.uk>
Subject: 	Re: Bugfixing Bugfix	
Date: 	Wed, 19 Jul 2000 23:40:53 +0100 (BST)	

> "Yes, but the same is true of the Default profile,
> but it IS used for
> Missions.  Metal is simply a Metal version of
> Default, and why should the
> final mission require anything different?  Krogoth
> is also a TA:CC profile,
> however it doesn't fit the map specs.  As you say in
> your own readme, you're
> assuming things like the presence of water, but
> there is no water on the
> final ARM mission, surely making metal the more
> logical choice?"
>
> Krogoth MAY be a water-included ai, but including
> water support isn't the
> same as REQUIRING it. If you've played and beaten
> the Krogoth Encounter,
> you'd know what I mean. What's more, I believe I
> used that profile in a CORE
> mission because KROGOTH ai profile originally
> controlled the CORE ai only on
> the Krogoth mission -- but in the CORE missions, it
> would control the ARM
> ai, which doesn't have to abide by the same set of
> restrictions as the Core ai.

But the problem here is that most people when making a Krogoth profile will likely turn off one or two factories to support the sea effort.  If they don't, then their AI can support them, and why didn't they add two extra factories to their original Metal profile?  What benefits do Krogoth offer over Metal? I only see disadvantages.

> Did you know that in missions you cannot FORCE the
> ai to build TA:CC units,
> even *IF* they're allowed in unit restrictions for
> the mission *AND* they're
> called for in the ai profile *AND* if the ai is
> given no other choice? The
> ai will instead do nothing! Tell me why that is the
> case and we'll both
> know. Thus, setting a mission to use a Default.txt
> ai which is made
> specifically for TA:CC would be like trying to use
> your Queller TA:CC ai on a TA-only PC.

Which is why Queller says to use Queller TA for the original missions, all ships as one package :).  But as you may have noted at AI Central, Queller is not a mission AI right now :)

But which missions do you mean?  And do you mean it just won't build them, or the AI will only build them when it has a choice?

> "By the way, talking of map OTAs, isn't Tropical
> Paridise a CC map?
> Currently it is WaterWrld, but I think it would be
> much better off as a
> SeaBattle map, it has land, there is room to build
> air plants on it to in
> places, and there is quite a bit of land metal."
>
> Ok, I'll probably change that! I think Tropical
> Paridise is a DOWNLOAD map,
> so you COULD use it without TA:CC if it doesn't have
> any underwater metal spots.

Hmmm, I only suggested it because someone told me it was a CC map.  I guess the less of those Gray boxes you generate the better?  If you did TP, you should do Long Lakes too, currently it's Default......they never learned.

Quitch(y)
--------
From: 	Quitch <quitchy@yahoo.co.uk>
Subject: 	Re: Questions	
Date: 	Thu, 20 Jul 2000 00:03:00 +0100 (BST)	

> >Subject: Questions
> >Date: Tue, 18 Jul 2000 00:00:37 +0100 (BST)
> >
> >Shouldn't CORE Mission 2 use METAL as the Hard
> >profile, rather than Default?
> >
> >I've noticed that in your missions you added
> support
> >for ARMAIVP and other AI plants, but not for the
> much
> >more common AAIVP, CAIVP and CAIAVP.  Or is ARMAIVP
> >something else?
> >
> >Quitch(y)
>
> CORE Mission 2 probably should use METAL...or
> mission/s.
>
> ARMAIVP is the special factory in my counterstrike
> ai. The others are "more
> common", but only in BAI-based ai's.

Queller uses the AAI and CAI versions, but it isn't BAI based :)
--------
"By the way, on the first Bugfix mission, I was given 3 Vamps which I couldn't control, patrolling over the enemy Mohos.  Another test you forgot to undo?"
This is another reason TA BF v1.5 is still a beta -- I was wanting to do some serious testing of missions and never got around to it! What you are looking at is my testing method for missions: I flip sides and turn LOS to permanent and mapped. Sometimes I forget to switch it all back when done, especially when I get sidetracked -- as I have on the mobile AA units not firing at aircraft issue.

"But the problem here is that most people when making a Krogoth profile will likely turn off one or two factories to support the sea effort.  If they don't, then their AI can support them, and why didn't they add two extra factories to their original Metal profile?  What benefits do Krogoth offer over Metal? I only see disadvantages."

Krogoth ai gets hurt very little if it doesn't have ships enabled. (I almost feel it helps... ;) Metal ai on the other hand should NOT have ships enabled.

"Which is why Queller says to use Queller TA for the original missions, all ships as one package :).  But as you may have noted at AI Central, Queller is not a mission AI right now :)"

"But which missions do you mean?  And do you mean it just won't build them, or the AI will only build them when it has a choice?"

It doesn't matter WHICH mission it is, or WHAT you do to the ai profiles. If you wanted the ai to build PELICANS in a mission, EVEN IF YOUR AI GIVES IT NO CHOICE BUT TO BUILD PELICANS OUT OF THE ADV. LAB, the ai will still refuse because PELICANS are NOT an original TA unit. It doesn't matter that you've gone to great lengths to ensure that Pelicans are allowed in the unit restriction (which means the player can possibly build Pelicans on that mission just fine), the ai *WON'T BUILD THEM EVER*! There's only one way around this for the ai, in the mission's OTA file, you can give a factory direct orders on what to build -- but that will only work for that particular factory, and will not react to any changes in resources or circumstances. It'll even build Pelicans from that factory in that case even if Pelicans aren't on its build menu, the ai has Pelicans limited to 0, and even if it's a CORE ai!

[Tropical Paridise a CC map?]
"Hmmm, I only suggested it because someone told me it was a CC map.  I guess the less of those Gray boxes you generate the better?  If you did TP, you should do Long Lakes too, currently it's Default......they never learned."

I'll knock out both of them, and whatever other new maps Cavedog released as well.

> I have a hacked counterstrike ai in an attempt to
> localize the problem, but as of yet
> don't know what the exact unit is that is causing
> it. If I send you the ai
> profile, could you eliminate a unit or 2 and leave
> it running on Painted
> Desert overnight/day while you're not using your
> computer? Even better would
> be if you RECORD a demo of it using 2 computers. But
> that would probably be
> a huge demo, since the bug seldom occurs before the
> 40 minute mark. :( I
> *THINK* the bug is in one of the rarely-built core
> mobile unit's death
> script, but I have no guarentee of this!

"Sure thing."

I'll send you my ai. The problem seems to be in Core only, but there may be a SIMILAR problem with ARM as well! Test it as-is on Painted Desert with 1 CORE ai (+you as its ally) vs 1 Core ai to guarentee it crashes, then eliminate punishers and anything other than Doomsday Machines and test again. IF YOU CAN, record a demo during every test, (this of course will require 2 computers, which is something I cannot do!) and check out what it does during the last critical seconds by replaying the demo using ".pos 99" (run it at a slower speed if things are happening too fast.) The demo should crash at the same point as it did when you first ran the test. You might want to run the game at minimum resolution with +NOSHAKE, +SFX, no shadows/shading graphical options, and no lifebars on to speed it up a bit. (and at +10 speed of course ;) It might help if you use the Helper c-air that I've sent you to keep you alive in the game. (I ussually just build a helper and pick up my com with it and hide it inside my ally's base.)

David
--------
From: 	Quitch <quitchy@yahoo.co.uk>
Subject: 	Re: TA BF v1.5 beta disaster	
Date: 	Sat, 22 Jul 2000 16:14:30 +0100 (BST)	

--- David Sellick 
wrote: > "By
the way, on the first Bugfix mission, I was
> given 3 Vamps which I
> couldn't control, patrolling over the enemy Mohos.
> Another test you forgot
> to undo?"
> This is another reason TA BF v1.5 is still a beta --
> I was wanting to do
> some serious testing of missions and never got
> around to it! What you are
> looking at is my testing method for missions: I flip
> sides and turn LOS to
> permanent and mapped. Sometimes I forget to switch
> it all back when done,
> especially when I get sidetracked -- as I have on
> the mobile AA units not
> firing at aircraft issue.
>
> "But the problem here is that most people when
> making a Krogoth profile will
> likely turn off one or two factories to support the
> sea effort.  If they
> don't, then their AI can support them, and why
> didn't they add two extra
> factories to their original Metal profile?  What
> benefits do Krogoth offer
> over Metal? I only see disadvantages."
>
> Krogoth ai gets hurt very little if it doesn't have
> ships enabled. (I almost
> feel it helps... ;) Metal ai on the other hand
> should NOT have ships enabled.

I'm still not understanding your reasoning.  The one difference between a Krogoth and Metal map, is water......and possibly Krogoth numbers :)

What possible benefit do you gain from using the Krogoth rather than Metal profile on an all Metal and basically, standard Metal map?

> "Which is why Queller says to use Queller TA for the
> original missions, all
> ships as one package :).  But as you may have noted
> at AI Central, Queller is not a mission AI right now :)"
>
> "But which missions do you mean?  And do you mean it
> just won't build them,
> or the AI will only build them when it has a
> choice?"
>
> It doesn't matter WHICH mission it is, or WHAT you
> do to the ai profiles. If
> you wanted the ai to build PELICANS in a mission,
> EVEN IF YOUR AI GIVES IT
> NO CHOICE BUT TO BUILD PELICANS OUT OF THE ADV. LAB,
> the ai will still
> refuse because PELICANS are NOT an original TA unit.
> It doesn't matter that
> you've gone to great lengths to ensure that Pelicans
> are allowed in the unit
> restriction (which means the player can possibly
> build Pelicans on that
> mission just fine), the ai *WON'T BUILD THEM EVER*!
> There's only one way
> around this for the ai, in the mission's OTA file,
> you can give a factory
> direct orders on what to build -- but that will only
> work for that
> particular factory, and will not react to any
> changes in resources or
> circumstances. It'll even build Pelicans from that
> factory in that case even
> if Pelicans aren't on its build menu, the ai has
> Pelicans limited to 0, and even if it's a CORE ai!

So you're saying the AI will only ever build OTA units, even in the CC missions, unless scripted otherwise?  Wow.

> [Tropical Paridise a CC map?]
> "Hmmm, I only suggested it because someone told me
> it was a CC map.  I guess
> the less of those Gray boxes you generate the
> better?  If you did TP, you
> should do Long Lakes too, currently it's
> Default......they never learned."
>
> I'll knock out both of them, and whatever other new
> maps Cavedog released as well.

Those are the only two I changed IIRC.

> "Seesh, I was only kidding :).  Deary me <g>"
> "Because I use Quitch everywhere, except when I came
> back to TA I didn't
> know my CD Forum password.  I needed a name that
> showed I was Quitch, while
> being different from it.  Thus what people used a a
> nickname for my nick,
> Quitchy, got brackets to show that the Y was simply
> tacked on."
> "So now you know :)"
> You've mentioned something about how you got stuck
> with Quith(y) before. Or
> is it Quitch[y]? Or would you prefer I just called
> you Quitch.

In e-mails and forums I don't mind, but in documents I prefer Quitch(y) so that people know exactly who is being talked about, without it being vague or only a possibility.
--------
"I'm still not understanding your reasoning.  The one difference between a Krogoth and Metal map, is water......and possibly Krogoth numbers :)"
"What possible benefit do you gain from using the Krogoth rather than Metal profile on an all Metal and basically, standard Metal map?"
"So you're saying the AI will only ever build OTA units, even in the CC missions, unless scripted otherwise?  Wow."

Yes, in short ALL the ai will build in missions is OTA units, not TA:CC units as well -- unless scripted to do them. THIS is why the Krogoth ai profile is neccessary, because I don't use many OTA units in my METAL ai profile! The fact I use Krogoth ai profile on metal maps without water just means that when the ai tries to build shipyards, it'll give up and go on to something else...
--------
From: Quitch <quitchy@yahoo.co.uk>
Subject: Wind
Date: Wed, 26 Jul 2000 21:58:23 +0100 (BST)

I noticed that in your Mostly Harmless readme you had a list of maps, with some of them marked down as possibly being better with a wind orientated profile.

Since Queller comes with OTA files, I thought I may as well finish tracking down every map would would work better using the wind instead of the sun.

Since I thought building a Wind Generator on every map and watching it would be unreliable method of checking on a maps wind potential, I decided to finally work out what those Wind values stood for.  Have you ever done that?  Is it documented?  Anyway I found that basically every 500 Wind units equals 3 energy, the cut off point is 30 energy, however the game doesn't ignore Wind numbers which would generate more than 30 energy, it simply cuts the amount generated by them down to 30.

So I got this table:


Wind		Energy

0		0
500		3
1000		6
1500		9
2000		12
2500		15
3000		18
3500		21
4000		24
4500		27
5000		30

*energy roof*

5500		30
6000		30
etc.

Anyway enough with the theory.  Basically I went through your list of potential Wind maps and worked out on which ones the average amount of Wind generated would be greater than the amount generated by a Solar Panel, and here they are, in the same order and groups as your list:

Biggie Bigs
Cluster Freak
Oooooweeee
Slate Gordon

Lake Shore
Trout Farm

Assault on Suburbia
Ice Scream
John's Pass

The Barrier Reef
East Indeez
Sail Away
Pillopeens

Eastside Westside

* This list excludes maps already changed by Bugfix.

Quitch(y)
--------
I'll look into making the appropriate changes, HOWEVER one size does not fit all. Some of these maps will need airwind, others might use seawind, hoverwin, etc...

David
--------
This is what I meant of course, I just didn't think I'd have to spell it out :) 
--------
Yes, but... some of these ai profiles (seawind) don't exist yet. BTW, MOST airbattle maps are PRIME candidates for wind useage -- so much so that airwind would be the rule rather than the exception.

David
--------
>Oh and *pokes Switeck again* I still think you should
>include the Cavedog 2.1b1 AI, as it is a "Bugfix" to
>my mind.  You never said what you thought of that
>idea.
>
>Quitch(y)

I am not a 100% perfectionist, and it isn't EASY to edit 100's of lines of text trying to find the words to describe errors that most people will NEVER see.

If I included a modified ai -- or even the Cavedog 2.1b1 AI -- people might never bother to get Queller ai. :)

David
--------
> If I included a modified ai -- or even the Cavedog
> 2.1b1 AI -- people might
> never bother to get Queller ai. :)

2.0b1 may be an improvement over 3.1 (which isn't hard), but it isn't THAT big an improvement.  This is simply one of those changes that should be made because it is inside the Bugfix mandate.  Anyway, 2.0b1 only had a limited number of profiles, 3.1 would still control some areas.
--------
SUBJECT: Queller UH ai conflict

There is a weapons id # conflcit with Queller UH ai and UH v2.1b4 -- weapon #234 in UH is BFIDO, which is also used by Queller UH ai's AI Stunner EMP Silo.

Dunno when BSR changed the BFIDO weapon id #, or even if he did, but that's a problem now. :(

David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: Re: Queller UH ai conflict
>Date: Mon, 11 Sep 2000 00:12:38 +0100 (BST)
>
>We've discussed this one :)
>
>It's fixed in my v7 BETA, this error was the reason I
>switched from TACC to Immermans Data Viewer (TACC
>missed the conflict).
>
>Quitch(y)
>
>--- David Sellick <switeck@hotmail.com> wrote: > There
>is a weapons id # conflcit with Queller UH ai
> > and UH v2.1b4 -- weapon
> > #234 in UH is BFIDO, which is also used by Queller
> > UH ai's AI Stunner EMP
> > Silo.
> >
> > Dunno when BSR changed the BFIDO weapon id #, or
> > even if he did, but that's
> > a problem now. :(
> >
> > David

Any chance you're going to release it soon with a warning to NOT use Queller UH ai v6 and earlier with latest UH versions?

David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: switeck@hotmail.com
>Subject: Bugfix v1.5 BETA 6 - AI Metal Makers
>Date: Tue, 12 Sep 2000 22:50:52 +0100 (BST)
>
>With BF1.5B6, the Cavedog AI doesn't use Metal
>Makers.........and I can't figure out how you've done
>it.
>
>I checked the SIDE= line, it's fine.  I ran searches
>for internal weights and limits.  None.  I looked in
>the Sidedata file, everythings fine.
>
>So you've got me stumped, how did you stop the AI from
>building them?  I haven't seen it touch them.
>
>Quitch(y)

I sincerely doubt this is a never thing -- although I did make them far less likely. :)

David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: Re: Bugfix v1.5 BETA 6 - AI Metal Makers
>Date: Wed, 13 Sep 2000 19:06:35 +0100 (BST)
>
>I've run a LOT of games with the 3.1 AI in recently,
>and I ain't seen a single MM.  I didn't even notice
>until one of the AIs launched an attack force at the 8
>(!!!) minute mark.
>
>Golly!
>
>Quitch(y)
>
>--- David Sellick <switeck@hotmail.com> wrote: >
> >From: Quitch <quitchy@yahoo.co.uk>
> > >To: switeck@hotmail.com
> > >Subject: Bugfix v1.5 BETA 6 - AI Metal Makers
> > >Date: Tue, 12 Sep 2000 22:50:52 +0100 (BST)
> > >
> > >With BF1.5B6, the Cavedog AI doesn't use Metal
> > >Makers.........and I can't figure out how you've
> > done
> > >it.
> > >
> > >I checked the SIDE= line, it's fine.  I ran
> > searches
> > >for internal weights and limits.  None.  I looked
> > in
> > >the Sidedata file, everythings fine.
> > >
> > >So you've got me stumped, how did you stop the AI
> > from
> > >building them?  I haven't seen it touch them.
> > >
> > >Quitch(y)
> >
> > I sincerely doubt this is a never thing -- although
> > I did make them far less
> > likely. :)
> >
> > David

I changed the categories on MM's to NOT include metal. Also, Cavedog's metal ai profile probably doesn't include MM's in it. The change I did with the map ota files probably just means the ai is now using a metal ai profile on a metal map.

David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: Re: Queller UH ai conflict
>Date: Mon, 11 Sep 2000 00:12:38 +0100 (BST)
>
>We've discussed this one :)
>
>It's fixed in my v7 BETA, this error was the reason I
>switched from TACC to Immermans Data Viewer (TACC
>missed the conflict).
>
>Quitch(y)
>
>--- David Sellick <switeck@hotmail.com> wrote: > There
>is a weapons id # conflcit with Queller UH ai
> > and UH v2.1b4 -- weapon
> > #234 in UH is BFIDO, which is also used by Queller
> > UH ai's AI Stunner EMP
> > Silo.
> >
> > Dunno when BSR changed the BFIDO weapon id #, or
> > even if he did, but that's
> > a problem now. :(
> >
> > David

Where can I get Immerman's Data Viewer?

Also, another reason why I haven't included ai profiles is for Mac users' sake. THEY can only get ai's to work if packed into UFO files, which WON'T work if TA Bugfix's ai profiles are always overriding them. Plus, I don't want to work with a crappy ai -- I spent TOO long trying to make them GOOD, I don't WANT to make a mediocre ai to go with TA Bugfix. Likewise, I don't want TA Bugfix to be TOO hard for people to play against the ai in skirmish. This way, THEY can control the difficulty level.

If you don't mind, I'll put a URL link directly in TA Bugfix's readme to point to your website to get better TA ai's.

David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: switeck@hotmail.com
>Subject: TA Bugfix
>Date: Wed, 20 Sep 2000 19:49:10 +0100 (BST)
>
>For Long Lakes, why did you make it UrbanWin?  It has
>two giant lakes seperating layers in a 1 on 1, so the
>AI is going to do a Hundred Isles on the player now,
>and I can't see how this is an improvement over
>Default.  Surely HoverWin should have been used?
>
>Quitch(y)

Hover maps are water maps that have TOO much land to be seabattle maps. Hover maps ai's should play almost like the seabattle ai. Hover maps do NOT have underwater metal spots.

Urban maps are maps with too little metal spots to use default AND have some water with underwater metal spots! Urban should have a STRONG land battle slant to it, with water added in almost as an afterthought.

See the differance I'm shooting for now? Plus, since SO few have HoverWin and UrbanWin, you can almost do whatever is needed for those rare few maps.

David
--------
From: 	Quitch <quitchy@yahoo.co.uk>
Subject: 	Re: TA Bugfix	
Date: 	Thu, 21 Sep 2000 20:15:29 +0100 (BST)	

> >For Long Lakes, why did you make it UrbanWin?  It
> has
> >two giant lakes seperating layers in a 1 on 1, so
> the
> >AI is going to do a Hundred Isles on the player
> now,
> >and I can't see how this is an improvement over
> >Default.  Surely HoverWin should have been used?
> >
> >Quitch(y)
>
> Hover maps are water maps that have TOO much land to
> be seabattle maps.
> Hover maps ai's should play almost like the
> seabattle ai. Hover maps do NOT
> have underwater metal spots.

This sounds more like a Mostly Harmless convinience to me.  I think you'll find in 99% of AIs the Hover profile will handle this map, where as the Urban one will not.  I cannot think of a single CD Urban map where ships would be useful, generally because of the low metal conditions (the AI after all, doesn't reclaim very well), they're usually a waste of metal.

> Urban maps are maps with too little metal spots to
> use default AND have some
> water with underwater metal spots! Urban should have
> a STRONG land battle
> slant to it, with water added in almost as an
> afterthought.
>
> See the differance I'm shooting for now? Plus, since
> SO few have HoverWin
> and UrbanWin, you can almost do whatever is needed
> for those rare few maps.
>
> David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: switeck@hotmail.com
>Subject: Internal AI stuff
>Date: Sun, 24 Sep 2000 12:37:03 +0100 (BST)
>
>You said giving a unit internal weights and limits can
>screw around with the AI.
>
>I cannot remember, please remind me, does this apply
>to weights and limits of 0?
>
>Quitch(y)

No, it doesn't -- however if your ai profile doesn't have individual weights and limits on those units they won't be built. It's a handy way to prevent the ai from accidentally building such units. I probably need to add that to many such units in TA Bugfix even. Reason why I haven't is because earlier tests showed the ai was somehow *IGNORING* the 0-weights period. And 0-limits on units in FBI files never seemed to work at all. :(

> Hover maps are water maps that have TOO much land to
> be seabattle maps.
> Hover maps ai's should play almost like the
> seabattle ai. Hover maps do NOT
> have underwater metal spots.

"This sounds more like a Mostly Harmless convinience to me.  I think you'll find in 99% of AIs the Hover profile will handle this map, where as the Urban one will not.  I cannot think of a single CD Urban map where ships would be useful, generally because of the low metal conditions (the AI after all, doesn't reclaim very well), they're usually a waste of metal."

Yes, it *IS* a MH ai convience. But then again Cavedog's TA ai's tend to diversify FAR more than anything I make does (other than maybe Krogoth), so it doesn't really matter much if it makes the ai play a little worse (is that even possible?) when not using a modified ai.

BTW, did you ever figure out why the TA default v3.1 ai won't build metal makers with TA BF v1.5 installed?

David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: Re: Internal AI stuff
>Date: Sun, 24 Sep 2000 20:12:47 +0100 (BST)
>
>Thanks Swi.
>
>I haven't used the MM thing yet.  I've gone through
>every way I know of stopping the AI from building
>something, but nothing.
>
>Quitch(y)

I know multiple ways to stop the ai from building something -- but some can be overridden by other things. Like missions can even build units out of the 'wrong' factories! I've even seen a Core ai building Shooters in a mission. I THINK it was out of a Core factory too.

I don't have much trouble preventing the ai from building something, rather it's the reverse that causes so much trouble -- I can't make it build needed resources (like Geos...) when they are needed. :(

David
--------
From: 	Quitch <quitchy@yahoo.co.uk>
Subject: 	Re: Internal AI stuff	
Date: 	Wed, 27 Sep 2000 23:36:32 +0100 (BST)	

> >Subject: Re: Internal AI stuff
> >Date: Sun, 24 Sep 2000 20:12:47 +0100 (BST)

Let me rephrase.  I haven't figured out yet why the Metal Maker is not being built yet, young sir :).

> On the other part, I know multiple ways to stop the
> ai from building
> something -- but some can be overridden by other
> things. Like missions can
> even build units out of the 'wrong' factories! I've
> even seen a Core ai
> building Shooters in a mission. I THINK it was out
> of a Core factory too.
>
> I don't have much trouble preventing the ai from
> building something, rather
> it's the reverse that causes so much trouble -- I
> can't make it build needed
> resources (like Geos...) when they are needed. :(

No what I meant is that I have checked every way I know of that stops a unit being built by the AI (e.g. internal weight and limit), but none of those methods seem to have been employed, yet still the MM is not built.

Still trying to figure it out....
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: AI LOS
>Date: Mon, 23 Oct 2000 19:17:03 +0100 (BST)
>
>This is part of an e-mail I sent to BSR:
>
>"Seems you were right about the AI needing LOS for
>things like Silos, I've had to rethink how the AI
>spots.  Seems it actually uses LOS just like everyone
>else, its ground troops will only fire at units that
>come into their LOS, but they fire at their main
>target at max range because they're given the attack
>command.
>
>I'm seeing what I can do, this really screws the AI in
>terms of LRPCs and Silos.....if only the LOS table
>wasn't so clumsy, then at least a 320000 range LOS
>could be generated and used on all long range AI
>stuff."
>
>This is a real pain in the ass.  I've only tested it
>using my AI Anti-Nukes, so it may not hold true for
>other units, but I think it probably will.
>
>Quitch(y) 8:)

I've never had any problems with the ai not being able to fire its nuke silos due to lack of LOS. What's more, it seems reasonably intelligent about its targets too.

David
--------
From: 	Quitch <quitchy@yahoo.co.uk>	
Subject: 	Re: AI LOS	
Date: 	Tue, 24 Oct 2000 19:52:22 +0100 (BST)	

Start a game unmapped with true LOS.  Build one of my insta-build Silos, and see if the AI attacks it with the insta-build AI Anti Nuke silos.

Try a second game starting with perm LOS mapped, then type in +LOS.  I've got a feeling the AI uses the original settings.

The file attached is bloatware by the way, it only needs two of the units in the, but I couldn't be bothered to wipe them, it'd be quicker for me just to attach it as is :)

I'm a thoughtful person :)

I'm typing this line to pass the time, bloody Yahoo, bloody slow.

Still waiting......

No green on the connect line, yay BT Internet!  You suck.

I hate modems.....

So now we've established I hate everything :)

Ho hum....

Wait, I've got a second browser window, I'll go play in that :)

Now the file failed.  God I hate everything!  Trying again.......

Good God it worked.........finally

Quitch(y) 8:)

> >Subject: AI LOS
> >Date: Mon, 23 Oct 2000 19:17:03 +0100 (BST)
> >
> >This is part of an e-mail I sent to BSR:
> >
> >"Seems you were right about the AI needing LOS for
> >things like Silos, I've had to rethink how the AI
> >spots.  Seems it actually uses LOS just like
> everyone
> >else, its ground troops will only fire at units
> that
> >come into their LOS, but they fire at their main
> >target at max range because they're given the
> attack
> >command.
> >
> >I'm seeing what I can do, this really screws the AI
> in
> >terms of LRPCs and Silos.....if only the LOS table
> >wasn't so clumsy, then at least a 320000 range LOS
> >could be generated and used on all long range AI
> >stuff."
> >
> >This is a real pain in the ass.  I've only tested
> it
> >using my AI Anti-Nukes, so it may not hold true for
> >other units, but I think it probably will.
> >
> >Quitch(y) 8:)
>
> I've never had any problems with the ai not being
> able to fire its nuke
> silos due to lack of LOS. What's more, it seems
> reasonably intelligent about
> its targets too.
>
> David

Attachment: ^^QuellerUH7(5).rar (682k)
--------
From: 	Quitch <quitchy@yahoo.co.uk>
Subject: 	AI Anti-Nukes
Date: 	Sat, 25 Nov 2000 13:36:32 +0000 (GMT)

I suddenly realised that I don't think I've asked you to add a tag to Bugfix units, that this unit would require if it were to work.

Basically, every unit, except those that the Anti-Nuke would normally affect (i.e. everything which has a weapon that Anti-Nukes can shoot down) should be given a tag, something like NOTSILO.

I would appreciate it if you could add this tag so that this (close as we're gonna get) AI Anti-Nuke idea can work.

Quitch(y)
--------
I don't have any intentions to add a category tag to almost EVERY unit in the game just to help 1 unit target a little better.
That's why the Arm ai EMP silo and Core ai Neutron silo sucks too.
However, if you can think of no more than 2 categories I can use to fix the ai antinukes AND those silos, I might consider doing it. :)

David
--------
Subject: 	Re: AI Anti-Nukes	
Date: 	Mon, 27 Nov 2000 23:32:30 +0000 (GMT)	

Target a little better?  Uh, NO!  They're needed to make a concept work that's as close to what people have been demanding since the dawn of TA time, as we're gonna get.  They're needed for the things to work, period.  How often do you think a human places a Nuke at the front of his base?

Why two categories?  What's so different about copy and pasting 3?  You'd still be adding just as many labels, to just as many units, except you'd have to press CTRL+C one more time........oh no! :)

Quitch(y)
--------
Why only 2 or so? Well because... these categories would be the "NOT" for every unit in the game except nuke silos. And that *STILL* won't be enough to make the ai antinukes to target worth a darn. Same goes with ai Arm EMP silos and ai Core Neutron silos.

I went to a lot of effort to remove a lot of seemingly worthless categories. You're asking me to add a category that only has relavance if a hacked ai is installed.

Sorry, I want a better reason -- because I know the ai's targetting will still suck really badly.

David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: Re: AI Anti-Nukes
>Date: Tue, 28 Nov 2000 20:03:15 +0000 (GMT)
>
>In my experience AI units like Nukes, with the right
>tags in place, target perfectly.
>
>Quitch(y)

I've run many tests on just that very subject. If anything, the ai nuke silos seem to prefer lone missile towers as targets over other targets. HOWEVER, that doesn't stop it from taking occassional shots at metal extractors or attack units. I would expect the ai antinuke silos to be no different. It IS random, the tags just up the probability some...
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: Re: AI Anti-Nukes
>Date: Wed, 29 Nov 2000 00:22:01 +0000 (GMT)
>
>Try running a test with a Comm and a Silo.  Tell Silos
>not to shoot Comms.  The Silos will shoot the other
>Silos every time.
>
>Quitch(y)

That's only a 50-50 test... but what if there's only 1 silo and 499 other units? I've seen ai nuke silos shoot almost everything under the sun with little rhyme or reason, and often contrary to what BOTH its bad targetting lines tell it to do. And if it won't obey either when legitimate targets are plentiful, I don't see how the ai antinukes will choose legitimate targets much at all!

I don't want to do what you're asking because it'll slow the game down (admittedly, very slightly.) I don't see it as being an adequate solution -- because the ai's too stupid to smartly choose targets. (Especially considering that the ai may have to get LOS to fire!!!) And I also know how annoying it is to manually cut-and-paste category line changes into all the unit fbi files. It's easy for be to screw up badly somewhere. (I'm still sorting out the occassional unit script mistake from changes I made over a year ago and thought I'd double checked.)

You want to run a check on this yourself, go ahead. But please don't demand that I do.

David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: Re: AI Anti-Nukes
>Date: Thu, 30 Nov 2000 00:13:22 +0000 (GMT)
>
>Well we'll just have to see how it goes in UberHack.
>
>Quitch(y)

Thanks.

If it goes well, I will add it to TA Bugfix. However, keep in mind that it would only be relevant if a hacked ai is installed which does utilize ai antinukes of this particular kind. Wouldn't it make more sense to make an ai mutator that changes these units for you? This is something that has to change a lot more game parameters than even BAI did -- which I objected to at the time of the release on the grounds that I considered it excessive. BUT... this change will not result in giving the ai an ability far beyond that of the player/s -- so you have more support from me on this issue than you realize. My fear is that the changes required to DO this is like those required for 3do weapons -- it sounds good, it looks good -- but it either doesn't work, or just results in the game being slowed down. And actually doing it without error is best left to someone who's not so burned out on making changes in TA than I am...

David
--------
Back on the ai antinuke subject -- how about make the ai antinuke a wee bit more powerful and allow it to stun all strategic weapons, such as nuke silos and berthas? Possibly even guardians... The more types of units it can affect, the greater the probability that it chooses one it can affect. Give it a large AoE and a reasonable firerate and cheap cost -- but only about a 1 min stuntime, so it must fire repeatedly. This way if the ai makes a BAD choice about targets, it'd still be likely to hit something it can affect.

David
--------
From: 	Quitch <quitchy@yahoo.co.uk>
Subject: 	Re: AI Anti-Nukes
Date: 	Tue, 5 Dec 2000 19:44:03 +0000 (GMT)

Maybe in a cheating AI, but for non-cheating if the idea doesn't work on the correct targets then I'll just abandon the project.

Quitch(y)
--------
Did the ai antinukes work?

David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: Re: AI Anti-Nukes
>Date: Sat, 30 Dec 2000 14:24:30 +0000 (GMT)
>
>BSR isn't adding the tags until the final UH release.
>
>Quitch(y)

Why wait that long? :)

Do some minor testing now to learn if it's worth your trouble -- or I will.

David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: Re: AI Anti-Nukes
>Date: Sat, 30 Dec 2000 20:58:49 +0000 (GMT)
>
>I have done minor testing with just a couple of units.
>  I told you, the Comm and the Anti-Nuke, and it always
>went for the anti-nuke.
>
>I don't feel like doing more than that.
>
>Quitch(y)

Even if the ai antinukes do work as specified, they will not stop the production of nuke missiles by the nuke silos... they will just stop them from firing for however long their paralyzer lasts. :(

David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: TA: Dark Suns AI
>Date: Tue, 1 May 2001 21:50:13 +0100 (BST)
>
>You may be interested in this.  I don't know if you'd
>enjoy the TC or not, but it has a quirk I've noticed,
>and I don't know why.  You can find the TC at
>http://www.unituniverse.com/tads/
>
>As I understand it, it has always been accepted belief
>that the AI will kick off with something that produces
>metal.  However, in this MOD (of which I have the TAM,
>no custom EXE version), the IAF AI at least, can place
>plants first.
>
>These plants don't seem to have any reference to metal
>that I can see, and I have, so far, been unable to
>figure out what makes them so different from standard
>TA plants that the AI can use them right away.
>
>Thought this might interest you.
>
>Quitch(y)

I'll look into it, but I cannot promise any news on it for a couple days till I figure out why.

I just hope someone can solve my GAF problem with TA Bugfix... :(

David
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: David Sellick <switeck@hotmail.com>
>Subject: Re: TA: Dark Suns AI
>Date: Wed, 2 May 2001 13:22:32 +0100 (BST)
>
>What's the GAF problem?  Why not put it on the
>UberHack forum?  That's where people who play with
>that sort of thing tend to be.
>
>Quitch(y)
--------
The Arm FARK GAF is messed-up, it is MUCH larger than 64x64 pixels!
Other build menu GAFs have similar problems in OTA. I do NOT know if it causes stability issues.

TA Dark Suns' ai works only for "ARM" not for "CORE".
The "Arm Commander" makes +5 metal/sec and +75 energy/sec, while the "Core Commander" only makes +1 metal/sec and +25 energy/sec.

So, the Arm ai starts building because the metal and energy failsafes turn off because it's over the threshold values on both. But the Core ai wants to build mexs -- and CAN'T, so it just sits there.

And the reason why it doesn't build mexs?

Mexs in TADS can only be built underwater...

Don't ask me WHY this is the case, it may in fact be accidental -- Brother Alpha may have wanted to code in mexs as having a maximum water depth of 1, but instead he put in a MIN water depth of 1!

David
--------
From: 	Quitch <quitchy@yahoo.co.uk>
CC: 	Dawn Falcon <dawnfalcon@upliftwar.com>, Wizzball <wizzball@gmx.net>
Subject: 	Re: TA: Dark Suns AI	
Date: 	Fri, 4 May 2001 14:18:00 +0100 (BST)	

--- David Sellick wrote: 
> The Arm FARK GAF is messed-up, it is MUCH larger
> than 64x64 pixels!
> Other build menu GAFs have similar problems in OTA.
> I do NOT know if it
> causes stability issues.

Can you just recreate the image but in the proper size?

>
> TA Dark Suns' ai works only for "ARM" not for "CORE".
> The "Arm Commander" makes +5 metal/sec and +75
> energy/sec, while the "Core
> Commander" only makes +1 metal/sec and +25
> energy/sec.
>
> So, the Arm ai starts building because the metal and
> energy failsafes turn
> off because it's over the threshold values on both.
> But the Core ai wants to
> build mexs -- and CAN'T, so it just sits there.

I'll try changing the TA'AK Comms production to those values and see if it works.  Interesting idea.  The TA'AK doesn't sit there though (at least not in my experience), but always makes a MEX first.

> And the reason why it doesn't build mexs?

I haven't had this problem?  Or have you?  Or did you just misunderstand my original e-mail?  I was saying that the IAF Comm can start with anything, where as the TA'AK Comm always plays the standard AI game of MEX first.

> Mexs in TADS can only be built underwater...
>
> Don't ask me WHY this is the case, it may in fact be
> accidental -- Brother
> Alpha may have wanted to code in mexs as having a
> maximum water depth of 1,
> but instead he put in a MIN water depth of 1!
>
> David

This is purposeful.  On space maps all asteroids are actually patches of water.  They did this because it stops the AI slapping MEXs all across the map on empty patches.  Works suprisingly well.

Quitch(y)
--------
From: 	Quitch <quitchy@yahoo.co.uk>
Subject: 	Just thought you should know...
Date: 	Sun, 6 May 2001 02:44:24 +0100 (BST)

Your idea about increasing Comm resource production worked out great.

Now, to figure out the minimum values..... :)

Quitch(y)
--------
>From: Quitch <quitchy@yahoo.co.uk>
>To: Switeck <switeck@hotmail.com>
>Subject: AI Debate
>Date: Mon, 27 Aug 2001 21:22:56 +0100 (BST)
>
>There's an AI debate you might be interested in, about
>what constitutes a cheating AI, over at the
>Thunderball forum on TAU, subject "Tried Thunderball
>last night".
>
>Quitch

Oh great... not THAT subject again. :)

Sure, I'll look into it -- and the more hotly it is debated, the more I'll try to throw facts at it.

David
--------
These ai 10-in-1 bombers are either flying Krogoths, thus guarenteeing that they take out whatever target they select, or they are a total joke and a waste of the ai's resources.

And then there's the 3rd possibility: BOTH!

The ai bombers will, more often than not, select whatever target is nearest to them, irregardless of bad targetting tags, and bomb the heck out of it. And if it's a MT or mex on the edge of a player's base, that's going to be a major waste because now it'll be trying to turn over enough AA to kill a Krogoth! Even if they don't take special damage from AA weapons, they will still die if it's a cheating ai.

Cheaper ai bombers that aren't 10-reg-bombers-rolled-into-one WILL split up SOME (but not much) to take out multiple targets. It is actually more efficient that way, not less, because of the ai's relatively poor targetting choices.

And Quitch(y), even a single Arm Thunder can take out an open solar in 1 pass. On the other hand, I've seen an extreme-vet Pheonix (it had about 1k kills -- I was testing something on the ai and bombing it when it built the 'wrong' things) take out a full-health Arm Fusion on a single pass. This was NOT a hacked bomber either...

In short I repeat once again that the ai not only does not NEED 10-in-1 ai bombers that it is a hinderance for it to HAVE to use them.

Of course if you're making an ai with +1k m/sec production for almost nothing, cost-per-bomber doesn't matter -- only how long it takes to build/replace. But it's hardly sporting, old chap.