Why, oh why does the poor AI...

[player1]

Until it's fixed by Soren in the next patch , this is another reason to keep the 8-10-1 infantry boost in the PTW version of the AU mod. The guerilla would become 7-6-1 or something.

You mean like in PtW version of Patch suggestion MOD?
(7-6-1/cost 80)

[Nor Me]

Giving facrories the abilities of barracks also seems to work and would have less of an impact. Even if you already have barracks in a city, the governor still suggests building a factory next.

If the guerrilla is considered to be as good an offensive unit as the infantry, why does the AI build Mech. Infantry over tanks. I've tested identical units with difierent tech requirements and it builds them 50-50.
Maybe it doesn't like using the rubber?

[Robber Baron]

Originally posted by Nor Me
Maybe it doesn't like using the rubber?

Well, to be fair, I don't think anybody likes using the rubber. :P (Sorry, couldn't resist.)

[alexman]

Originally posted by Nor Me
Giving facrories the abilities of barracks also seems to work and would have less of an impact. Even if you already have barracks in a city, the governor still suggests building a factory next.

Another good idea, but then barracks become obsolete with factories. Who wants to pay that extra maintenance if you have a factory that does the same thing?

Originally posted by Nor Me
If the guerrilla is considered to be as good an offensive unit as the infantry, why does the AI build Mech. Infantry over tanks.

The AI builds Mech. Infantry for defense. It builds tanks for offense, since they have a better attack factor and are cheaper. You see some AI offensive Mech. Infantry in the field, but they are all from upgrading Infantry created for attack, and keeping the offense flag.

[player1]

On the other hand, if you deselect Offesnse AI flag from Mech. Inf. all upgarded Infatries will become Defensive Mech. Infantry.

Which would lead to higher amount of AI tanks.
Since we would think that he needs more Offense units.

[alexman]

After a few more tests with PTW, I have not yet found a case where the AI does not build a unit that comes later in the list of units that it can build for that purpose.

Examples for an offensive unit:
(With both units in the build queue)

• Always builds Chariots instead of Swordsmen
• Always builds Horsemen instead of Swordsmen
• and of course... always builds Guerillas instead of Infantry

I even added a new unit, exact copy of the Warrior but without an upgrade, to the end of the list and called it DummyWarrior. The AI always build this crappy unit even though it can build Med. Infantry...

[Edit: I'm not so sure about this any more. It seems this happens only when you have offensive units as a build-often and defensive units as a build-never.]

[Soren Johnson]

now _this_ is a great thread. thanks for all the work, everybody... even though a lot of these fixes are only really going to affect the AI's performance for mods. I can't promise that everything here will be fixed, but I will do my best.

[player1]

Thanks Soren!

[player1]

By the way is there any chance to make unit/building type priorities, more open for editing?

For example, you can say "when AI get Infantry, it NEVER builds Guerrila", or always build Marketplace before Library.

It's huge job for one person, and if it could be open for editing it could help a lot.

And it would make moding much more effecive.

I hate to think that by changing stats of one unit could ruin AI performance since MODers don't know inner build queue of AI (how and when would AI build particular unit).

Or at least some testing utility which would answer MODers question about which unit AI likes and which don't (partialy already done with debug mode).

Or calculated AI's value of unit (based on stats), and which units he would build alongside that one (since they are as good), and which unit would it ignore (since this unit is much better)?

[ducki]

Originally posted by player1
By the way is there any chance to make unit/building type priorities, more open for editing?

For example, you can say "when AI get Infantry, it NEVER builds Guerrila", or always build Marketplace before Library.

NEVER and ALWAYS are extremely radical decisions for specific units.
For example, if you tell the AI ALWAYS build MP before Lib - what if the AI has only 1 lux, or no lux, or is in a game for Cultural Victory? Seems like ALWAYS can often hamstring the AI when applied to specific units/improvements.

Of course, there may not be any other way with the current program structure, but I'd prefer to see the AI doing a cost-benefit analysis, so that IF and ONLY IF UnitA is more beneficial per shield used than UnitB, then build UnitA, otherwise, Default to UnitB.

From some of the reading here, though, Units and Buildings seem to be purely hierarchically arranged, such that A is always build before B if possible - instead of if preferable.

Then again, computers are woefully inept at "preferable" and other fuzzy logic decisions. We humans will lock an unoptimally placed town at 6 population in order to save tiles for the metropolis next door, while I don't think the AI has any idea what to do about this. I think the AI ALWAYS tries to have metropolises.

Anyway, point is, "always" and "never" are very tough to balance when dealing with "AI" when put into specific terms. Which is probably why units and building have flags like Offense, Defense, Culture, Wealth, etc.

What the AI needs is a better sense of "better", IMO.

I hate to think that by changing stats of one unit could ruin AI performance since MODers don't know inner build queue of AI (how and when would AI build particular unit).

I agree here, but changing a unit stats is probably less dangerous than telling the AI to "never" or "always" build it.

And I definitely agree with more transparency about "unit value" to the AI. More docs from Firaxis, like you suggested.

[Chemical Ollie]

Re: "Simulated Intelligence"

Originally posted by Zachriel
The "AI" is actually an "SI," that is, Simulated Intelligence...

This is one of the compromises that makes Civ possible on a "primitive" 20th century computing machine...

So true! Think of chess - a game with much simpler rules than CIV3. You need a NASA-grade super computer (Deep Blue) to build an AI that can beat a good human player.

I always lose against the AI in chess...

On the other hand, I occasionally beat the Emperor level AI in CIV3... Perhaps because I like CIV better...

[cgannon64]

Originally posted by player1
And here is the reply:

>>>>>>>>>>
That is intentional. I'm not sure why, but I know it is listed that way in
the design document.

Barry Caudill
QA Manager
Firaxis Games
>>>>>>>>>>

That is insane. Do they want to cripple Spain permanently? Unless they fix that, I doubt I'll EVER see Spain win a game.

CG

[alexman]

Originally posted by Soren Johnson Firaxis
now _this_ is a great thread. thanks for all the work, everybody... even though a lot of these fixes are only really going to affect the AI's performance for mods. I can't promise that everything here will be fixed, but I will do my best.

Soren! I was hoping you would read this thread!
Thanks for letting us know you did!

Weird things happen when you have offensive units as a build-often and defensive units as a build-never. The AI never builds any units. If you have everything else flagged as build-never as well, the AI always builds the last unit in the editor list in this case. (Provided it has the necessary tech and resources). This is not a big deal at all, it just confused me like hell when I was testing...

[Konquest02]

Originally posted by Zachriel
The "AI" is actually an "SI," that is, Simulated Intelligence...

This is one of the compromises that makes Civ possible on a "primitive" 20th century computing machine...

I admit the AI has some bugs, but I could be worse.

The following screenshot is from Transport Tycoon Deluxe, set on hardest level. The goal is to build a successful transport company. The AI often can't get a single line running efficiently. And yes! it's a Sid Meyer game!

--Kon--

[Theseus]

I'm glad all of this is getting dealt with...

On a side note, let me just say that I have been so expressive to my fiance about my appreciation for Firaxis' ongoing involvement, that I actually overheard her (an interior designer) on the phone with her sister (an actress), explaining that *some* companies ("you know, that Civilization game he's always talking about") really try to pay attention to their customers, and it makes a difference!!

Soren, you is da man, and anything that helps development that keeps bubbling up outta 'poly and CFC is because you guys continue to listen. Thanks.

(But you should STILL put alexman on that per diem!)