AI Victory Condition.

[Spoonwood]

Do the civ III tribes appear to pursue a victory condition?

In my experience, the answer is never. I might get conquered. The AI might happen into a 100k victory. They might launch the spaceship after a while. But, they never seem to behave as if they could pursue a victory condition. One might say the AIs (I prefer to think of each tribe as a separate AI) have no strategy. It seems almost sad to say, but one might say... no AI ever wins this game, only the human player wins or loses.

It might seem silly to actually find an AI that would build two cities, take one city to size 12 early, try to build wonders, and research accordingly, since they probably would be easy to conquer. But, at least then you could say "yep, the AI had a strategy." The AIs don't use ICS spacing and whip like no tomorrow in Feudalism... as they basically never adapt their city spacing. I can't say I've seen the AIs avoid researching optional technologies to expedite climbing the tech tree for a spaceship or diplomatic victory. Or say researching Literature to help with this. Or how about an AI that would refuse to build any wonders, refuse to capture cities after a certain number or percentage of territory, and burn cities to the ground in endless wars?

How would the game play if you had 1-3 tribes going for a 20k win, 1-3 for a spaceship win, 1-3 for a diplomatic win, and 1-3 for a militaristic win?

 

[GamezRule]

In my experience, the AI plays turn by turn, so they can't have a long term strategy. It would be nice for the modders to implement.

 

[Spoonwood]

In my experience, the AI plays turn by turn, so they can't have a long term strategy. It would be nice for the modders to implement.

I basically know nothing about computer programming, but I'd guess that changing how the AI behaves would require access to the source code. Here's the catch...

It might make the game easier in some respects if not many. But, if it did, I wouldn't mind so much, because at least then the game would play differently every time, instead of having the ability to predict things like "Monty will build a temple after he finishes on his spear." (o. k., I don't know that much about the AIs, but I think you get the picture). I'd much rather have an interesting set of AIs to play against than AIs which were more difficult to beat. By "interesting" I mean a variety of behavior patterns, which the victory conditions naturally fall into.

 

[MysteryX]

I haven't tracked it in detail, but it does seem to me that the AI will follow production and research strategies that favor the particular civ's abilities. That makes certain victory conditions more likely. I've played quite a few games in PTW as Egypt, using culturally linked starting locations, and got a lot of familiarity with what Rome, Carthage, and Greece would do. Rome researched military techs first, Carthage favored commerce techs first, Greece favored techs that enhanced research first. That influenced the buildings they built first.

That isn't the same as long-term strategy, but it mimics long-term strategy fairly well for an AI.

Unfortunately, I don't have a lot of experience with the AI achieving a victory condition, because I will concede the game when it becomes hopeless. However, in one of my earlier games, I think it was Greece that came up with a Space Race win that took me by surprise (I wasn't keeping track of where it was in the Space Race).

When it comes to wonders though, it appears AIs build the first available. I ought to pay attention when cascades start to happen, to see if an AI civ will pick one over another when having a choice.

I have also noticed that altering the victory conditions does not seem to change the way the AI researches tech. Some of the tech lines on the modern tree have the most use for a Space Race victory, but minimal value outside of the Space Race. They will still go up trees like Space Flight -> Superconductor long before I do.

 

[Ataxerxes]

The AI really does play turn-by-turn without a long-term strategy. This is true in both III and IV (IV has some of the most mind-numbing contradictory moves you can imagine). The only victory it "pursues" (using the term very loosely) is conquest by some of the more warlike leaders.

AI hasn't a clue about culture victories and doesn't bribe to get the UN win. If it techs well it might lurch into space but the journey will have a lot of "wait a minute, why did it do that?"

 

[GamezRule]

I don't think that the AI will win by anything other then Domination, Space, 100k, or Diplomatic. And they can always get the top score for victory if time runs out.

 

[Takhisis]

I basically know nothing about computer programming, but I'd guess that changing how the AI behaves would require access to the source code. Here's the catch...

Yes you do, and Firaxis refuses to release it. What you can do in scenarios is preplace resources to induce the AI to attack and other tricks.

It might make the game easier in some respects if not many. But, if it did, I wouldn't mind so much, because at least then the game would play differently every time, instead of having the ability to predict things like "Monty will build a temple after he finishes on his spear." (o. k., I don't know that much about the AIs, but I think you get the picture). I'd much rather have an interesting set of AIs to play against than AIs which were more difficult to beat. By "interesting" I mean a variety of behavior patterns, which the victory conditions naturally fall into.

Sadly, as I said above, you can't do that. You can customise civs to the extent that they'll have different units, wonders and even techs, but you can't tell the bloody thing what to do unless you make it a one-tech path with no options or combinations and give it a straight-line of units customised, once again, for each civ.

 

[Aabraxan]

I know diddly-doo about computer programming, but my feeling was always just the opposite. It's not that the AI doesn't pursue a VC; it's that it pursues them all. It just that it: (a) doesn't have any focus; and (b) doesn't pursue any of them well. Where a human player may pick a VC at 4000 BC (& this seems particularly important for something like a 20K game), the AI tries to build for all VCs, all the time.

 

[Spoonwood]

I know diddly-doo about computer programming, but my feeling was always just the opposite. It's not that the AI doesn't pursue a VC; it's that it pursues them all. It just that it: (a) doesn't have any focus; and (b) doesn't pursue any of them well. Where a human player may pick a VC at 4000 BC (& this seems particularly important for something like a 20K game), the AI tries to build for all VCs, all the time.

I've felt the AI acts this way also. But, pursuing all victory conditions at once has much the same effect, in my opinion, as pursuing no victory condition at all, as unless you play like SirPleb did in one game you have to win by one victory condition or another.

 

[Aabraxan]

Agreed. Chasing all VCs is very much like chasing none.

 

[dominatr]

the AI doesnt go for victory conditions, they usually just quabble amongst themselves while i stab them in the backs.

 

[Quintillus]

It seems to me that the AI might be a bit more likely to declare war on you when you're quite close to a victory, especially domination, but this seems to hurt them as much as help. The AI certainly doesn't look at it beyond an individual turn - if it did, it would have to save information about them so they'd work after a reload, and no one's ever found any evidence of that.

It would be rather interesting if the AIs more heavily pursued certain paths - MysteryX is right that certain civs are naturally inclined a bit toward certain victories by their traits, but they never go all out towards any one victory. Might make for a more warlike game, too.

 

[Spoonwood]

It seems to me that the AI might be a bit more likely to declare war on you when you're quite close to a victory, especially domination, but this seems to hurt them as much as help. The AI certainly doesn't look at it beyond an individual turn - if it did, it would have to save information about them so they'd work after a reload, and no one's ever found any evidence of that.

It would be rather interesting if the AIs more heavily pursued certain paths - MysteryX is right that certain civs are naturally inclined a bit toward certain victories by their traits, but they never go all out towards any one victory. Might make for a more warlike game, too.

It might if the AIs went after a military victory condition. Though, there could still exist some variation here say if an AI went for a conquest win instead of a domination win. Like say you played against Shaka, Montezuma, the Incas, and Khan. In that game say Shaka and Khan always raze cities (too much of a hint?) while Montezuma and the Incas ignore all wonders but the Temple of Artemis, Sun Tzu's, and Leo's Workshop, ignore all libraries, etc., and always keep cities.

But say you play against the same 4 in another game, but this time Shaka puts all his cities as close together as possible. Khan researches Literature and keeps offering you 10 gold every 20 turns. He also gives you optional techs at a discounted price every so often. Montezuma occasionally starts to build wonders and has very few cities. The Inca leader also researches Literature, but also Republic and refuses to revolt again. Sounds like a lot less war to me.

I certainly know how a game goes 7 AIs will behave when they all play in the same aimless way. But, how does a game go when a priori each of the 7 AIs have 6 different possibilities of playing the game?

Now some of this, *might* lead to the game becoming easier sometimes. That's a possibility, sure. But, could you imagine trying to beat Deity or even Demi-God level AIs with its production bonuses if *you and the AIs* all played for a 100k game? Or a 20k game if they actually had *half* a clue what to do with workers in such a game? Or even a military game where they concentrated on building units and conquering instead of trying to do everything at once? Even if some games became easier, and you could predict what victory conditions the AIs went after, beating the AIs at the same victory conditions they selected might not prove so easy.

I'm not saying that any of these ideas would make for an easy change of the AI programming, just it would make things more interesting.

 

[Raliuven]

I'm not a computer programmer, but four options come to mind:

1) The AI is randomly assigned a VC. The AI will seek to execute this VC regardless of how suicidal it may be. We all know how tenacious the AI can be when it gets a bad idea stuck in its head. But occasionally you will be matched with a handful of AIs that are well suited and positioned for their VC and they may be a true challenge. Draw back - once you have identified the VC of a Civ, you can cripple and crush them.

2) The AI assess 'the land' at a certain time in the game - maybe around 2000BC and it decides at that time what VC to pursue. That might make 20K difficult to attain. Thereafter, the AI continues along like option 1, hell or high water. Of course we all know how bad the AI is with surveying land that is not within its immediate borders and even land that is within its borders. Same draw back as option 1, but each AI would have a better chance of being competitive.

3) Same as two but it assesses at multiple times - 2000 BC, 10AD, etc. Whatever. But this could result in some pretty schizophrenic AIs - not that they don't act insane already.

4) The AI continually assesses its position and sets the VC that it is closest to achieving (or has the highest probability). It seems to me that would kill the processing time and would require every aspect of the game to be assigned a VC point system for the AI to make a decision. In large developed games that could be a huge amount of data. But as you work to upset their VC, they will switch gears to a new VC if the present one is obstructed by the human.

Honestly I don't know if any of these options would actually improve the game. (And at higher levels might actually make it impossible to win?) There is a long list of 'stupid AI' actions that haven't been remedied, any one or three of which could seriously change the challenge of the game without the complexity of a VC.

 

[GamezRule]

Well, it might make the higher levels, but then again, if the AI's knew what they were doing, a production bonus would make them high near impossible to beat, so regent would no longer be an easy level, and the game would max at monarch.