Is this gossip about the AI true?

[r_rolo1]

Maybe Cultural could be another option that can be programmed, but if you look at the way the AI wages war, they can't even defend themselves properly.

I don't think that Civ4 cultural victory is a easy to program one ( civIII cultural victory was a entirely different issue, just requiring settler spamming and temples) because the degree of long term planning required ( expand religions to use cathedrals, theaters for artists, mercantilism and or SoL for free specialists, the rightful placement of cathedrals, Hermitage, and broadcast antenas, the Eiffel tower,broadway, rock and roll, you know, all that stuff) is, in my point of view, way above the capacity of Civ4 Warlords unpatched AI. But diplomatic would be even worse: how to teach Monty to not be such a moron and to keep strategic friends? And how to convince Toku that and Open Border agreement could be a necessary step for a diplomatic win?

 

[DrewBledsoe]

I posted a while ago that in principle a cultural win should be impossible,and I still think this to be true. If an ai is lets say about to complete a cultural victory, then all you have to do (probably to your own detriment) is concentrate every possible available troop with the task of destroying one of those 3 cities. Given even reasonable tech parity, I dont see how this can be stopped, even though it would probably ruin the players winning chances.

The only reason a player ever gets a cultural victory, is because the ai at present allows it. If the ai were programmed with the same "spoiler" mentality as noted above, the player would never get a cultural victory either, making that type of victory obsolete.

Kinda catch 22 ?

 

[Alsark]

Indeed - when people go for a quick cultural victory then they are essentially "giving up" in all other areas of the game. With their commerce going entirely into culture (for at least three cities), and their culture slider maxed out (or close to it), they'll take somewhere around 100 turns to research anything once when they have the religion that they want and some of the artistic technologies. In fact, the player will probably avoid the "Scientific Method" in any way that he or she can, because it will obsolete their monasteries, which provide a culture boost (or would that culture bonus still count as long as it was already built?), and since the Scientific Method leads to anything that can be considered remotely modern, the player would basically be stuck in the high medieval era for some time.

At any rate, the end result would be that somebody going for a quick cultural victory will lack all technological power and military might, allowing for any other player or AI to stomp them to the ground. Considering the game tells you who has the highest cultural cities, a player would be smart enough to realize that you're going for a cultural victory and stop this. So, like DrewBledsoe said, players can only win the cultural victory against the AI because they allow it, and it should probably be integrated into their AI that they should stop people as they draw close to a cultural victory. All they'd need to do would be take one of their three main cultural cities and then they'd have delayed the player a good hundred or couple hundred turns (unless they have a back-up fourth cultural city, but frankly it wouldn't be difficult to take over the entire territory of somebody who has gone only for culture).

 

[Sisiutil]

In fact, the player will probably avoid the "Scientific Method" in any way that he or she can, because it will obsolete their monasteries, which provide a culture boost (or would that culture bonus still count as long as it was already built?), and since the Scientific Method leads to anything that can be considered remotely modern, the player would basically be stuck in the high medieval era for some time.

No, cultural buildings keep producing culture; you just can't build them anymore. And for monasteries, the 10% science boost is what goes away with SM.

You can achieve a cultural win late in the game from a position of strength, using Caste System and artist specialists rather than the slider. In some ways this might be an easier way to program the AI to go for a cultural win.

 

[ojchase]

IIR--if i remember

but you might want to double check that as im also a noob. theres a list of common acronyms in here somewhere.

Thanks....I didn't even think that it might be a common acronym since it came right after BtS. I just assumed it must be a subtitle or something!

 

[r_rolo1]

I posted a while ago that in principle a cultural win should be impossible,and I still think this to be true. If an ai is lets say about to complete a cultural victory, then all you have to do (probably to your own detriment) is concentrate every possible available troop with the task of destroying one of those 3 cities. Given even reasonable tech parity, I dont see how this can be stopped, even though it would probably ruin the players winning chances.

The only reason a player ever gets a cultural victory, is because the ai at present allows it. If the ai were programmed with the same "spoiler" mentality as noted above, the player would never get a cultural victory either, making that type of victory obsolete.

Kinda catch 22 ?

Agreed. But the AI in this moment can't take the necessary countermeasures against an easily recognizable Domination victory (just read the numbers , if necessary bribe someone and attack the guy with all you've got ), so thats not a scenario in the moment

You can achieve a cultural win late in the game from a position of strength, using Caste System and artist specialists rather than the slider. In some ways this might be an easier way to program the AI to go for a cultural win.

Maybe that can be done. Something like:
1)Beeline to Philosophy

2)Tech normally, like the AI in his clear mind moments do, and maximixe and P.S : Build decent army

3) Somewhere in the early industrial age ( or before ) switch to Caste system and pile a lot of Artists in three cities

4) Any GA that pops up must culture bomb the most culture-weak of those three cities

5)Try to get the +50% late wonders and the Eiffel tower or broadcast tower in the three cities

6) With luck, you can win a cultural win

Of course that can be devised some countermeasures for that. But it would make the game more interesting.

The problem is that, even if the AI could follow this path, how to make the AI to choose wisely between victory types? And even more difficult, how to program the AI to act against a player (human or not) that is near to win the game?

 

[LlamaCat]

in my first game playing Warlords, China tried to win diplomatic victory a couple times.. they lost by 1 single vote in the latest poll!! I was amazed how close I came of course i went on to retire and lose later anyway... I find Warlords AI behavior has definitely been updated somehow

I think the AI limitations on victory attempts are a joke, they need to get their asses in gear for the next expansion. In fact, why doesn't this game allow 10-15 different victory types at least for the human player? there are so many units, buildings, etc that you could easily come up with more ideas. diplomatic, fine. space race -- boring, but ok. need other science type of victories. cultural -- very boring to play, but it could be expanded into different types of cultural victories. conquest and domination, fine, but those could be expanded. religious victory (ever play Fall From Heaven mod?) definitely needs to be added....
how about a victory having to do with spying - steal secret plans or something? (that unit needs to be strengthed and given more related units with more effectiveness, by the way, like diplomats that can be upgraded to spies, that can be upgraded to secret agents, etc)

 

[bardolph]

I've played games with BetterAI where the AI went for Cultural Victory. My solution of course was to raze one of the cities. It wasn't all that hard, unfortunately, but it was fun to have threats other than space race to deal with.

I agree that Cultural Victory is too easy to spoil, due to razing. Then again, maybe it IS possible to defend your three core cities, with good unit management.

 

[Yakk]

AIs going for cultural victory can force you into a war you don't want to do.

They could be a vassal of a huge friendly military state, for example.

And in Blake's AI, the AI is slightly better at war: it sends stacks of doom at you. Considering that was written by two people part-time, I am convinced that civ programmers can do better.

 

[Sisiutil]

Maybe that can be done. Something like:
1)Beeline to Philosophy

2)Tech normally, like the AI in his clear mind moments do, and maximixe and P.S : Build decent army

3) Somewhere in the early industrial age ( or before ) switch to Caste system and pile a lot of Artists in three cities

4) Any GA that pops up must culture bomb the most culture-weak of those three cities

5)Try to get the +50% late wonders and the Eiffel tower or broadcast tower in the three cities

6) With luck, you can win a cultural win

Of course that can be devised some countermeasures for that. But it would make the game more interesting.

The problem is that, even if the AI could follow this path, how to make the AI to choose wisely between victory types? And even more difficult, how to program the AI to act against a player (human or not) that is near to win the game?

That's pretty much it, except you also want to be the first one to Music (free GA) and also build the Sistine Chapel (+2 per specialist). Neither are essential, but they are tremendously helpful.

These early targets could serve as indicators to the AI as to whether or not the cultural victory is worth pursuing. If it fails to get the free GA and to build the Sistine, then it could be programmed to punt a cultural victory.

 

[largedarryl]

I agree that the betterai has done wonders for the challenge given by the ai, but as mentioned, basically when an ai starts down the path of a cultural victory it ussually means 3 of the best cities in the world will soon be mine, hence making me even stronger.

I am also worried about what the firaxis group is going to do to the ai, especially since they make note about expanding the space race victory, and since the current space race is the only victory that an ai can competantly persue, making that more difficult may make a time victory the only thing that the ai can accomplish.

 

[bardolph]

AIs going for cultural victory can force you into a war you don't want to do.

They could be a vassal of a huge friendly military state, for example.

And in Blake's AI, the AI is slightly better at war: it sends stacks of doom at you. Considering that was written by two people part-time, I am convinced that civ programmers can do better.

These are all good things, aren't they? When an AI takes actions that force me to respond, then the AI is doing its job. Otherwise, I might as well just play solo vs the Barbarians.

 

[r_rolo1]

That's pretty much it, except you also want to be the first one to Music (free GA) and also build the Sistine Chapel (+2 per specialist). Neither are essential, but they are tremendously helpful.

These early targets could serve as indicators to the AI as to whether or not the cultural victory is worth pursuing. If it fails to get the free GA and to build the Sistine, then it could be programmed to punt a cultural victory.

Nice point, and probably it is easy to write the necessary code for that, even making some leaders ( thinking on Creative or Industrious traits) to be more inclined to follow that path ( something like the "Oooh, shiny!!!" factor for the wonders ). Do you have some ideas about other victory types, like Dom/ Conquest or Diplomatic? In your ALC's ( I've read them all ) you already showed a lot of ways to win this game in a lot of different conditons with some of the more difficult leaders of the game, and maybe some of the methods you used could be used for an algorythim for a certain kind of victory ( maybe axe or cat rush for a dom/ conquest victory, with the necessary teching?).

 

[Sisiutil]

Nice point, and probably it is easy to write the necessary code for that, even making some leaders ( thinking on Creative or Industrious traits) to be more inclined to follow that path ( something like the "Oooh, shiny!!!" factor for the wonders ). Do you have some ideas about other victory types, like Dom/ Conquest or Diplomatic? In your ALC's ( I've read them all ) you already showed a lot of ways to win this game in a lot of different conditons with some of the more difficult leaders of the game, and maybe some of the methods you used could be used for an algorythim for a certain kind of victory ( maybe axe or cat rush for a dom/ conquest victory, with the necessary teching?).

Well, the diplomatic victory in the Inca ALC game showed that in many ways that's a matter of number crunching (is my diplomatic score with the other civs higher or lower than whoever would be my rival in UN elections?). On the one hand, then, programming the decision to go for a diplomatic win late in the game shouldn't be difficult. The difficulty is everything before that--monitoring and adjusting relations as needed through the whole game to make sure you're well-liked enough to win a diplomatic victory.

Of course, the very notion that the AI civs would give you a diplomatic win seems strange to me. They won't trade you techs that give you the ability to build spaceship parts, so why would they vote for you when you're trying to gain a diplomatic win? Of course, if they didn't, it would be impossible to every win that type of victory. Just another Civ paradox.

Conquest wouldn't be too hard: am I more powerful than anybody else? Right then, I'm off to kill everybody! Don't worry about keeping my dinner warm, honey!

Domination strikes me as the hardest of all for the AI. It's one thing to raze cities, quite another to capture them, defend them, expand their cultural borders, manage war weariness, and keep the economy and research going while pursuing a domination win. Especially when the AI is so pathetic at war.

 

[Nikis-Knight]

But diplomatic would be even worse: how to teach Monty to not be such a moron and to keep strategic friends? And how to convince Toku that and Open Border agreement could be a necessary step for a diplomatic win?

The key would be to not have every ai try for every victory, but have each go for one of two, or both perhaps, depending on what their traits and personality incline them towards.

 

[Sisiutil]

The key would be to not have every ai try for every victory, but have each go for one of two, or both perhaps, depending on what their traits and personality incline them towards.

This makes complete sense to me. Mansa would be focused on launching the space ship, Louis would be building wonders towards a cultural win, while Monty would be trying to kill everybody.

 

[r_rolo1]

The key would be to not have every ai try for every victory, but have each go for one of two, or both perhaps, depending on what their traits and personality incline them towards.

That might not be a good idea, because it could put the AI in a complete stalemate if none of the coded victories were realistically acheivable. Imagine, for the sake of example, Monty. Type of victories : maybe Dom or conquest ( fits him nicely). Put Monty in a small isolated continent ( like ALC 14 ) and Monty can't win. A human player in that situation would think cultural ( Myst as starting tech = early religion ), but Monty wouldn't pursue it because it's blocked.
A weakened form of your idea could be useful, one or two prefered victory types, but keeping the other options opened.

 

[r_rolo1]

This makes complete sense to me. Mansa would be focused on launching the space ship, Louis would be building wonders towards a cultural win, while Monty would be trying to kill everybody.

And Ghandi would try to win by UN vote, and Alex would conquer first and then lightbulb things until exaution, and Toku would ... er .... "grasp" ...
be a jerk? ( Really ,I don't like the guy's mood; to me a good Toku is a dead one )
We have to admit the game would be more interesting

 

[DrewBledsoe]

Its actually far more interesting already, just play with random personalities on, it makes the game (sometimes a lot harder), but always more fun. Its far too easy when you meet say Mansa (with standard personalities) in 3000Bc and know exactly how he will behave, and exactly how to treat him....in fact its one of the biggest edges you have as a player ( the ai never knows this about you in contrast).

You might get a dozen psychos as opponents, you might get a dozen "monopoly tech" traders, thing is you don't know for a long while, and have to work all of the ais out each new game. Its mostly a little mix and match. Believe me, Ragnar going Religious and tech trading mad, whilst still up for a fight at the drop of a hat, is a monster.

On the topic in question (which got me to this) how on earth can you say "x" should go for this or that victory, and "y" should go for that or that? For someone who now always plays random personalities, it just doesn't make sense. If the ai is to be programmed to go for different types of victories, then it should consider all of them, given the right circumstances.

After all I've often played as say Gandhi or Hatty, and found my only real option to be conquest. I'm sure something similar has happened to us all. Having Gandhi only really ever (as an ai) go for a diplo or cultural victory, makes the game even more predictable, which im sure no-one wants....

 

[r_rolo1]

On the topic in question (which got me to this) how on earth can you say "x" should go for this or that victory, and "y" should go for that or that? For someone who now always plays random personalities, it just doesn't make sense. If the ai is to be programmed to go for different types of victories, then it should consider all of them, given the right circumstances.

After all I've often played as say Gandhi or Hatty, and found my only real option to be conquest. I'm sure something similar has happened to us all. Having Gandhi only really ever (as an ai) go for a diplo or cultural victory, makes the game even more predictable, which im sure no-one wants....

I agree with you on the point that all of the victories should be at the disposal of the AI, because the last thing that we want is a more dull AI that behaves like a swiss watch. But I also think that every AI, based on his traits, UU and UB and general behaviour, should have preferences about the victory type ( like they have on civics, but more difuse ). But preferences, not hard coded victory types, because, like I said above, if one of then is a impossibility, the AI would stall, with no precise goal ( like they look right now). Maybe something like the "Oooh, shiny!" factor for the wonders atached to the victory types ( I've already saw Monty, the guy with the smallest "Oooh, shiny!" factor, enter in wonder crazy mode sometimes).

But what I would really like to see would be an AI "think" to itself: "Ok, I've got this traits that give me X and Y , this starting techs, this UB, that UU, this start position, and I like of doing things in a H manner ,so maybe I would try to go Z victory, teching like this... OH #"&% , I can't go for Z victory because of W.... maybe I'll try K victory". That kind of algorithym would make the AI behave in a more flexible and efficient way ( more like an human player , that consciously or not, uses the think path described above ), would be usable in random personalities and would provide a more challenging and interesting game