Can you spoof the AI?

[geofelt]

I have noticed that if I start a wonder, the AI will research the required tech and try to start the same wonder if it can. I think you can fool the AI by starting a wonder that you don't want, and switching to your desired wonder at the last minute. This strategy seems to work for me, but I do not have definitive proof. Does anyone have any insight on this?

I have noticed that the AI tends to build about the same number of cities as you do. In order to build many wonders, I have tried some games where I restrict my city count to 4 or 5. The AI seems to restrict the number of cities it builds also, making my wonder buiding job easier.
Conversely, If I expand like crazy, the AI seems to follow, making it harder to take them out later. Comments?

 

[Peaster]

Interesting observations. I build wonders only from vans (instantly) so I just dont' know about the AI behavior you mention in the first paragraph.

It seems the AI does not build as many cities as I do, but I build pretty fast. For example, in a recent GOTM, I built about 35 cities (100AD) and the AI's built about 6 cities each. Another player in the same GOTM built 12 cities, and his AI's built an average of 8 cities each (500AD). Maybe an OCC player can share their experience here.

The way I play, I like to have a lot of cities to build the wonders, ships, and elephants I need for conquest. I don't care too much how many cities the AIs build, because the AIs don't know how to use them properly [especially in the ancient era]. And once their capitol falls, the rest is usually very easy. I do care if the AI builds walls (or the Great Wall), but I've never decided what provokes them to do that.

 

[The Judge]

I have played a few OCC games (though I am no expert) and so I have seen AI civs expand to have more cities than I do but I am unsure whether my non-expansion influences their rate of expansion and I think this would be interesting to test. In those games I have in some instances kept a high power rating (especially as I only had one city) for a while which suggests their may be some truth to this.
With regards to Wonders, I am a little skeptical about whether the Wonders the human player picks has a bearing on the AI choices. The only reason I say this is that in my experience the AI tends to be content building Wonders which I don't want. In a game recently I deliberately avoided building the Manhattan Project (to avoid any possible nuclear strikes and also so I could use caravans to build SS parts by keeping an unmade Wonder aside). Only for the Indians to relentlessly pursue it and begin building it.

 

[Downstrike]

I have noticed that if I start a wonder, the AI will research the required tech and try to start the same wonder if it can. I think you can fool the AI by starting a wonder that you don't want, and switching to your desired wonder at the last minute. This strategy seems to work for me, but I do not have definitive proof. Does anyone have any insight on this?

This has always worked for me.

I have noticed that the AI tends to build about the same number of cities as you do. In order to build many wonders, I have tried some games where I restrict my city count to 4 or 5. The AI seems to restrict the number of cities it builds also, making my wonder buiding job easier.
Conversely, If I expand like crazy, the AI seems to follow, making it harder to take them out later. Comments?

That never worked for me. However, I always play with 7 tribes, so the influence of the other tribes may outnumber my influence.

 

[Ace]

The AI seems to love some wonders, Pyamids, Hanging Gardens, and ignores others, like J.S. Bach. I cannot recall an AI ever even attempting to build that one.

The number of cities an AI civ builds is related to its characteristics. An agressive, expansionist civ will build many more cities than, say, a perfectionist civ. Given a chance by the human player, civs like the Japanese, Russians, Mongols et al, will expand like crazy if the terrain permits. (After all, no civ can expand if its trapped behind a mountain range...).

 

[Ali Ardavan]

The number of cities an AI civ builds is related to its characteristics.

True.

An agressive, expansionist civ will build many more cities than, say, a perfectionist civ.

Only being expansionist matters. Aggressiveness is besides the point in this context.

Given a chance by the human player, civs like the Japanese, Russians, Mongols et al, will expand like crazy if the terrain permits. (After all, no civ can expand if its trapped behind a mountain range...).

Japanese are perfectionist. They rarely build more than a few cities. Russians are not expansionist either; they are the nameless middle (neither expansionist nor perfectionist). The expansionist civs are: Mongols, Vikings, Greeks, French, English, Spanish, Romans, and Celts.

 

[ElephantU]

I cannot speak authoritatively to the issue of number of cities the AI builds, as I tend to play a lot of One City or One Citizen games and the rest are usually Early Landings. But regarding the issue of AI wonder preference, there is definitely an unquantified preference in the wonders the AI pursues which may be tied to the "characteristics" or individual traits of AI civs. Early in a normal random game the AI rarely starts a wonder before 2000BC, and often that can be extended to 1000BC. Of the seven Ancient wonders the AI usually will start with Pyramids, occasionally with Hanging Gardens. There seems to be great interest in Great Library, particularly if the human player is making rapid progress in tech advances. Gifting Lit to an AI civ that is working on HG or some other wonder (especially those less valued by the AI, like Colossus, Lighthouse, or MarcoPolo) will often spur them to swap to GL. The AI rarely pursues Oracle (then again, so do us humans...), and I've only seen a few games where they went after GreatWall. There may be a trigger there between early wars (the "peace" counter?) and GW.

In the next group, the Renaissance wonders, the AI seems more interested in KingRichards and SunTzu, but I think if the human player gets ahead in tech the AI gets interested in Copernicus. Michelangelo, Magellan, and Shakespeare don't seem highly prized by the AI, and the "sleeper" of this group seems to be MarcoPolo. Warfare pressure seems to elevate the interest in SunTzu, and I think very militaristic civs like Mongols, Russians and Greeks are more likely to start it than more civilized civs.

The Industrial wonders seem to be dominated by the AI's interest in Leonardo's, followed by Darwins and the Statue. Rarely do I see the AI try Newtons, even in a tech deficit, except when the others are already taken. Same for AdamSmith and JSBach, and even worse interest in EiffelTower. A note should also be made of wonders that have been made obsolete before they have been started - the AI very rarely will pursue them if they are of no use. I've extended a few games in the modern age to see if the AI will build obsolete wonders and rarely seem them try. If you are going to use caravans or freights to build spaceship parts you need a wonder to load them into, so prioritizing a wonder-obsoleting tech before any AI starts on the wonder will often yield a forgotten wonder in the modern age. My favorite is Oracle, which fits nicely with the number of shields for a spaceship module if you need to rush-buy parts (assuming your city produces at least 20 shields).

In the Modern age of wonders I think the AI is most interested in Hoover, with the rest of the pack being fairly mixed and more likely to be influenced by what tech the AI has at the moment. Sometimes I've seen the AI start a less-desired wonder, even in two or three cities at the same time, then switch to a better one when they get the tech for it. If the human is well ahead in tech the AI will pursue SETI, and if warfare pressure is great they sometimes try to start Manhattan. If there is a "sleeper" in this age it may be Apollo, but I may not be a good judge on that because I tend to play mostly spaceship games.

I'd be interested in what others feel are AI preferences for wonders in different ages, particularly those who tend to play conquest games...

 

[TimTheEnchanter]

I have not paid that close attention to these things or done any systematic research or documentation, but just from casual observations:

The Vikings are never early starters on wonders, but once they do they seem to favor Lighthouse more than most civs do in the early going.

Adam Smith seems to be somewhat more desirable to the AIs than Newton or Bach, probably at least on par with the Statue, but Leo is clearly King during that era.

In the early modern, a lot of them seem to really like Suffrage as well as Hoover.

As far as trying to obsolete a wonder, Oracle seems like a curious choice in OCC games since you really don't need Theology.

 

[ElephantU]

All good points. If I can get Metal or Indust early enough to obsolete GW or KRC I am much happier, but in the late stages of a game I am often forced to take a few off-path techs. First priority is toward Espionage, but only once the AI has gotten Communism. I try to delay getting the nuke techs till I run out of other tech paths, just so there is less time for thefts and attempts at Manhattan. If I'm not OCCing I really like JSBach for that extra edge in city pacification and celebration.

 

[TimTheEnchanter]

Oh, definitely. In a Non-OCC game, getting Theology/Bach is a priority on its own. Obsoleting the Oracle is just a pleasant side benefit.

 

[temurleng]

I can't say I've experienced any connection with my expansion and that of AI civs. However, there does seem to be a connection with my war/research strategy and that of all the other AIs. I've had a number of games where peace that stretches for thousands of years collapses into a warmonger heaven when I started conquering one civ. In the game here, my war with the Indians apparently triggered the Anti-German Alliance to collapse, while my eradication of the Indians and sneak attack on the Arabs seems to have led to the defeat of the English, Romans and Spanish.

I've extended a few games in the modern age to see if the AI will build obsolete wonders and rarely seem them try.

I usually play conquest games that go into the modern era- some into the 3kAD. I've gotten into competitions with AI civs to get the last obsolete wonder, usually Oracle. We try to build it in our greatest cities to get or hold onto the number one city in the world.

the AI usually will start with Pyramids, occasionally with Hanging Gardens... less valued by the AI, like Colossus, Lighthouse, or MarcoPolo.

In the games I play, the AI loves Pyramids, Hanging Gardens and Colossus. Expecially Colossus. They almost seem to pursue it. I think it's a little underrated- cheap bonus trade that doesn't end till flight.