Old AI vs. Better AI?

[Gr8scott]

I think it would be sweet to have a game where the old AI battles against the new AI. Is that at all possible?

I guess if you fooled around with the source code, you could assign certain civs to use the old AI and other to use the new AI, but that is beyond me

Anyway, some things that would be interesting would be:

-Is the new AI better at pop/research/production? You could compare the average new AI to old AI at 0 A.D.
-Does the new AI usually win? My guess is 4 new AI vs. 4 old AI would almost always end in a new AI win, but it would be nice to see.

If you really wanted to get fancy, you could try the 4 vs. 4 scenario and see who wins, and then restart the game with the AI's changed from old to new and vice versa.

Anyway, just throwing it out there and see if I can entice a coder to mess with this

GS

 

[Mutant]

Sound pretty cool, I'd like to see that, perhaps with a human player propping up the old AI side.

Would it also be possible to assign personalities to the new AI's as well? Like Aggressive, Passive etc.?

 

[Dignitas]

Just start two games. One with the old, the other with the new AI. You can prepare a map and use the same one in both games. Put your own civ on an island surrounded by mountains to put you in safety. Then just hit "End Turn" over and over.

At 0 AD come back and post your results. I'd love to see them.

 

[Gr8scott]

Just start two games. One with the old, the other with the new AI. You can prepare a map and use the same one in both games. Put your own civ on an island surrounded by mountains to put you in safety. Then just hit "End Turn" over and over.

At 0 AD come back and post your results. I'd love to see them.

OK I gave this a shot. Here is the basic setup:

Vanilla Civ
Difficulty: Noble
Speed: Normal
Map: Pangea

I walled myself in and just let the game progress. I let 4 games progress, two with the old AI and two with the new AI. Not a big test, but I was hoping to see some changes.

I recorded when religions were founded and when Wonders were built. There was not any clear difference between the two AI settings. The new AI built stonehenge and Pyramids a bit earlier and the old AI built the Oracle a bit earlier. Pretty mixed though.

I also recorded the power scores.

At 0 AD the top power scores for the old AI were 625 for the first run and 600 for the second run. The new AI top CIV got a score of 602 and 616 in their two runs, respectively.

I thought perhaps the new AI might start to shine if I let the game run out a bit more, so I let all 4 test shots run until 1500AD. Here were the top scores for that:

Old AI top scores at 1500 AD: 2166 and 2058
Better AI top scores at 1500 AD: 2153 and 2083

As far as wars go, the old AI had one conflict in one of the two runs (and I wasn't involved). The new AI had one game with one conflict and one game with three conflicts (I was involved in one of these, so that one was a meaningless war).

Anyway, lots of caveats for this analysis. It is really only a single data point since I only looked at one game and just let different AIs play it out. Also the new AI is playing against the new AI, so that could be limiting somewhat.

Caveats aside though, I was surprised that there didn't really seem to be much difference. I checked that the new AI was indeed engaged when it should be, so I don't think that is an issue (I took the .dll file out of the custom assets folder when I wanted to engage the old AI).

Kind of underwhelming, but probably just a setup problem or a quirk of the game I created.

If anyone else has other data please post it. I thought the new AI would be obviously better, but sadly I couldn't document it

GS

 

[Impaler[WrG]]

Hypocritically if the jdog500 autoplay mod was merged into two DLL's one dumb the other smart and a multiplay game were started between players with each of those DLL's (they would have to ok the Modified contents) and they turn over control the the Autoplay for the whole game you would effectively have an old vs new game to see how the new AI sticks it to the old.

 

[druidravi]

I think emperor and plus would be a better testing ground as it magnifies the advantages of one AI over the other

 

[Monkeyfinger]

Well, what did you expect? You were having the better AIs compete with one another, same for the old ones. What needs to happen is for some civs piloted by Blake's AI and others piloted by the old AI to be put together on the same map, on the same landmass, and play it out.

 

[Yakk]

Gr8scott -- some questions.

1> What handicap file did you use, the BetterAI one or the Default one, for the BetterAI games?

2> Did you have Aggressive on or not?

Second, if you want to test this, you should mix the AI autoplay mod in... It gets rid of the human player problem, where an AI will try to figure out how to attack that weak player.

 

[Gr8scott]

Gr8scott -- some questions.

1> What handicap file did you use, the BetterAI one or the Default one, for the BetterAI games?

2> Did you have Aggressive on or not?

1: Default handicap file in all cases
2: Aggression was off