Histograph worries

Simpleton · October 24, 2001, 14:24

Has anyone else noticed that in every histograph in-game screen shot released the Human player is always the most powerful? I mean the computer players are never even close! This makes me wonder how good the AI really is.....

Granted, the person could have been playing on the easiest level but even the one's released from Firaxis showed this and supposedly they weren't winning on the highest levels or still playing playing on chieftan or warlord.....

I hope I'm just a worry wart and this is not the signs of a poor CTPish AI....

drake · October 24, 2001, 14:27

All civ AI is pathetic...why would civ3 be any different?

An Artificial Idiot isnt going to give any of us a challenge for long. Its only a matter of time before the patterns are figured out and deity level becomes a walk in the park.

Your.Master · October 24, 2001, 15:07

Quote:

Originally posted by drake
All civ AI is pathetic...why would civ3 be any different?

An Artificial Idiot isnt going to give any of us a challenge for long. Its only a matter of time before the patterns are figured out and deity level becomes a walk in the park.

What you do is NOT follow the correct counter-pattern and see if you can do it...

I hope the AI is better than you think.

If it isn't, what would be neat is if the AI recognized if you were doing an anti-AI pattern and changed IT'S strategy completely to counter, in a rock-paper-scissors fashion with ~three separate AI scripts (too much to ask, I know).

But hopefully the AI poses a challenge.

*remembers CivII manual - "not even the people who made the game can beat the computer on Deity consistently"*

drake · October 24, 2001, 15:13

Might be worth playing a few games. I'm just very pessimistic. I have yet to play any game, of any genre that has an AI worth a crap. Enough playing time and just about any AI can be cracked.

Quote:

"not even the people who made the game can beat the computer on Deity consistently"

GodSpawn · October 24, 2001, 15:23

Quote:

Originally posted by drake
I have yet to play any game, of any genre that has an AI worth a crap.

You must be a really good chess player, then...

drake · October 24, 2001, 15:34

Never ventured into that realm gs

Chess certainly has a lot of variables, but not even close to the same amount as a game like civ.

I am guessing it would be harder to write a very difficult AI code for civ, than a very diificult AI code for chess.

Course I could be wrong

Earthling7 · October 24, 2001, 15:41

I am concerned as well. Yin posted a Korean review where the AI seems to be acting very... AI.

We shall se...

Akron · October 24, 2001, 18:44

Quote:

Originally posted by drake
Never ventured into that realm gs

Chess certainly has a lot of variables, but not even close to the same amount as a game like civ.

I am guessing it would be harder to write a very difficult AI code for civ, than a very diificult AI code for chess.

Course I could be wrong

Actually, I'm not sure about that. Chess is a very well known game around the world, and there are many programmers who have worked and competed against each other to create difficult AIs for chess. Also, the rules of chess are standard . Civ, on the other hand, is constantly evolving and there is only one small group of people with little competition (CTP doesn't count) developing an AI for it. If they released the source code or whatever is needed to modify the AI and several groups of programmers were to compete to make the best AI for it, then they might eventually get an AI that could hold its own on deity.

I really feel that AI is something that games often neglect to develop fully in order to cut corners, as stuff like cool graphics and features seem to be noticed far before the AI by most casual gamers. Also, casual gamers may get frustrated by difficult AI.

dennis580 · October 24, 2001, 20:16

Dan has said sevral times that Firaxis testers had a cheap mode they were using. So don't judge the AI until you play the game. You just might be shocked at how skilled it is.

Lawrence of Arabia · October 24, 2001, 20:24

Well, to tell you the truth, I can't even read the histograph. The one for Civ2 was better because it compared you with a 0. THis one doesn't make any sense. Is it just whoever has the thickest line or what?

Simpleton · October 24, 2001, 20:38

Quote:

Originally posted by dennis580
Dan has said sevral times that Firaxis testers had a cheap mode they were using. So don't judge the AI until you play the game. You just might be shocked at how skilled it is.

Yes, perhaps that's it!!

But my mind would be more at ease had they posted some histographs with the computer ahead.....

Simpleton · October 24, 2001, 20:39

Quote:

Originally posted by Lawrence of Arabia
Well, to tell you the truth, I can't even read the histograph. The one for Civ2 was better because it compared you with a 0. THis one doesn't make any sense. Is it just whoever has the thickest line or what?

I read it as whomever is the thickest at the bottom of the graph is ahead or has won.

Lawrence of Arabia · October 24, 2001, 20:51

Quote:

I read it as whomever is the thickest at the bottom of the graph is ahead or has won

Which sucks because your strength is only relative to the other players in the game. If i want to compare myself from game to game, I may be stronger in one of them, but couldn't tell because all oft he other civs are 3x stronger in my game than in the last one.

SITS · October 24, 2001, 20:55

Well, the AI doesn't matter that much as we can get a good challenge from MP.

squid · October 24, 2001, 22:49

Quote:

Originally posted by drake
Never ventured into that realm gs

Chess certainly has a lot of variables, but not even close to the same amount as a game like civ.

I am guessing it would be harder to write a very difficult AI code for civ, than a very diificult AI code for chess.

Course I could be wrong

Actually, chess is hard due to the horizon effect....one move outside of the search "window" (move 8 if you can search 7 moves ahead) can upset the board, forcing something called a quiescence search to be made. Second to Go, chess is the most challenging AI to do well. Most chess AI's are technically not doing 100% AI work, but are consulting opening and endgame databases for as long as possible.

Though chess has a smaller branching factor than Civ, Civ is not a game where the board is 100% visibile to all sides. The AI must then rely on perceptors.

Being Civ is not a zero sum game with a revealed board, AI comparisions between Civ and Chess are a mute point.

Any other C.S. majors feel free to dispute....I just don't think it's a valid comparision.

Inverse Icarus · October 24, 2001, 22:50

$ai < $human;

require($multiplayer);

squid · October 24, 2001, 22:53

Also, there are two fundamental approaches to AI (see Russell&Norwig)....you either program to make logical decisions or you program to make it think as a human would....the second type is more cognitive. Chess is more of a brute force hack....until we get neural nets involved in the process and through out the databases, chess AI does not learn.... it does lots of math and generates lots of data. Chess AI's do not think even close as to the way they think. They just evaluate every single option and every result, with a few optimizations. There isn't anything advanced or noble about a Chess AI, but it is hard to make a good one run quickly and to make one think like we do.

Civ has it easier in some ways but harder in others. The Civ has priorities other than "get more points", it has strategic goals. Rather complex. I suppose much could be learned from looking at what FreeCiv is doing....maybe I'll read up on it.

Sava · October 25, 2001, 07:46

Comparing chess to civ is bad. What makes chess so difficult is the limited number of movement options. Also, both players start with the same amount of pieces and the only way to go is towards your opponent.

They are both strategy games, but they are so different you can't make an effective comparison.

Earthling7 · October 25, 2001, 07:58

Quote:

Originally posted by Lawrence of Arabia

Which sucks because your strength is only relative to the other players in the game. If i want to compare myself from game to game, I may be stronger in one of them, but couldn't tell because all oft he other civs are 3x stronger in my game than in the last one.

Many games use this method, but have an option to use the normal graph as well. I wonder if this is the case with Civ3.

Pyrodrew · October 25, 2001, 08:12

This new histograph scoring will definitely be different & interesting. After the game is out someone should put a poll up whether it is better or worse. From what I understand it seems that the Civ that wins could even be a Civ that died decades ago... that you killed.