jmb777 said:
I also do not buy into the crap that a computer can't beat a human at war games. A computer can beat a human at chess, so why not? The answer is they just don't put the effort into building the AI. A human designed the game, so you're telling me a human can't program a great player with the rules and know-how to win? Sorry, I don't believe that.
You make an interesting point, but I'm not dissuaded. The gameboard of chess is always the same size, yes, but the pieces have different rules of movement, which Civ 4's pieces do not.
Chess is very much an exception in terms of AI ability. Firstly a huge amount of work has been put into it, and its really only been in the last decade that computers have clearly over taken the very best human players.
Secondly chess is quite conducive to computers. The board is very small compared to the Civ boards, and you don't have the huge array of different
types of square which are present in Civ. As to different rules of movement for pieces, doactually think about what you're saying. Civ has dozens of different units, many with distinct abilities (sea, land, air, siege, mounted, missle, nuke, sub etc.) and promotions means that not even all units of the same type can be treated identically.
Just consider for a moment the starting position of a civ game, OK you've only got 2 pieces - your settler and a scout or warrior. Your warrior can move in any direction (8). Your settler could found in place (1), move 1 tile and found (8), move 2 tiles (16), or move 1 and stop (a human will immediately eliminate this as stupid, but how do you tell an AI that?). That's 25*8=400 possible positions after turn 1, and I haven't even considered reasearch, production or goody huts. It's not even consistent how many positions there are due to sea, mountain, forest etc.
Now consider chess - you have a choice of exactly 20 possible positions to go to on your first move, and the board is always identical - start to see why civ is so much more complex to program an AI for? Chess engines still have a strong element of considering every possible move (there's generally only a few dozen legal moves in any chess position, and many of them are obviously wrong). This is obviously not an option in Civ. I've just shown how many possibilities there are on turn 1, and the difficulty of evaluating which are worthwhile. Now consider a late game Civ position, where there are literally thousands of possible moves, and millions (probably grossly underestimating here) of possible positions for the next turn. The brute force approach used in chess, even with reasonable pruning of posibilities, really won't get you very far.
Consider the board - let's take an 80*80 map - 6400 tiles. Each tile can be any of ice, tundra, desert, plain, grassland, mountain, sea. Each of those may or may not have an overlay (hill/jungle/forest/goody hut or some combination of these), and may have one of up to a dozen different resources on it. Anyone want to do the math for how many possible starting positions there are in civ? It's probably on the order on the number of atoms in the universe.
So that's a short summary of why civ AI is worse than chess. Humans work by spotting patterns, which computers are abysmal at, and so are much better at dealing with positions which are similar, but not quite the same, as in Civ.