[Moozh]

[ QUOTE ]
[ QUOTE ]
Nobody will ever build a program to play poker at the same level programs exist to play chess, because as everyone points out every single thread about bots, poker is a game of incomplete information.

[/ QUOTE ]

And, as someone points out in every single thread about bots, this is irrelevant. The fact that a game has incomplete information has nothing to do with the possibility of writing bots for it that play at a certain level compared to humans. See mine (and others) posts above.

[/ QUOTE ]

Actually, it is a significant difference. The reason is that there is a very simple and effective approach to teaching a computer how to play a game with complete information. It is called the min-max algorithm (usually with alpha-beta pruning).

Using this algorithm, you can program a computer to play 'perfectly' as long as it knows the possible moves and the winning conditions of the game. The only factor that prevents the computer from playing perfectly is that the complexity of the analysis grows exponentially with each possible move in the future. Thus, programmers are forced to limit the search depth at some level with an imperfect evaluation function. The deeper a computer can search, the more it can overcome the limitations of its evaluation function. Chess and Go are not solved only because the number of possible moves is too great for a computer to fully compute.

Poker on the other hand is an entirely different beast. Why? Because it's a game of incomplete information. There are adjustments that can be made to min-max for a game of random chance (i.e. Backgammon), but poker involves more than random chance. That's because how a player plays the hand affects the judgement required on that hand (and on future hands too). There just isn't any simple effective algorithm to deal with a problem like that.

That is why the incomplete information issue is a very serious one.