Most AI game research is concentrated on deterministic fully-observable games like chess and checkers. The challenge in such games is to find algorithms that search the states' tree as deep as possible and intelligently evaluate states when this is impossible. Poker on the other hand, is a totally different paradigm. The fact that it is only partially-observable (each player can only see its own cards) means that we don't really know in which state we are, and the fact that poker is non-deterministic (we don't know which cards will be dealt next) forces us introduce statistical elements into our algorithms. The need to make economical decisions without having enough information and the need to handle unreliable information about our opponents (opponents might or might not be bluffing) makes poker similar to many real world problems. Brute computational power, great as it may be, is not enough to create a good poker player. It is imperative to develop creative algorithms in order to play the game successfully.

We want to thank Poker Academy for their software contribution.