Reported on PhysOrg (
http://www.physorg.com/news134831424.html) on July 9th, the University of Alberta computer program called
Polaris won a poker tournament in Las Vegas recently.
The article describes some of the features the team used to implement and wring out
Polaris as well as the approach the program takes in playing the game.
The University of Alberta's Poker Research Group page (
http://poker.cs.ualberta.ca/) provides the details of the matches, the project participants, and an online Java poker playing client named
Poki. I did not have a chance to play against
Poki, but perhaps if I get some free time I might have a go at it.
In his 2006 PhD dissertation (
http://poker.cs.ualberta.ca/papers/Papers/billings.phd.pdf), Darse Billings outlines four algorithmic approaches that have been explored in the evolution of poker game-playing research:
knowledge-based systems,
simulation,
game-theoretic methods, and
adaptive imperfect information game-tree search. I enjoyed reading through the dissertation quite a bit. Billings did a respectable job of reviewing some of the computer game playing research and the general approaches that have evolved and succeeded. He then goes into detail on the four approaches that are the main subject of his dissertation. I feel the paper is worth reading if you are interested in game-playing. There are plenty of "real-world" applications where lessons learned from game-playing research can be applied.
-Stu
