Previous |
Next |
July 24, 2007, 04:13 PM ET
Robot Rounders
Fresh off of proving that a perfectly played checkers match will end in a draw, computer scientists at the University of Alberta are going back to the game that has become their bread and butter: poker.
The university has pitted its most advanced poker-playing computer program against Phil Laak and Ali Eslami, a pair of high-stakes professional players, in a limit Hold ‘Em match billed as the First Man-Machine Poker Championship. Alberta’s program, Polaris, is the latest in a long line of poker bots designed by the university — including Hyperborean, which won a computer poker tournament last year, and Poki-X, which dueled Mr. Laak (but lost) in 2005.
So far, Polaris is more than holding its own. Yesterday the program squared off in two 500-hand matches against Mr. Laak while simultaneously playing Mr. Eslami with the same hands reversed. In the first session Polaris edged out the human players by a statistically insignificant amount. In the second session, Mr. Laak beat Polaris out of $1,560. But the program, holding the same cards as Mr. Laak, finished $2,515 ahead of Mr. Eslami.
The poker pros and the poker program will play two more sessions today, and if Polaris wins, it can reasonably claim to be an elite limit Hold ‘Em player. But all is not lost for the humans. Limit poker — so named because it places a cap on the amount of money players can bet — is a highly mathematical game, which gives number-crunching computers something of an edge. But the most popular poker variant, no-limit Hold ‘Em, includes healthier doses of guile and brinkmanship, and it will probably be quite some time before computers learn to simulate those traits. —Brock Read
Categories: Research
Add Your Comment
Commenting is closed.