On July 23 and 24, the University of Alberta Computer Poker Research Group pitted the latest generation of its computerized poker program, called Polaris, against two top poker professionals, Phil Laak and Ali Eslami. On the line: bragging rights and $50,000 as the two sides were out to determine who, man or machine, is the superior poker player. This time, the humans won. In order to attempt to remove the “luck factor” from the game, the contest was set up in a “duplicate match” style. Both human competitors played simultaneous matches (in seclusion from one another) against two separate instances of the Polaris program. In both of these matches, the same five hundred hands were dealt. In the pair of matches, the human in one room would receive the same cards as the computer in the other room. After all the hands are complete, the combined chip stacks of the computer duo was compared to the stacks of the human team to determine the winner of the round. The first round was a draw, although the computer team did come out slightly ahead. The difference, however, was small enough to be considered a statistical tie. Polaris crushed Laak and Eslami in the next round to take a one to nothing lead. Eslami commented, “Polaris was beating me like a drum.” The professional poker players, relying heavily on their high mathematical aptitude, since there were no physical tells to be seen, bounced back to win the next round. Then the Polaris programmers tried to throw a curveball at the humans, switching the computer’s strategy. Instead of sticking to the same bot that it had been using (the team had developed ten bots, each employing different poker strategies), it implemented a multiple personality attack, alternating amongst three bots to confuse the players. It didn’t work. Laak and Eslami took down Polaris without much trouble. The men were humble in victory, saying the computer program was the toughest opponent they had ever played.
|