Poker hotshots narrowly defeat bluffing computer

After two thousand hands and countless "flops", "rivers", and "turns", two elite poker players have narrowly defeated a formidable computer opponent. The result means that, while chess world champions have fallen to computers, humans still hold sway in poker, a game where psychology plays a huge role. Phil "The Unabomber" Laak and Ali Eslami took […]

After two thousand hands and countless "flops", "rivers", and "turns", two elite poker players have narrowly defeated a formidable computer opponent. The result means that, while chess world champions have fallen to computers, humans still hold sway in poker, a game where psychology plays a huge role.

Phil "The Unabomber" Laak and Ali Eslami took on Polaris, software developed by researchers at the University of Alberta in Canada, in a set-up designed to reduce the role that luck normally plays in a game of poker.

The pair played Polaris simultaneously in different rooms, with computer and human playing opposite hands in each game. In other words, if Laak was dealt a full house, Polaris would have exactly the same hand, at the same time, in its game against Eslami.

The game was played at a meeting of the Association for the Advancement of Artificial Intelligence in Vancouver, Canada.

Hidden information

At the end of play on Monday, Polaris had tied the first round of 500 hands and triumphed in the second, finishing almost $1000 up against the humans. But two wins in Tuesday's sessions earned Laak and Eslami overall victory. The pair, who said they were exhausted by the rapid rate of play, admitted to being impressed with the ability of Polaris.

Poker is harder than other games for computers to crack because of the importance of studying other players' tactics and behaviour. In chequers, for example, a computer can work out the best move simply by knowing the rules of the game and the current position.

Earlier this month, computer scientists said they had created software that would never lose a game of chequers.

But poker is different because it contains "hidden information", says Graham Kendall, a computer scientist at the University of Nottingham, UK.

'Over for humanity'

A good player will look at opponents' facial expressions in a bid to guess what cards they might hold, for example. Bluffing is also an important way of fooling an opponent, but the strategy behind this is hard to programme into a computer.

Polaris is, however, one of a few new poker-playing machines that have begun to master the tactics behind bluffing, although Kendall says a good human player can probably deduce a computer's tactics more easily than vice versa. "I'd probably go for the humans," said Kendall when asked to predict the winner before the match.

Yet Kendall and others admit that machines are catching up. Kendall predicts that a computer could become world champion in around 10 years, provided tournament organisers allowed machines to enter.

In a previous match against Polaris, Laak folded a hand and declared "if that is a bluff, it's over for humanity". Ominously, Polaris had indeed been bluffing.

Game, Gaming, Poker, Computer, Game News

Source:→ NewScientistTech