Poker program goes for a rematch
In Las Vegas, of course
POKER SOFTWARE called Polaris will play a rematch against human players during the 2008 World Series of Poker in Las Vegas.
Developed by an artificial intelligence group at the University of Alberta in Canada, Polaris will be pitted against several professionals at the Rio Hotel between June 3rd and 6th. Its human opponents will include Stoxpoker.com coaches Nick Grundzien and Ijay Palansky along with Matt Hawrilenko, all of whom have well over $1 million in lifetime winnings from playing poker.
In the Canadian poker playing software's first tournement showdown in July 2007, Phil "The Unabomber" Laak and Ali Eslami edged it out, winning two matches, drawing one match and losing one match.
This year's contest format will be the same as last year's, seven-card limit Texas Hold'em, with each match to consist of 500 hands. The hands will be dealt out in duplicate, such that Polaris gets the same cards dealt in one room as the professional gets in another room. The duplicate format is intended to balance out the luck of the cards and test the poker-playing skills of each contestant. Limit poker requires more consistency of play than no-limit poker.
Theoretically, a computer should be able to beat human players, according to the leader of the university's computer poker research programme. "It's possible, given enough computing power, for computers to play 'perfectly,' where over a long enough match, the program cannot lose money," said associate professor Michael Bowling. "Humans will always make some mistakes, meaning the program will have an advantage."
We're not so sure about that, though. Poker is a very complicated game incorporating not only mathematical betting and statistical odds but also the important skills of expectation, observation and learning, psychology and deception, intimidation and subterfuge.
We give Polaris no better than an even chance to beat its human opponents this time out.
There's more here.
Comments
500 not enough
500 hands is not enough to be statistically significant, I can play 500 in two hours online(4tables at once)Any pro will tell you you need at least 100,000 hands before you can make judgements on your results without luck being involved
It "should" win if it can card count, rigjt?
If it can card count won't it win? I don't know poker shuffling rules, are cards shuffled out of orde after each game or just placed face down on the table and then stacked on the deck so the pc will see every card in the order after couple of hands?We're doing duplicate matches and more than 500 hands
Hi BobCFC,My name is Mike Johanson; I'm one of the programmers on the Polaris project. You're absolutely right that 500 normal hands means nothing, which is why we're not doing that.
All told, we're going to do six duplicate matches of 500 hands (6000 hands total). Duplicate poker is like duplicate bridge. In each match, two humans are on a team together against two copies of the bot. A human and a bot are in room A, and the other human and bot are in room B. The cards dealt to the human in A are also dealt to the bot in B, and vice versa. The players in each room aren't allowed to communicate during the match. After each 500 hand match, we add the scores for each team member together and compare the team scores. This means that the human team and the bot team have both had exactly the same opportunities - the same lucky outcomes and the same bad beats. This takes out *a lot* (but not all) of the luck in the game.
Duplicate poker drops the standard deviation of each hand from 6 to less than 2, meaning that you actually can say something meaningful about each match. After all 3000 hands, we'll have a pretty tight confidence interval to determine the statistical significance of the result.
As to the other poster - the deck is shuffled after every hand, so it's not possible to count cards.