The Second Man vs Machine Poker Championship

Share Article

Superstar players from the online poker coaching site http://www.stoxpoker.com recently did battle against the latest version of Polaris, a poker AI designed by the University of Alberta's Computer Poker Research Group (http://poker.cs.ualberta.ca/).

One goal of the CPRG is to create a poker program that plays better than any human being

On July 3-6 a lineup of elite players from Stoxpoker.com, an internet poker training site, went head to head with the latest version of Polaris, a poker AI designed by the University of Alberta's Computer Poker Research Group (http://poker.cs.ualberta.ca/).

The match was held at the 2008 Gaming Life Expo, just next door to this year's World Series of Poker in Las Vegas. The format for the tournament was 6 matches of one on one limit hold'em. Each match consisted of two humans playing against the computer in separate locations, and the contest used a "duplicate" format wherein the set of cards dealt during the match would be the same at both teammates' tables. At one location a human would play one side of the cards, and in the other location his teammate would play the opposite side of the cards. After 500 hands the winner was determined by whether the humans or the AI had the highest net score. This duplicate format greatly helped to reduce the short-term element of luck and make the matches a much truer contest of skill.

After the first five matches the tournament was tied at 2 wins for the humans, 2 for Polaris, and 1 tie (according to the rules of the tournament a win of less than 25 "small bets," the betting unit on the early streets in limit hold'em, would be declared as a tie). That meant that the last match of the tournament would also be the tie-breaker, and to add to the dramatic element Polaris' last set of opponents was arguably the strongest. Polaris won the match in convincing fashion, winning both sides of the duplicate match for a net win of 90 small bets. This gave Polaris a final record of 3-2-1, and provided the CPRG with their first win over human opponents. The CPRG had created a similar event last year against professionals Phil Laak and Ali Eslami where Polaris finished with a final record of 1 win, 2 losses, and 1 tie.

THE HISTORY OF THE CPRG AND POLARIS:

The Computer Poker Research Group is a not-for-profit research group founded by Darse Billings and Jonathan Schaeffer over ten years ago, well before the "poker boom." Over the years the group has created numerous poker AIs of increasing sophistication. In recent years a number of universities have begun competing against one another by pitting poker AIs against one another in competitions such as the AAAI Computer Poker Competition.

"One goal of the CPRG is to create a poker program that plays better than any human being". Life at the CPRG has not been just about fun and games, however. Many of the techniques involved in solving large games with unknown elements, such as poker, can be adapted to help humans make informed decisions with the help of AIs in many other fields. A variant of one of the algorithms developed for Polaris, for example, has been shown to be useful for laying out sensors in smart buildings in order to identify patterns of use and to improve efficiency. In addition, several members of the CPRG have been drafted into the lucrative online poker industry in order to help online poker rooms maintain security.

The first version of Polaris played against Phil Laak and Ali Eslami last year in Vancouver during the 2007 AAAI conference, suffering a narrow defeat after four matches. This year's model represented some substantial improvements over its predecessor. According to professor Michael Bowling, "There are two really big changes in Polaris from last year. First of all, our poker model is much expanded over last year's. The poker AI now has far fewer holes in its play that humans can exploit. And secondly, we have added an element of learning, where Polaris identifies which common poker strategy a human is using and switches its own strategy to counter them. This complicated the human players' ability to compare notes, since Polaris often chose a different strategy to use against each of the humans it played."

2008 MAN VS. MACHINE POKER CHAMPIONSHIP PARTICIPANTS:

The 2008 Man vs Machine Poker Championship consisted of 6 matches played by 7 elite professionals spread across 4 teams. Below is a short biography on each participant.

Nick Grudzien: Nick Grudzien left a lucrative Wall Street position in 2005 to pursue a career as a professional poker player full-time. He has since become one of the longest-lasting marquee players in online poker, with well over $1M in cash-game winnings in both limit and no-limit hold'em. In 2006 Nick founded stoxpoker.com, a website which provides economical poker coaching through access to instructional videos and recordings of real online play by some of the world's top pros.

Matt Hawrilenko: Matt Hawrilenko is one of the world's most formidable heads-up limit hold'em players, having won well over $1M last year playing online poker. He can be frequently found waiting for challengers under his alias "Hoss_TBF" at online poker's largest cash games. Matt is very familiar with the mathematical and theoretical aspects of poker and advocates playing a balanced, difficult to exploit style.

IJay Palansky: IJay Palansky retired from a career as a litigator in a major law firm to pursue his career as a professional poker player. IJay specializes in limit hold'em, and has won over $1M playing short-handed online cash games. IJay is well known for his aggressive style , often driving his opponents to frustration by forcing them to continually play in unordinarily large pots.

Kyle Hendon: Kyle Hendon is a shorthanded no-limit hold'em specialist with well over $1M in lifetime cash-game winnings. While Kyle may be better known for his no-limit hold'em play he has strong roots in limit hold'em and can field a very competitive one on one game.

Mark Newhouse: Mark Newhouse boasts over $1.7M in lifetime tournament winnings, including the 2006 Borgata Open WPT Championship Event. In addition, Mark is a limit hold'em specialist who has spent a fair amount of time playing in online poker's largest cash games.

Victor Acosta: Victor Acosta has been playing poker professionally for over six years, and is currently a specialist in high-stakes one on one limit hold'em cash games. In addition to his poker career he is also a 5th year grad student at UC Berkeley, studying atomic physics.

Richey McRoberts: Rich McRoberts graduated with a BS in finance in 2004 and moved straight to a career as a professional poker player. His poker career has become specialized in online one on one limit hold'em.

Bryce Paradis: Bryce Paradis is another heads-up limit hold'em specialist with over $2.5M in lifetime cash-game winnings. Bryce retired from poker at the age of 23 to pursue interests in real estate, but is still actively involved in the poker community through his coaching at stoxpoker.com and his work with the Computer Poker Research Group. As Bryce has been directly involved in the development of Polaris his match against the AI (which resulted in a tie) has been counted as an exhibition match.

MATCH RESULTS AND INFORMATION:

The results of each match, as well as logs of the hands played, highlights, photographs and additional information can be found at http://www.stoxpoker.com/man_vs_machine.html .

ABOUT THE SPONSORS:

Stoxpoker.com is an online poker training website which aims to help subscribers play better poker through instructional videos, and is the sponsor of the 2008 Man vs Machine Poker Championship. Stoxpoker features nearly 500 videos where the actual online play of elite professionals is recorded and combined with instructor's commentary to provide amateur players with insight into the mind of a professional player.

We would also like to thank Poker Academy (http://www.poker-academy.com) for providing the poker room software on which the tournament was run.

Contact:

Jim Varnon
jim@stoxpoker.com

Michael Bowling
bowling@cs.ualberta.ca

###

Share article on social media or email:

View article via:

Pdf Print