Online Pros Barely Beat Machine In Poker Challenge

Doug Polk, WCGRider, Bjorn Li,  Claudico,

Bjorn Li, who leads the human race against the latest no limit hold’em artificial intelligence program, Claudico. (Image: pokerstars.com)

Mankind has triumphed against the machines in a two-week 80,000 hand no-limit hold’em battle Rivers Casino in Pittsburgh, but just barely.

In fact, the results were close enough to statistically be called a draw, according Dr Tuomas W. Sandholm of the University Professor of Computer Science at Carnegie Mellon University, whose team developed Claudico, a no-limit hold’em playing bot.

Humankind, represented here by Doug “WCGRider” Polk, Dong “Donger Kim” Kim, Bjorn Li and Jason Les, won nine of the 13 days of heads-up cash play against the computer, with a total difference of $732,713 in theoretical money.

The bot came fifth out of five, but the results suggest that artificial intelligence is becoming more adept at playing no-limit hold’em than we had previously imagined.

Another computer program, Cepheus, developed by the University of Alberta, claimed earlier this year to have “solved” the game of fixed-limit heads-up hold’em by playing literally trillions of hands of poker against itself, while learning and memorizing optimal moves as it goes.

But the no limit variant is a vastly more complex proposition for artificial intelligence to master. A recent study placed fixed-limit hold’em 14th in a list of strategy game complexity, while no limit was 140th.

Imperfect Information

Because poker, unlike, say, chess, is a game of imperfect information, game theory algorithms developed by computer scientists can never quite be truly perfect; instead they must approximate a “perfect” poker strategy, which is defined as a strategy that cannot be exploited by any other counter-strategy.

Despite failing to trump humanity, Sandholm said he was thrilled by the results. “We knew Claudico was the strongest computer poker program in the world, but we had no idea before this competition how it would fare against four Top 10 poker players,” he said.

“It would have been no shame for Claudico to lose to a set of such talented pros, so even pulling off a statistical tie with them is a tremendous achievement.”

A.I. Will Catch Up With Us

“Beating humans isn’t really our goal; it’s just a milestone along the way,” Sandholm continued. “What we want to do is create an artificial intelligence that can help humans negotiate or make decisions in situations where they can’t know all of the facts.”

Sandholm believes the algorithms involved could have future applications in business, military, cybersecurity and medical arenas.

Bjorn Li finished with the biggest lead over Claudico, with $529,033, which earned him $44,676 from the $100,000 prize pool the casino and co-sponsor Microsoft put up for the players.

“We know theoretically that artificial intelligence is going to overtake us one day,” said Li. “But at the end of the day, the most important thing is that the humans remain on top for now.”

Philip Conneller
Written by
Philip Conneller
As part of the team that launched Bluff Magazine back in 2004, and then as Editor of Bluff Europe, Philip Conneller has (probably) written thousands of articles about poker and has travelled the globe interviewing the greatest players in the world, not to mention some of the sexiest celebrities known to man in some of the world’s sexiest destinations. The highlight of his career, however, was asking Phil Ivey (as a joke) how to play jacks, and emerging none-the-wiser. Philip once won $20,000 with 7-2 offsuit. He has been told off for unwittingly playing Elton John’s piano on two separate occasions, on different sides of the Atlantic Ocean. He became a writer because he is a lousy pianist. He lives in London where he spends his time agonizing about Arsenal football club, yet in Wenger he trusts.

Comments

Refriedrice wrote...

I’m pretty sure some of the developers at Deepmind, a new Google acquisition based in London, has algorithms that beat most players at NL Hold’em. It was explained to me thus- AI should always play optimally- and (most) humans will not. However it is flawed, in as much as it will monitor your style of play, and as long as you don’t change your style, you will be beaten. The trick is to vary your play- the suggestion given to me was to have favourite hands based on suits- that way you will play what the computer sees as equivalent hands differently. e.g. if you decide to bias spades, you will play 9s10s more often than , say, 9d10d, and so adding natural (unreadable) variation to your play.

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Did you know about our poker forum?

Discuss all the latest poker news in the CardsChat forum