The New York Times The New York Times Technology July 10, 2003



NYT Store
Photo: An early computer, the "mechanical mind" developed at MIT, 1927.Photo: An early computer, the "mechanical mind" developed at MIT, 1927.
Price: $195. Learn More.

Real Estate
Sign up for E-Mail Alerts!
Receive the latest property listings in your inbox...

Sell or Rent Your Home
Post a property listing on…

Find a Mover
Get instant quotes for full-service, self-service and last-minute moves...

Get Mortgage Quotes
Get instant mortgage quotes and calculate payments...

The New Card Shark

Artificial intelligence programs that can beat human opponents at poker are the focus of Darse Billings, a doctoral researcher at the University of Alberta.
Ian Jackson for The New York Times
Artificial intelligence programs that can beat human opponents at poker are the focus of Darse Billings, a doctoral researcher at the University of Alberta.

Email This Article E-Mail This Article
Printer Friendly Format Printer-Friendly Format
Most E-mailed Articles Most E-Mailed Articles
Reprints Reprints
Single Page Format Single-Page Format
Sponsored by Sprint

Subscribe to Circuits
Sign up to receive a free weekly Circuits newsletter by e-mail, with technology news and tips and exclusive commentary by David Pogue, the State of the Art columnist.

Computers and the Internet
Personal Finances
Create Your OwnManage
Most Popular AlertsTake A Tour
Click Here to Subscribe

Enlarge This Image

VICTORY - Chris Moneymaker celebrating after winning the World Series of Poker in May. He had never before played in a live, no-limit poker tournament.

HOME GAME - Gautam Rao, a skilled Canadian player, stopped going to casinos after his daughter was born "because of the smoke and distance'' and has turned instead to online play.

(Page 3 of 3)

"It understood the math perfectly," Mr. Rao said. "It knows the value of any holding at any point in time, allowing it to make proper calls." For example, many poker players consider 10's and jacks to be relatively low cards with little winning potential, but game theory suggests that they can still win hands in some situations with few players. The bot used knowledge like this to good effect. "It will win its share of pots," Mr. Rao said. "If you think you can over-aggress it, you will lose."

After several thousand hands, Mr. Rao shifted away from aggressive play. Instead of raising early and forcing the bots to react to his bets, he hung back more often to learn from their actions. This gave him more control, and he won more frequently.

Mr. Rao said that experience taught him a lesson. "Whenever you enter a new situation, don't assume you can execute a strategy that will win,'' he said. "Be quiet and listen instead of presuming or assuming."

Peter Muller, a friend of Mr. Rao's who has played against the same bot, said the approximations in the game-theory model left a weakness and limited the bot's chances to do more than break even. Game-theory models usually assume that every player uses the best possible strategy, something that rarely if ever happens with humans.

"An optimal game theoretic strategy might ensure that you don't lose, but it won't be effective at exploiting an opponent's weaknesses," Mr. Muller said. "The best players learn how to exploit predictability, but don't do it often enough so that the opponents catch on."

Mr. Billings is working on giving the next generation of bot the ability to track the behavior of an opponent and adapt to his moves. He believes that the foundation of game theory gives the bot the ability to manage losses, a crucial skill for winning in the long run.

All of this knowledge can have a downside, because analysis can kill a game as easily as it can a joke. Games like tic-tac-toe are well understood and therefore rarely satisfying. While poker is far from being understood at the same level, the deeper knowledge of a broader range of players is squeezing the margins for everyone.

Mr. Wilson spoke almost wistfully about the days in the early 1990's when card rooms first opened in California and began offering Texas Hold'em. "The games were crazy and loose," he said. "The games were wild. There were very few people who understood the game. Then the real turkeys ran out of money and stopped playing. They got smarter and started understanding the more subtle areas of the game."

When the skill is more even and well distributed, the effects of chance grow stronger, leading to more turbulence in the game. No one knows this better than Robert Varkonyi, the unknown who surprised everyone by winning the World Series of Poker in 2002. This year he was eliminated the first day.

<<Previous | 1 | 2 | 3

newspaper Expect the World every morning with home delivery of The New York Times Newspaper.

Click Here for 50% off

Advertiser Links

$7 Trades, NO Inactivity Fees

Click for a free Rauch report on Oracle9i Database Release 2

. Tools for Learning When to Hold 'Em (July 10, 2003)
. Technology Briefing | Software: International Game To Buy Acres Gaming  (July 1, 2003) 
. Technology Briefing | Internet: MGM Mirage Will Close Online Gambling Unit  (June 6, 2003) 
. E-Commerce Report; Congress wants to put more restrictions on online gambling, and the sites look for ways around the rules.  (March 31, 2003)  $
. HORSE RACING; Real Winners Of Pick Six Are Waiting  (January 12, 2003)  $
. BackTalk; Conditions Favorable For Point Shaving  (December 22, 2002)  $
Find more results for Computers and the Internet and Gambling .

. The New Card Shark
. How It Works: By Land and Air, a Live View of Cycling's Big Race
. Yahoo Profit Rises in Second Quarter
. What's Next: And Now, the Weather Report for Your Neighborhood
. Microsoft to Award Stock, Not Options, to Employees