Indiana Poker Programming Contest Computers Playing Poker

About

Why poker?

"They didn't call it Texas hold 'em at the time, they just called it hold 'em... I thought then that if it were to catch on, it would become the game. Draw poker, you bet only twice; hold 'em, you bet four times. That meant you could play strategically. This was more of a thinking man's game."
—Crandell Addington, World Series of Poker founder

Besides being entertaining, Texas hold'em poker is a game well-suited for such a coding contest. First and foremost, poker is accessible. Most people at least have some idea of the object of the game—to get the best five-card hand—and anybody can learn the game in a short lesson. In recent years, Texas hold'em has become very popular in the United States, replacing seven-card stud as the standard casino poker game. Television broadcasts of the World Series of Poker (WSoP) have brought poker into living rooms across the nation and made celebrities out of its champions.

More importantly, playing poker well is difficult. Of course, the probability of each type of hand can be computed mathematically using elementary combinatorics, but the rounds of betting (up to four rounds in hold'em) are what make poker challenging. A computer can memorize the exact probability of each combination of pocket cards, but can a computer bluff effectively or "read" their opponents? Can a computer strategize in response to their opponents' playing styles? Can a computer "condense meaning out of the vapor of nuance"? Can a computer think?

Chess is most often pointed to when pressed for examples of artificial intelligence. After all, Deep Blue's victory over Kasparov in 1997 drew world-wide attention—driving up IBM's stock in the process—and ignited fresh speculation on when computers would out-think humans. However, compared with hold'em, chess is a very complicated game. Players must consider up to thirty-two pieces on the board, each with complex interlocking relationships, and then make any of a hundred of possible moves. Deep Blue was a supercomputer with dedicated chess hardware that essentially brute-forced the problem, and still required humans to intervene between matches to beat Kasparov. In poker, a player can check/call (that is, meet the current wager), raise, or fold. Even a typical move database for chess is gigantic compared to the average poker hands database (with only 182 distinct combinations of pocket cards).

So, poker offers accessibility, challenge, an open-ended problem, and momentum in popular culture, making it an ideal choice for such a contest.

Who are we?

This contest is the creation of Eric Jiang, along with help from Carlo Angiuli, Jeff Catania, Dustin Dannenhauer, and Joseph Morwick. Our faculty advisor is Gregory Rawlins. We are sponsored in part by the Indiana University School of Informatics and Computing under the direction of Dennis Groth. We would be remiss to not mention the Annual Computer Poker Competition, whose poker software we are using, although they are not affiliated with nor endorse this competition.