Main Menu
Top of Page
Top of Page
Top of Page
Top of Page
  | Home   | Index   | Info   | This Week   | Poker   | News   | Email

Welcome to the News desk.

Poker Bot Libratus crushes Professional Poker Players 01/02/17
• Astounding win rate of 14.7 big blinds per 100 hands

Artificially intelligent, poker-playing software developed at Carnegie Mellon University, challenged some of the game's best human players to a rematch over a 20 day tournament that involved 120,000 hands. Libratus, a computer program developed by Tuomas Sandholm, professor of computer science at CMU, and Ph.D. student Noam Brown, took on 4 poker pros at Rivers Casino on Pittsburgh's North Shore, Jan. 11-30, 2017 for $200,000.

The competition, called Brains vs. Artificial Intelligence: Upping the Ante, took place between the new AI poker bot Libratus (latin: 'balanced') and Jason Les, Dong Kim, Daniel McAulay and Jimmy Chou, considered to be 4 top professionals.

CMU's computer software lost to four professional players during the inaugural Brains Vs. Artificial Intelligence poker tournament in 2015. The 80,000 hands played against a computer program named Claudico weren't enough, however, to establish human or computer superiority with statistical significance.

For this rematch the four players were split into two pairs, one pair playing in the open and the other in a separate room with no external communication. The hands that Libratus received in the open match were played by the humans in the hidden match to reduce variation based on the quality of the cards dealt.

The humans teamed up at the start of the tournament with a collective plan of each trying different ranges of bet sizes to probe for weaknesses in the Libratus AI’s strategy that they could exploit. During each night of the tournament, they gathered together back in their hotel rooms to analyze the day’s worth of plays and talk strategy.

The human strategy of playing weird bet sizes had its greatest success in the first week, even if the AI never lost its lead from the beginning. Libratus held a growing lead of $193,000 in chips by the third day, but the poker pros narrowed the AI’s lead by clawing back $42,201 in chips on the fourth day. After losing an additional $8,189 in chips to Libratus on the fifth day, the humans scored a sizable victory of $108,775 in chips on the sixth day and cut the AI’s lead to just $50,513.

But Libratus struck back by winning $180,816 in chips on the seventh day. After that, the “wheels were coming off the wagon” for the human poker pros, Sandholm says. They noticed that Libratus seemed to become especially unbeatable toward the last of the four betting rounds in each game, and so they tried betting big up front to force a result before the fourth round. They speculated on how much Libratus could change its strategy within each game. But victory only seemed to slip further away.

Sadly for the humans the machine kept on learing and getting better. Late each day, after the poker play ended, Mr. Brown connected Libratus to the Pittsburgh Supercomputer Center’s Bridges computer to run algorithms to improve its strategy overnight. In the morning he would spend two hours getting the newly enhanced bot back up and running.

In the end Libratus was ahead $1,766,250 and with big blinds of $100 that translates into 14.7 big blinds per 100 hands, a crushing result in any analysis.
Player Position Result
Dong Kim 1 -$85,649
Daniel MacAulay 2 -$277,657
Jimmy Chou 3 -$522,857
Jason Les 4 -$880,087
The prize money of $200,000 was shared exclusively between the human players. Each player received a minimum of $20,000, with the rest distributed in relation to their success playing against Libratus.

Despite the historic victory over humans, AI still has a ways to go before it can claim to have perfectly solved heads-up, no-limit Texas Hold’em. That’s because the computational power required to solve the game is still far beyond even the most powerful supercomputers. The game has 10160 possible plays at different stages, more than the number of atoms in the observable universe, which is around 1082.

However as the tournament came to an end many a viewer on the Twitch live-stream expressed their concern with comments such as “Dude poker is dead!!!!!!!!!!!” before adding “RIP poker.” Brown tried to reassure the Twitch chat that invincible poker-playing bots probably would not be flooding online poker play anytime soon.

“People are worried that my work here has killed poker: I hope it has done the exact opposite,” Sandholm said. “I think of Poker and no limit [Texas hold’em] as a recreational intellectual endeavor in much the same way as composing a symphony or performing ballet or playing chess.”

But for the Carnegie Mellon University staff this was not just about poker. The algorithms that power Libratus aren’t specific to poker, which means the system could have a variety of applications outside of recreational games, from negotiating business deals to setting military or cybersecurity strategy and planning medical treatment – anywhere where humans are required to do strategic reasoning with imperfect information.

“Poker is the least of our concerns here,” said Roman V Yampolskiy, a professor of computer science at the University of Louisville. “You have a machine that can kick your ass in business and military applications. I’m worried about how humanity as a whole will deal with that.”

For Brown, Libratus challenges preconceptions about machine intelligence versus human intelligence.

“People have this idea that poker is a very human game and that bots can’t bluff, for example. That’s totally wrong. It’s not about reading your opponent and trying to tell if they are lying, it’s about the cards and probabilities,” he said.

Most online poker players have nothing to fear from Libratus right now. The system only works in Heads Up poker, where only two players are involved. A game with three players or more would be too computationally intensive, and require a totally different strategy and algorithmic approach.

Note : Throughout the competition, CMU said Libratus recruited the raw power of approximately 600 of Bridges' 846 compute nodes. Bridges' total speed is 1.35 petaflops -- approximately 7,250 times as fast as a high-end laptop -- with its memory coming in at 274 terabytes, about 17,500 as much as you’d get in that laptop.