New breakthrough in poker AI
This is another breakthrough in poker AI research. A new poker bot beat a line-up of 5 elite poker players at a time in a session of 6-handed No Limit Hold’em cash game.
The poker bot who was named “Pluribus” is a product of the collaborative efforts between the Facebook AI department and Carnegie Melon University.
Computer science professor Tuomas Sandholm and his Ph.D. student Noam Brown lead the project on CMU’s part. They have quite some experience with making poker playing bots.
Libratus Beats Poker Pros Heads-Up
In January 2017, Prof. Sandholm’s creation named Libratus managed to beat a 4-player team of high stakes poker pros in a series of heads-up matches.
Before that successful experiment, the professor called poker “the last frontier of game-solving in AI”. The reason is that unlike chess or go – games that have already been solved by computers – poker is a so-called “imperfect information game”.
Imperfect information game means that the player has access to some information their opponents don’t. In poker’s case, it’s the hole cards.
This January, we reported that Libratus the poker bot was actually used by the United States military on some projects.
Poker Bot Pluribus beats elite poker pros including LLinusLLove and Nick Petrangelo
This time, the researchers conducted two experiments. They had one bot play against 5 human players. They also had one person take on 5 bots at the same time.
In both cases, the bots managed to win the sessions by statistically significant margins.
In the test, 13 poker pros took turns playing against the AI software. The participants included Linus “LLinusLLove Loeliger, and 2-time WSOP bracelet winner Nick Petrangelo.
Jason Les and Jimmy Chou were also part of the team that eventually lost to Libratus two years ago.
This time, Pluribus beat the human side by $5 per hand on average. The game was played on $50/$100 stakes.
Noam Brown wrote this in a blog post on the Facebook AI website about the game of poker in general:
“No other game embodies the challenge of hidden information quite like poker, where each player has information (his or her cards) that the others lack. A successful poker AI must reason about this hidden information and carefully balance its strategy to remain unpredictable while still picking good actions.”
Poker Pros opinions on Pluribus
Many mainstream outlets outside the poker-specific media covered the story. Those include The Times, the BBC, The Verge and the MIT Technology Review.
Within the poker rounds, Doug Polk had this to say about the issue:
“There is one important note, though. The bot has to play at a stack depth that is the same as its own. So essentially, it’ll be very strong when it’s the shortest stack, but as it gets deeper it would have to reset down to a smaller amount of big blinds. So not exactly perfect multi-handed AI at this point, that would be a major leap when you’re playing online today.”
Yes, the rapid improvements in poker bot technology can be unnerving for many online poker players.
Already we’ve seen cases where major poker sites issued refunds to their players after it was uncovered that cheating bots were playing at their tables. Some fear that with the ongoing developments the online bot situation will only worsen…
Get the best partypoker rakeback deal on the internet!
SIGN UP through us with partypoker and earn up to 60% rakeback on partypoker via their Diamond Elite Club plus take part in our high-value VIP-Grinders Promotions:
- $30 Welcome Bonus in Spins Tickets
- Up to 60% Cashback via the Diamond Elite Club
- 4 x $250 Private Poker Freeroll per month
- Exclusive Party & Bwin Chase
Terms and conditions apply. New customer offer and 18+ only. Should you require help regarding your betting pattern, please visit www.begambleaware.org.
