Poker Players Go “All In” on Artificial Intelligence

Some of the world’s greatest poker players have a secret weapon up their sleeve these days: It’s artificial intelligence. Find out how AI is being used to create the optimal poker hand.

Poker is a notoriously difficult game to master.

But now, scientists have developed artificial intelligence (AI) that helps professional players learn more about optimal poker strategy, so they can become better players and win bigger in high-stakes games.

Professional poker player Seth Davis was playing in the World Series of Poker when he came to the end of a hand with no valuable cards and not a lot of options. With his $250,000 tournament entry fee (not to mention his pride) on the line, Davis decided to bluff and declared that he was “all-in” for the round.

To Davis’s relief, his opponent folded, and Davis won the hand. That night, he analyzed the hand using a program called PioSOLVER, an AI-based tool that has radically reshaped the way poker is played in the past few years. Within a few moments of inputting the details of the hand, the computer told Davis he had played the hand close to perfectly, and his bluff was the right move at that moment.

While AIs have had success beating humans at games like Go and chess, winning at poker is significantly more challenging due to the fact that an opponent's hand in poker is kept secret. This informational deficit increases the complexity of developing a program that can play poker.

Winning at poker requires intuition, strategy, and reasoning based on hidden information. But despite these challenges, artificial intelligence can now play poker (and win!)

How Researchers Developed Poker-Playing AI

In an article for Science magazine in 2015, researchers at the Department of Computing Science at the University of Alberta reported that they had developed a computer program that could calculate the best course of action for a variant of poker called Limit Texas Hold’em.

The same year, Polish computer programmer and online poker player Piotrek Lopusiewicz started selling the first version of his program PioSOLVER for $249. The program calculates solutions for no-limit poker, which is more complicated than Limit Texas Hold’em. PioSOLVER can help players remake their approach to the game by learning what works best in different situations. The program even figures out the best and worst hands to bluff with.

After the development of PioSolver, scientists at Carnegie Mellon University partnered with Facebook to develop Pluribus, a poker-playing bot that can beat some of the best Texas Hold’em players in the world.

Pluribus played 10,000 hands of poker against five other people who have earned at least a million dollars in poker, and the bot won an average of $480 from its human competitors for every 100 hands – which is similar to the earnings that professional poker players shoot for.

One of the reasons Pluribus is groundbreaking is because it plays against more than one person, and doesn’t just rely on game strategy to win.

With its solutions to complicated problems, Pluribus can be the gateway to solving some of the biggest problems in the world, like self-driving cars, fraud detection, drug development, and cybersecurity.

Some Players Love AI (and Some Don’t)

Although many players are thrilled with the recent development of AI to understand the optimal strategy for playing poker, there are some players who feel AI-based approaches have changed the poker landscape for the worse.

Higher-stakes competition has indeed gotten much greater with the advent of new solvers like PioSOLVER. Some players also say that AI changes the essential nature of the game.

Doug Polk, a retired poker player who raked tens of millions of dollars during his career, says, “I feel like it kind of killed the soul of the game, from who can be the most creative problem-solver to who can memorize the most stuff and apply it.”

Others believe the new generation of poker-playing AI tools is just a continuation of a long pattern of tech innovation in the game. For example, top players have used software to collect behavior data about their opponents and analyze it for weaknesses for many years.

Whether they’re playing online at home or at the World Series of Poker, players may have to do battle with a new contender in their favorite game: A program that can bluff as well as they can.