Pioneering Poker AI Research

During the past decades, given its direct relationship with real-world applications such as negotiation, auctions, cybersecurity, and more, poker has served as a challenging problem for the fields of game theory and artificial intelligence (AI). Indeed, no other popular recreational game captures the challenges of hidden information as eloquently and effectively as poker.

The poker industry has always lagged behind its scientific counterparts when it comes to implementing academic technological advances in commercial products.

Research in those areas led to the development of powerful algorithms, like Counterfactual Regret Minimization (CFR), the core pillar behind poker solvers. However, since the publication of CFR in 2007, the poker industry has always lagged behind its scientific counterparts when it comes to implementing academic technological advances in commercial products. Indeed, it took eight more years after the publication of CFR for the poker community to have access to a poker solver, and many more to see widespread adoption.

Thanks to the introduction of GTO Wizard AI, we are proud to say that we’ve taken a step towards bridging the gap between academia and industry by democratizing years of scientific research and experimentation to a larger audience.

We believe that fostering a culture of collaboration between the scientific community and the poker industry is essential for driving innovation.

Contrary to the traditions of past years, we believe that fostering a culture of collaboration between the scientific community and the poker industry is essential for driving innovation, solving complex challenges, and shaping the future of our ecosystem. In our efforts to lead the charge, GTO Wizard has demonstrated its commitment to the academic community through various initiatives, including giving lectures, connecting with leading researchers, and supporting academic endeavors.

Giving Back to the Community

GTO Wizard AI team members recently shared their expertise with Massachusetts Institute of Technology (MIT) students by delivering a lecture on the current landscape of poker AI research, its future, and how its adoption in study tools like GTO Wizard is revolutionizing the industry. We also had the privilege of connecting with Noam Brown, a world-class researcher renowned for his groundbreaking contributions to poker research, such as Pluribus, the first poker AI to beat top pro players in 6-player No-Limit Hold ’em. Our interaction with Noam Brown and MIT students has been invaluable, offering insights that have propelled our research and development efforts forward.

In our initiative to connect with academia and empower the talent of tomorrow, we were thrilled to be one of the main sponsors of MIT Pokerbots. This month-long class challenged students to apply their mathematics and computer science knowledge to develop a poker agent. In this isolated academic setting, students were tasked with creating agents capable of competing against agents from other teams in a specialized poker variant.

This year’s variant was Auction Hold ’em, a No-Limit Texas Hold ’em variation that introduces an exciting twist to the game dynamics. In Auction Hold’em, a bidding round, known as ’the auction,’ is added immediately after the dealing of ’the flop’ cards. During this unique phase of the game, players have the opportunity to bid for the chance to receive a third hole card, which can potentially strengthen their hand. What makes Auction Hold ’em particularly captivating is its implementation of a sealed-bid second-price auction, where players submit bids without knowing the bid of other participating players in the auction, and the highest bidder wins the right to receive a third card but only pays the second-highest bid amount. However, if players submit identical bids, they all receive a third card. The bid amount is then added directly to the pot, enhancing the stakes of the hand.

To foster interaction with students and as one of the event’s main sponsors, our team spent five days working on our own entry to the tournament, which we made open-source here.

Looking Ahead

The team behind GTO Wizard AI brings together a diverse group of talented individuals from across the globe. From programmers who showcased their knowledge on the international stage of programming competitions to the author of an open-source solver and alums from world-renowned deep learning labs, our team has proven the effectiveness of combining neural networks with traditional solving algorithms.

We can accelerate the integration of revolutionary features into our application by partnering with academic institutions.

As a pioneering entity advancing theoretical poker AI research and pushing its boundaries, we are on a path to change how poker is being studied by hundreds of thousands of players worldwide. While our efforts to enhance our solver’s capabilities—extending its reach to solve any ICM and PKO structure and enabling custom multiway solving—are well underway, we can accelerate the integration of revolutionary features into our application by partnering with academic institutions.

In our quest to solve any poker variant in a few seconds, we are always on the lookout for exciting collaborations and fantastic talent. Whether you are a research lab looking to apply its algorithms to larger games, or a researcher wanting to push the boundaries of poker research, don’t hesitate to reach out at work(at)gtowizard(dot)com.