A Modern Era for Poker
Like many industries once hesitant to embrace Artificial Intelligence and Machine Learning, the Poker industry has this last decade undergone a transformation, moving to the forefront of innovation and technology.
For most stakeholders, its futureproofing was welcome news and provided opportunity, stability and perhaps most importantly, increased credibility for the industry for both potential investors and participants.
A catalyst for this transformation was the shift from classical pot-odds and game theory- based strategy to one of big data and statistical models. Unlike many games such as chess and checkers, the decision tree for players of no-limit poker games (no restrictions on the amount a player can bet) is infinite, meaning there are infinitely many choices a player can make. Although the vast majority of possibilities aren’t ever taken in practise, formalising a winning strategy that stands up to all player types and skill levels was considered by most, impossible.
The Classic Strategy
From the late 1970’s to the late 2000’s, professional players such as “Texas Dolly” Doyle Brunson, published several popular books outlining winning strategies, developed from decades of experience playing the World Series of Poker (WSOP) and World Poker Tour (WPT). These strategies would exploit favourable situations whilst avoiding unfavourable ones as early as possible - with opponents having little chance to combat them. What many pros didn’t appreciate, however, was that although the games of that era were for all intents and purposes ‘solved’, poker was not. The games they participated in, shaped by the strategies and playing conventions of the time, were essentially a simplification of the raw game defined only by its rules. These players’ strategies were vulnerable to exploitation, especially by those who viewed the game through a different un-simplified lens using big data and statistical modelling.
By the mid-2010’s computer software packages called “solvers'' were readily available on the market and through a combination of hand-histories, probability theory and simulations, players were able to very quickly determine which aspects of their strategy were -EV (loss making), or which strategies were +EV (profit making) to employ in certain situations. The modern game became one of who could most effectively study, memorise and apply insights at the tables than one of psychological warfare and heuristics.
This shift in game strategy helped transform the industry as a whole in 3 ways:
The introduction of modelling and simulation kickstarted an industry of its own, one of research; software production; digital infrastructure; educational tools; chat forums and IP - lots of IP.
Players from all sorts of scientific, financial and gaming backgrounds who’d otherwise overlooked poker began to participate, especially in online environments where they could study and practice at home. The benefit of playing poker over other games for these people was obvious; if they became good enough, they were essentially paid to enjoy their own pastime!
Online poker providers were pressured and regulated into taking security extremely seriously, which also brought a raft of development efforts towards cyber security; cryptography; anti-cheating technologies and random number generation. The developers of the PokerStars client use quantum physics and lasers to shuffle the playing deck.
The poker industry has reinvented itself into being a modern hub of research and innovation.
Enter: Artificial Intelligence & Machine Learning
The game of poker and the challenge it posed to game strategists caught the attention of several academics and data scientists. Notably, in 2015 Professor Tuomas Sandholm of Carnegie Mellon University, Pittsburgh, pitted Claudico, an Artificial Intelligence program, against four of the best poker pros in the world. Sandholm was quoted:
“Poker is now a benchmark for artificial intelligence research, just as chess once was. It's a game of exceeding complexity that requires a machine to make decisions based on incomplete and often misleading information, thanks to bluffing, slow play and other decoys.”Tuomas Sandholm
Over the course of 13 days, Claudico fought Doug Polk and 3 others in a series of heads-up matches (one-on-one) at Rivers Casino. At the time, Doug Polk was considered the best heads-up no-limit (HUNL) player in the world.
The blinds for the match were 50 and 100 and each hand started with both opponents having 20,000 chips with those won or lost being tallied along the way. By the end of the match the professionals had won over 700,000 chips. In fact, the team won chips at a rate of over 7 big blinds per 100 hands (7bb/100), a resounding victory in the world of poker.
Despite their defeat, the academic challengers came back with a new and improved Artificial Intelligence in 2017 called Libratus, pitted against four of the worlds’ best players in a series of HUNL poker matches.
Programmed to train through each night of the 20-day challenge and learn from its missteps in a way akin to the pros, Libratus blew away its opponents netting over 1.1m chips at a rate of over 14bb/100. Dong Kim who performed the best, losing only 85,000 chips said;
“I didn’t realize how good it was until today. I felt like I was playing against someone who was cheating, like it could see my cards. I’m not accusing it of cheating. It was just that good.”Dong Kim
This is just one example of several staggering results produced by Machine Learning, with the AI producing a phenomenal exploitive strategy which swept aside its opponents, unmatched by human processing.
Applying the Lessons to Business
Artificial Intelligence and Machine Learning are pushing the boundaries of what we can achieve in terms of speed, efficiency and quality in many industries - Poker being just one example. The introduction of new processes can be daunting for some and technology sceptics often reference outliers capable of stumping infant algorithms, (as seen with Claudico in 2015), to prevent adoption. It’s clear to see with the victory of Liberatus, however, that Machine Learning is developing at an exponential rate, much to the gain of those who adopt it. Now in 2021, this has expanded to problem-solving within business.
Using this same practice SamsonVT utilises Machine Learning applications, specifically Anomaly Detection, to support their clients in turning asset data into smart insights that drive better decision making. The developed solution, SamsonBASE, recognises patterns of normal regimes for any equipment and automatically delivers alerts when anomalies are detected - before machine failure - preventing downtime, increasing productivity and enabling maximum overall equipment effectiveness (OEE).
To learn more about how Machine Learning can work for you, book a demo today by clicking here, or call our sales team on 0161 820 2115.