Machine Learning in the Casino: Why businesses should care about winning at poker

by | 2 April 2019 | Basics

Machine Learning is at the centre of the competition between humans vs. Artificial Intelligence (AI). About two years ago, there was another major breakthrough on the machine side. In this blog article, we explain why this date is significant, what machine learning is exactly and why this method is so important for the economy.

The Poker AI Libratus was able to beat several champions in the poker game No-limits Texas Hold'em. This news came for two reasons Sensation the same. On the one hand, because the Bluffing is a key skill to win the game. Secondly, because winning requires the ability to take deliberate risks. Libratus even played so well after a certain point that his opponents felt he could see their cards. Today, the poker bot has a new job where he uses the skills he trained in the game for a completely different purpose. But from the beginning.


By loading the video you accept YouTube's privacy policy.
Learn more

load Video

Libratus was developed and trained with the supercomputer at the Pittsburgh Supercomputing Center, which is operated by Carnegie Mellon University, among others.

The real surprise in the victory of the poker AI was: it was essentially based on the use of machine learning

On the other hand, Libratus' victory was a minor sensation for another reason. Until now Deep Learning and Neural networks as the beacon of hope in the development of artificial intelligence. Unlike AlphaGo, however, Libratus' success is not based on Deep Learning or neural networks. So how did Libratus manage to teach poker?

Reading tipIn this series of articles we show different Machine Learning Methods on.

The surprising (and slightly abbreviated) answer to this question is: Machine Learning. Although Deep Learning is also a form of machine learning, the creators of Libratus relied on relatively conventional forms of machine learning. All in all, this is reason enough to explain this concept in more detail and to explore the economic potential of the Data Science Projects to illustrate.

Reading tipIn our basic article we explain all the common concepts that exist around the topic of KI gives.

What is Machine Learning?

The most general, non-technical formula to describe machine learning is: A self-optimising Algorithmwhich it is possible to learn from experience. If we stay with the poker bot Libratus, the functioning of this ability could be observed in real time during the poker tournament. The decisive indication was that the longer the poker tournament lasted, the better Libratus was able to assess his opponent and develop successful strategies.

Machine learning are fundamentally different from those we humans use. In this respect, the idea that Libratus can play poker in a similar way to humans is also wrong. To win at poker, humans need first and foremost Two essential competencesMathematical or statistical talent and knowledge of human nature. In comparison, machines develop their own, completely different approaches and strategies.

The Libratus learning method: Counterfactual Regret Minimisation

Libratus, on the other hand, can win at poker because he statistical procedures in Perfection mastered. While still playing, he makes new assumptions and checks them against the respective results. If the partial result and the assumption do not match, the Strategy varied and adapted for future moves.

The underlying learning method is called "Counterfactual Regret Minimisation"(roughly: "Subsequent minimisation of regret"). After each move, the AI returns to the previous decision, evaluates it and rates it according to how high the "regret" is. Libratus has asked itself a question about this countless times: How much better or worse would the game results have been if I had made different decisions?

In order to grasp the enormous potential that machine learning has for the economy, however, it is necessary to describe the method in even greater detail. Machine learning as a method is, however, much more complex than this exemplary description in the case of the poker AI Libratus.

The whole truth about Libratus' success and why companies should care about poker

As already indicated, the whole truth about Libratus' success is a little more complicated. This is because Libratus' capabilities are not "just" based on machine learning or reinforcement learning. Rather, the poker bot consists not just of an intelligent algorithm, but of a System of a total of three different machine learning methodswho work together. Libratus is considered an important milestone, among other things, because it masters a game in which a large part of the Relevant information concealed remains.

In this context, games are a Key to reality. Intelligent software that masters a particular game can, in principle, also master challenges in the real world. Strategic thinking in situations where not all information is known, but is not only relevant in poker. One example is pricing algorithms that automatically determine the optimal price customers are willing to pay for a certain product or service. Incidentally, the poker AI Libratus itself now works for the Pentagon, where it helps optimise diplomatic negotiations and war strategies.

Machine learning is a key to turning data into economic value added

There are also numerous areas of application for this ability in economic contexts and in everyday life. Negotiation situations or challenges in the field of cyber security can be understood as a game with partially hidden information. It does not always have to be an AI of the calibre of Libratus. Less elaborate programmes based on machine learning can also be used to Improve decision-making processes and achieve economic success.

You can find out more about machine learning in the CRISP Research Study!

<a href="" target="_self">Michaela Tiedemann</a>

Michaela Tiedemann

Michaela Tiedemann has been part of the Alexander Thamm GmbH team since the early start-up days. She has actively shaped the development from a fast-moving, spontaneous start-up to a successful company. With the founding of her own family, a whole new chapter began for Michaela Tiedemann at the same time. Hanging up her job, however, was out of the question for the new mother. Instead, she developed a strategy to reconcile her job as Chief Marketing Officer with her role as a mother.