Inside PokerSnowie's brain reveals the work of Johannes Levermann, the head of research of Snowie Games. As the PokerSnowie "master trainer", Johannes explores first hand how the brain of PokerSnowie evolves and learns advanced strategic concepts, on its own.

PokerSnowie's ultimate aim is to produce the perfectly balanced game, find the ultimate un-exploitable equilibrium for all No Limit Hold'em configurations. Join us on this fascinating journey, which is just starting, into the future of poker.

The PokerSnowie Video Quiz series aims to answer the question: "what is the right play?" It is a set of poker coaching videos presented by French professional poker player Sharp. In each video, Sharp sets up an interesting hand in the "Scenarios" tool of PokerSnowie and explains how to analyse the situation and learn from PokerSnowie's advice, based on the Game Theory Optimal model.

New AI - Introducing the 25% Pot Bet Size

What's New?

At Snowie we strive to provide the best AI quality possible and are constantly trying to improve the PokerSnowie advice. Multi-player NL Hold'em is an extremely complex game and the AI will never be perfect. Nevertheless, with the latest AI update we believe we have taken an important step in the right direction. The new AI has trained many months more than the old one, and this fact alone should lead to a stronger and more consistent strategy. The main difference, however, is the introduction of a new bet size: 25% pot.

Why 25% Pot?

In principle, it would certainly be good to add a bunch of bet sizes, the more the better. If there are too few bet sizes, the strategic constraints for the AI become too big, and the resulting strategy may be far from the optimum. In practice, the situation is different though. The more bet sizes the AI has to deal with, the more difficult it becomes to balance the ranges for each bet size. This is why we chose to just add one bet size, which seems like a reasonable increase in complexity. Many players use bet sizes of between 50% and 100% pot, so introducing a new bet size of 70% or 75% would have made sense as well. We finally chose 25% as the new bet size to increase the overall bet sizing range of the AI and to make it possible for the AI to develop completely new strategies by employing min-raising and probe bets.

What is the result?

If this new bet size was too small to be effective, the AI would have just ignored it or employed it only rarely. Then the resulting strategy would have been very similar to the old AI. However, it turned out that 25% pot was a very good choice indeed. The AI likes this bet size and chooses it quite often as the optimal size. This is actually proof that the strategy has improved and that the new AI must be stronger than the old one. As a rule of thumb, the lower the bet sizes chosen in overall play, the higher the overall betting frequency. In most situations, the betting range for 200% pot is much lower than for 50% pot, for example. Using the new 25% pot frequently also means an overall more aggressive play with a higher betting frequency. In the next post, I'll present two examples where the new bet size makes a lot of sense.


Inside PokerSnowie's brain

Johannes Levermann is the co-founder and head of research of Snowie Games. Having previously contributed to the development of the world famous backgammon Snowie, he nowadays spends most of his time developing and fine-tuning the training algorithm of PokerSnowie.