The ai

Extremely solid play

In the learning process, bad moves and very unsuccessful betting lines that cost a lot of EV have a big impact on the neural network. At all costs the neural network tries to avoid playing blunders (big errors) and will adapt accordingly. As a result, PokerSnowie is very aware of situations where one certain action is absolutely right or where one certain action absolutely must not be played. PokerSnowie will not make bad calls or raises out of despair, and is not susceptible to 'tilt'. It plays a very disciplined game and does not give away EV to its opponents easily.

This alone is a big advantage over human players. Even pros make big errors quite regularly. Sometimes they play on too many tables at the same time, ending up unable to concentrate 100% on every single hand. Other times they make their decisions in the few seconds available, and only upon longer reflection would they agree with PokerSnowie's advice.

Very balanced play

PokerSnowie has to defend itself against exploiting agents all the time. Each time a part of PokerSnowie's strategy is not correctly balanced, it is punished by the agents. PokerSnowie cannot focus on a single hand and how to play it; it has to construct proper hand ranges for every game decision. In any situation, for example, the fold percentage over the whole hand range should not be so high that opponents could just run over PokerSnowie and profitably raise with garbage. At the same time, PokerSnowie should avoid paying off too much to the opponents' strong hands. The same balance is needed for bets and raises: if the ratio between bluffs and values is not correct, a hefty punishment is the reward.

We think it's safe to say that PokerSnowie plays an extremely well balanced game.

Difficult to read

There are two factors concerning why PokerSnowie is so difficult to read. It keeps you guessing all the time: does PokerSnowie really hold a strong hand or will I end up folding the best hand?

First of all, there is no tell given by the bet size. The same bet size is used for the whole hand range. Whether the actual hand is a bluff or a monster has no effect on the bet size.

The second factor is PokerSnowie's balance. If PokerSnowie checks, it does so with a range that leaves you in the dark. If it bets or raises, it holds a good mixture between bluff and value hands.

Learned all conceivable strategies

PokerSnowie has the freedom to learn whatever strategy it can come up with. There are no limitations given by expert rules or anything else. As it turns out, PokerSnowie has successfully learned all theoretically known concepts and strategies, and has built up an impressive repertoire of strategic weapons! Here is an incomplete list of learned strategies, some of which will be discovered and explained in further articles

  • Bluffing: an obvious concept for PokerSnowie. As bluffing often works, it was one of the first things PokerSnowie learned.
  • Check-raising: a basic principle that can be well applied if the checking range is correctly balanced.
  • Slow-play: knowing that fast play often scares away the opponent, PokerSnowie sometimes slow-plays strong hands.
  • Semi-bluffs: if there is a choice of hands to use for bluffing, PokerSnowie always prefers hands with outs to improve to the best hand and makes semi-bluffs instead of pure bluffs.
  • Inducing bluffs: PokerSnowie checks more often than most humans and so creates bigger pots when the opponent is weak.
  • Controlling the pot size: out of position with medium hands, PokerSnowie often plays passive, but therefore maintains good control of the pot size.
  • Implied odds: PokerSnowie is aware of implied odds and sometimes prefers hands with a lower winning percentage but higher implied odds.
  • Positional awareness: full ring game pre-flop is quite a positional game, and PokerSnowie shows quite drastic strategy changes depending on the position.
  • Floats: in certain situations, PokerSnowie calls with surprisingly weak hands. The reason is that many times a bluff on the next street will be successful.
  • Bet sizing: PokerSnowie adapts the bet sizes according to the exact situation. Different sizes are chosen if PokerSnowie sits in or out of position, if it's a probe bet or a bet that is meant to show strength, etc.
  • Donk bets: in general PokerSnowie doesn't like donk bets very much, but it found certain types of situations where they should be applied.
  • Blocking bets: PokerSnowie has even learned this rare concept. While normally a bet is either a value bet or a (semi-)bluff, in some occasions PokerSnowie plays a blocking bet with a medium strength hand, to avoid getting bluffed out of the hand and to set the bet size
  • Card removal: especially on pre-flop, PokerSnowie has learned that the chances of being re-raised also depend on the cards held. Therefore it likes to use hands with blockers for bluff-raising (like A5s, which reduces the chances that the opponent holds AA or AK).


As you can see, PokerSnowie doesn't play perfect poker, but PokerSnowie's weaknesses cost very little EV in terms of overall performance. With PokerSnowie's strengths covered here, we believe that PokerSnowie is not only by far the most advanced artificial intelligence-based software for poker, but most likely also comparable to the world's best poker professionals.

PokerSnowie's balanced poker style works well against all kinds of opponents, from low stakes up to nose bleed stakes. Beginners and professionals alike can benefit greatly by trying to understand PokerSnowie's strategy and the underlying concepts.