Inside PokerSnowie's brain reveals the work of the Snowie AI Team. It 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.

PokerSnowie Video Quiz: Facing a 4bet preflop with Ax suited

preflop -  facing a steal from the cutoff

In Today’s PokerSnowie Video Quiz, Sharp goes through a preflop situation and reviews PokerSnowie's advice in light of the preflop theory.

Hero did 3bet light with A5s from the BB and faced a steal from the cut off.

What is to be done when facing a 4bet?

  • Fold, hero was bluffing in the first place
  • Call to see a flop, the odds are too good to fold
  • 5bet for half pot, the cut off might be bluffing
  • 5bet all in, hero is committed to the pot and should maximize his fold equity.

Watch the video to find out the surprising answer from PokerSnowie.



PokerSnowie Video Quiz

Sharp is a French maths postgraduate and a probability engineer turned professional backgammon player in 1999. French number 1 for 4 years, having won all the largest Backgammon tournament titles, he moved on to Poker in 2004. He has since then built an impressive record and stands amongst the best French MTT players. Nowadays Sharp spends most of his time coaching and producing coaching videos, in particular for, where he is one of the lead instructors.



Hi Sharp. Nice video. Against a very straightforward lower limits player, where they are typically only 4-betting their premium hands pre-flop, would you make adjustments for that, or still try to play a GTO strategy?
Hi, I never really played at low stakes, so it is difficult for me to answer this question. But I think you are right. A 5bet bluff is a high variance move and you have to be sure that villain is 4betting bluff with a high frequency before risking an extra 34BB. Even with a good win rate, you would need a lot of hands to get them back. Biggest wins in poker are not coming from bluffs. They are coming from your ability to build big pots with big hands and get pay off. At low stakes one doesn’t need to be balance to be a winner, because most of the opponents will play according to their own hand and not to his range. Bluffing is of course still a big part of the game, but I would not advise to run big and costly ones at low stakes.
Thanks Sharp, that's good advice. I'm coming from SNGs to cash games, so it's a good reminder of some of the differences and subtleties.
This was really illuminating. The idea of 5-bet bluffing at 100bb seems so weird to someone like me that grew up with the PNLH idea of pot-commitment, but I *think* you've convinced me that it's correct. ;) Thanks a lot for your videos, Sharp!
Hello, yes I think the 5bet click back is a great weapon in practice both for value and as a bluff. For value since it’s leaving room to our opponent to do something stupid, while he would fold his bottom range on a shove. As a bluff since it’s risking less and leveraging our opponent stacks. The same trend occurred in MTTs, a couple years ago, everybody agreed that with 10BB or less it was wrong to open and fold so any hand played were opened shoved. Today many good players open bluff with 8 blinds with the plan to fold to a raise.
Hi, its good case to show how exploitable Snowie is at the moment. It reccomend 4bet 44bb and fold to 100bb shove after that 76% of time??! CO can print money just shove any2 after Snowie's 4bet from BB. Too many bluffs in Snowie's 4bet range that it folds to shove. Tell me if I'm wrong with that math
Hi and thanks for being such a fan. I agree with the math but disagree with the conclusions. First before 6 betting shove villain has to get there. And you will agree that if it will win money when Snowie is 5betting bluff, opening any2 and then 4betting them cannot be a winning strategy overall. Beside the joke, in my opinion, when Snowie 5bets clicking back, the program is not trying to be balanced it is exploiting the 4bettor. According to Snowie in this situation we are no more in a GTO model. In GTO the cut off should have 4bet pot and not to 22BB in the first place. Snowie’s estimations are based on pattern recognition. From experience it knows the cut off is very polarized, and Snowie seizes a great opportunity to bluff him back with very good odds and the maximum leverage. I am not interested in printing money against Snowie, since it is a tough player and it wouldn’t pay if I win. But I am very interested to learn from Snowie how to exploit other player’s leaks in their game. And I think this trick can be printing money against much kind of players at mid stakes or higher when they sit on the cut off.
Hi Sharp, thx for your detailed answer. You prompt me to check the spot with a standart Snowie's bet sizing... and i was confused even more. I used for CO $7 - $44.5 - $200 (AI) and for BB $22-$89.5 Its almost the same sizing as in the original case but Snowie's recommendations changed drastically: - call push 63% times (was 24%) and BB EV for min 5bet A5s is 0.00 EV (was +9.14), that is way more credible and balanced. Also, according Johannes Levermann's interview, Snowie wasn't trained on real HH or in play with human players, that is why it cant 'exploiting' mid stakes field. My assumption is that Snowie's advices are close to GTO if all players in a hand use fixed betsizes (0.5, 1 or 2 PSB) and could be very wrong if players use 'strange' betsize. And i still cant understand how it is possible to train Snowie for every possible bet size as soon game tree in that case should be huge and for any particular game state hard to get enough samples to balance strategy.
Hi, This is probably more a question for Johannes Levermann, but I will do my best to answer it with my little knowledge and my strong opinions. NLHE is too big of a game for the bots to train with every possible bet-size. I think the 3 sizes Snowie team had chosen are great to cover most of the situations and that Snowie should be able to adjust quickly his evaluations to those small betting variations. I read somewhere team Snowie will introduce an intermediate amount between pot and half pot for developing the next PokerSnowie. I don’t think it would change a lot his way of playing. I personally would much more prefer a small blocking bet size like quarter pot or even less, which make sense in theory on some turn and river situations when pots get big. I don’t know if Snowie wouldn’t be able to exploit mid stakes field. I would not be scare to bet on it at a 6max table. And I think for sure we can learn a lot from Snowie to exploit mid stakes field.
While I'm willing to entertain clicking back in position, I think clicking back out of position is suicide for a few reasons 1) at mid-stakes heroes are 3betting linear ranges and villains are 4betting to 2x where heroes are forced to either 5bet all in with a linear range, 5bet all in with a tighter range and call 4bets with a middling range or 5bet fold and call 4bets with a middling range as possible strategies i.e. 5bet AQ+, 88+, AJs, KQs all in or 5bet AK, 88+ all in and call AQo, AJs, ATs, KQs, KJs, KTs, QJs, QTs, JTs and then decide whether or not to 5bet less than all in to create a 5bet fold range of, let's say, KQo. I think Snowie is assuming that 4betting less than pot is a mistake when it's not clear whether or not it is, like wise I think Snowie is assuming that villain's range is polarized when in fact as 4betting sizes decrease and villains start calling 4bets OOP the likely hood is that 4betting AQo, ATs or 99-88 type hands have value, so it's possible we either have snap calls with AQo, ATs in our range or snap 6bets with 99-88 in our range where the dead money he's laying turns hands we are essentially 4bet/folding for value vs a 5bet all in into calls and 6bets and that's assuming we don't just 5bet bluff off A5s. So this doesn't seem like a GTO or balanced strategy and Snowie is making a lot of assumptions about range composition that it thinks are accurate because it's bound by its understanding of bet sizing and range construction at set increments and as soon as people stop 3betting and 4betting to pot it goes wonky because it's seeing polarity where there isn't any. If anything, it's really interesting tho' that villain can't take advantage of those pot odds with the bottom of his range to call because either his Ace is dominated or his bottom pairs and gut shots are -EV to call with considering we're going to shove the flop so it seems like clicking back creates better leverage vs ranges that must be polarized as long as we're not exploitable ourselves.