Podcast: Play in new window | Download
The Computer Poker Research Group’s Michael Johanson and Dustin Morrill discuss Andrew and Nate’s recent battle with their heads-up no-limit hold ’em AI, DeepStack.
Johanson has previously been our guest on Episode 79 and Episode 110. Before the match, we discussed DeepStack with Michael Bowling on Episode 208.
Follow @DeepStackAI on Twitter for details of their upcoming Twitch matches as well as links to replays of past matches and other news.
Strategy
PokerStars – $150/$300 NL (2 max) – Holdem – 2 players
Hand converted by PokerTracker 4: http://www.pokertracker.com
Hero (SB): 69.67 BB
Hero (BB): 63.67 BB
Hero posts SB 0.5 BB, Hero posts BB 1 BB
Pre Flop: (pot: 1.5 BB) Hero has 6d 6c
Pre Flop: (pot: 1.5 BB) Hero has Qs Qd
Hero raises to 2 BB, Hero raises to 6 BB, Hero calls 4 BB
Flop : (12 BB, 2 players) 7s Th 5c
Hero bets 7 BB, Hero calls 7 BB
Turn : (26 BB, 2 players) 6s
Hero bets 14.17 BB, Hero raises to 50.67 BB, Hero calls 36.5 BB and is all-in
River : (127.33 BB, 2 players) 9h
Hero shows 6d 6c (Three of a Kind, Sixes)
(Pre 19%, Flop 13%, Turn 95%)
Hero shows Qs Qd (One Pair, Queens)
(Pre 81%, Flop 87%, Turn 5%)
Hero wins 127.33 BB
Details of the 85o shove, as well as further thoughts on it and the match in general, are here.
About the low but non-zero percentages given to various non-folding options with 85o in the pre-flop spot and the question of whether or not it was just not having had enough time to converge to 100% fold – I would be interested to know whether or not it was ever folding 85s. If so, then it would improve it’s strategy by folding all the time with 85o and moving that extra play to 85s as it’s an almost* strictly dominated strategy to ever play 85o if you are ever folding 85s.
* The “almost” is a rare situation I can’t see applies to this all-in. Results of preferring an off-suit hand to a suited hand have been observed in HRC when the game is constrained in such a way that villain has to either fold or jam to a 3-bet. In that case our 3-bet/fold range never sees a flop so we are just choosing it by blockers. Hands like K9o and K9s have almost the exact same blocker effect, with the exception that K9s blocks 1 combo of K9s and 6 combos of K9o, whereas K9o blocks 2 combos of K9s and 5 combos of K9o. If villain is 4-bet jamming K9s but not K9o, then K9o has the superior blocker effect and is preferred to K9s in our 3-bet/fold range. Therefore we can’t say that sometimes folding a suited hand when we ever play the equivalent off-suit hand is strictly dominated in a sense that might satisify a hard-core game theorist, but I think it’s a 99.9% sure way to find out if the 85o was due to convergence or balancing in your all-in.