Podcast: Play in new window | Download
Michael Bowling of the University of Alberta Computer Poker Research Group, who has previously appeared on Episode 79 and on Episode 110 after “essentially” solving Heads Up Limit Hold ‘Em, returns to discuss the group’s latest work creating an AI that it claims can beat top human players at the far more complex game of No Limit Hold ‘Em.
Can it? Join us Wednesday at 5PM Eastern on Twitch to find out as Nate and Andrew battle in real time against DeepStack.
Links
DeepStackAI Article in Science
Timestamps
Strategy
From a €500 deep stack tournament in Dublin. Blinds are 150/300/25 and effective stacks are 100-110k.
UTG limps, folds to MP who makes it 700, folds to Hero on the button who 3-bets to 2100 with AQo. UTG 4-bets to 7000. MP folds, Hero calls.
Pot: 15,450
Flop: 9c 7s 2h
UTG bets 6500. Hero calls.
Pot: 28,450
Turn: Qd
UTG bets 11,200. Hero calls.
Pot: 50,850
River: 5s
UTG bets 35,000. Hero?
Thinking about the hand. If I remember the discussion, much is made of the stack size being 300BB.
This sounds a bit unconventional, but I’m wondering how important it is when we are never going to let 4 streets of value go in anyway postflop after that pre-flop action. It’s almost as if the full stack doesn’t play (like a limit poker hand where it makes no difference whether you have 40 or 80 big bets behind). Any thoughts on that?
In the hand I agree with Nate that we should fold pre to this limp 4-bet. I probably even prefer flat the iso depending on what position is meant by “MP” at a 9 handed table (it can sometimes mean HJ, LJ or even the player to the right of the LJ, whom I call U7).