Podcast: Play in new window | Download
Nate and Andrew discuss Play Optimal Poker 2: Range Construction, available now from Amazon (paperback/Kindle) or the Nitcast Store (all e-book formats). Includes chapter-by-chapter discussion of the book’s contents.
The original Play Optimal Poker e-book is currently 67% off on Amazon or Nitcast.com.
I think this episode of the podcast illustrates how poker coaches’ obsession with solver-driven theory has gone way too far.
Yes, Andrew gives the usual caveats about how if you know how another player is going to play then you can exploit them, but then we get to the example of the guy who calls turn, and makes a set of 4s on the river. It’s explained that if we bet out then opponents will work that we have a strong hand and respond by overfolding – this may be true, but if we know opponents play like that then we certainly shouldn’t be playing GTO against them, we should be floating the turn and leading the river with all kinds of bluffs. The actual reasons not to lead 4s are that with villain’s polarised range, his valuebet+bluff range is wider than his call range (which is no more than the same hands that would be his valuebets, regardless of reads) and the chance of a check-raise.
This comes up a lot and not just in this podcast; it’s even worse in most of the other podcasts I’ve enjoyed listening to in the past. The coaches use solvers as a crutch and the chance to make an exploitative play is becoming seen as a weird exception and they’re reluctant to get out of their comfort zones and recommend it. I think in a recent episode it was compared to “playing Rock Paper Scissors” – the analogy used to be “playing poker”.
We hear all the time about “don’t make move X because villain will know what you are doing and exploit you”, and we never hear “villain does X, we obviously know what he’s doing here so let’s exploit him” , which seems inconsistent.
The irony is, that solvers don’t think about balancing their ranges at all. They just choose the most exploitative strategy for each individual hand against how they think the opponent is going to play. They only start to become balanced over multiple iterations if there are at least two players with ranges unlocked – i.e. at least two players who can see into each other’s heads – which is not how actual play looks.
When I came to poker in 2012 I started by writing what would basically be called a solver for myself, in order to learn the game from watching it play against itself. The thing that really surprised me was how difficult it was to get a strategy to “settle” when the correct move depended so much on the other player’s strategy – so it ended up in a loop (effectively, you throw rock, so I throw paper, so you throw scissors, so I throw rock, so you throw scissors ….) – which doesn’t happen e.g. in Chess where mixed strategies are not needed to solve a position. I put in mixed strategies, but the main conclusion I drew from that was that essentially I was going in the wrong direction and poker was mainly about knowing how the other person’s ranges were likely to be unbalanced. It’s strange to me that the rest of the poker world has gone in the opposite direction.
I enjoyed listening to Carlos talk about how he integrates his ideas of game theory as a basis for heavily exploitative play (e.g. he notices how the player pool is not check-raising flops anything like widely enough) and would like to hear more about that.