Node locking problem

LeChiffreLeChiffre NetherlandsRed Chipper Posts: 558 ✭✭✭
In GTO+, and I bet in many other solvers, you can lock your opponent's strategy on a certain street and see what the maxEV line is against that strategy. That's all great, especially on the river where there is no more streets to be played.

However, I think it could be misleading to apply node locking on the flop or turn. The issue is that our counter-strategy to a locked strategy on that particular street is still influenced by future streets -- and indeed, our opponent playing the rest of the hand optimally, even though he arrives at future streets with suboptimal ranges.

Let me give an example. In last week's RCP webinar with Ross Glover, he shows how we can modify OOP's strategy to always check to IP player, like most players do if IP is PFR. He demonstrates that, on that flop, IP's c-betting frequency should go down compared to if OOP has a donking range.

But in my mind, it could be perfectly possible that the optimal play for IP is to actually c-bet more (or even a lot), since villain makes some mistakes either on this street facing c-bets, or in future streets. You can't just lock one decision and draw conclusions from the counter-strategy.

Of course, if we are talking about mistakes on this street, those could be locked as well. But the point stands I think that as long as we can not lock our opponent's future decisions, the counter-strategy derived from locking our opponent's strategy on this street could, theoretically, have a lower EV than our strategy against a perfect opponent. I think it would be possible to create a toy game which demonstrates this.

This is not a knock on the idea of node locking, but I think it could be misleading and actually lead to mistakes if not used appropriately (like Ross quickly did in the video - of course it's possible this was intentional given that it was just a quick demo).

Comments

  • eselspieleselspiel Red Chipper Posts: 43 ✭✭
    Since it appears you have GTO+, check out CREV. When you buy one, you get the other.
    With CREV, you can set an opponent's play for multiple streets.
    I don't know a way to set up in GTO+ (Please let me know if there is).
    CREV is sometime awkward to use, but I find it good to exploit population tendencies.
  • porterporter Red Chipper Posts: 313 ✭✭✭
    LeChiffre wrote: »
    But in my mind, it could be perfectly possible that the optimal play for IP is to actually c-bet more (or even a lot), since villain makes some mistakes either on this street facing c-bets, or in future streets. You can't just lock one decision and draw conclusions from the counter-strategy.

    Of course, if we are talking about mistakes on this street, those could be locked as well. But the point stands I think that as long as we can not lock our opponent's future decisions, the counter-strategy derived from locking our opponent's strategy on this street could, theoretically, have a lower EV than our strategy against a perfect opponent. I think it would be possible to create a toy game which demonstrates this.
    Right. When you node lock flop a flop or turn spot, the solver assumes that all other nodes are still played perfectly/clairvoyantly by the opponent. It therefore maximally capitalizes on the mistake at the current node that is locked.
    You could node lock every possible turn and river, but that is not feasible as a general approach. An understanding of poker theory is necessary.
    eselspiel wrote: »
    Since it appears you have GTO+, check out CREV. When you buy one, you get the other.
    With CREV, you can set an opponent's play for multiple streets.
    I don't know a way to set up in GTO+ (Please let me know if there is).
    CREV is sometime awkward to use, but I find it good to exploit population tendencies.
    This approach has flaws as well because CREV does not account for equity realization and implied or reverse implied odds on later streets. Similar to using a solver, you could set a strategy in CREV for both players on all possible run outs, but it isn't feasible. I discuss this more in depth here.
  • LeChiffreLeChiffre NetherlandsRed Chipper Posts: 558 ✭✭✭
    edited December 2
    porter wrote: »
    Right. When you node lock flop a flop or turn spot, the solver assumes that all other nodes are still played perfectly/clairvoyantly by the opponent. It therefore maximally capitalizes on the mistake at the current node that is locked.

    Exactly. My point of the post is two-fold:
    1) to ask to which extent it's possible to lock villain's entire strategy, specifically that of future streets.
    2) to warn readers of the danger of not locking all aspects of villain's strategy if they are known. Only locking a subset of them could make the counterstrategy lower EV in reality than the equilibrium strategy. It is crucial to lock all aspects of villain's strategy that are known.
    eselspiel wrote: »
    This approach has flaws as well because CREV does not account for equity realization and implied or reverse implied odds on later streets. Similar to using a solver, you could set a strategy in CREV for both players on all possible run outs, but it isn't feasible. I discuss this more in depth here.

    Thanks for the link, interesting video.

Leave a Comment

BoldItalicStrikethroughOrdered listUnordered list
Emoji
Image
Align leftAlign centerAlign rightToggle HTML viewToggle full pageToggle lights
Drop image/file