No announcement yet.

What is the solver doing?

  • Filter
  • Time
  • Show
Clear All
new posts

  • What is the solver doing?

    Hi. New member here. I have a few questions that perhaps ya'll can help me with. If there is already a discussion of this somewhere, you could possibly just direct me to it.

    I have some idea of what a solver is doing, at least if I'm correct in thinking it is playing millions of hands, trying different strategies, and adjusting until it achieves an optimal strategy against an opponent who also is trying to achieve an optimal strategy.

    But I have some specific misunderstandings of what it can possibly be doing that might be illustrated as follows:

    100BB deep, UTG+1, playing against a raise: we are told to fold K9s and K8s, but to 3bet with K7s and K6s some percentage of the time (42% and 67% respectively). There are no calls on this particular chart, which I find interesting but I don't really question that part. What I do question is why would you always fold K9s but sometimes call K7s, other than that this is what the solver has found to work best so just do it and don't ask questions. Because my question is why and or how can that be correct?

    Now I can speculate myself that, against very solid opposition, we want to have some lower cards in our range, maybe not all the time but some of the time, so they have to consider the possibility. But on the other hand we don't want to have them all the time because they aren't worth it. Or, similarly, that we "need some bluffs" in our range, and possibly K9s is too weak to justify a play but also too good to use as a bluff. Are these the reasons? Or are they wrong? Or only partially right?

    Along the same lines, suppose that instead of playing K7s and K6s the recommended percentages, we decide to just play K7s 100% and never play K6s. Is the reason here that someone with a perfect memory and many thousands of hands against us can realize that we play the 7 but not the 6? If that is the reason, does it really have any practical application (is it something a human can use it in a live game)?

    These are my specific questions but hopefully you can see where I'm going with this and tell me what I'm missing or where it is to be found.


  • #2
    One of the main issues with solvers, is that they can tell us what to do, but not why to do it.

    If you've ever watched videos of top pros going through solver analysis you'll have no doubt seen them stumped many times.

    It's often common sense such as the need for balance, board coverage, blockers etc. But sometimes it does appear fairly random. This is perhaps just due to the solvers being as close to perfectly balanced as possible.

    A key point to working with solvers is to not try and memorise their ranges exactly, but to try and look for patterns. If a solver is betting a bunch of hands roughly 10% then we can safely never bet with them. Sure we lose a tiny bit of EV, but we stand to lose a lot more by trying to implement a very complex strategy and getting it wrong. If a solver is taking every top pair hand and betting it nearly 100% of the time in a certain spot, then we should consider doing the same.