Quote:
Originally Posted by statmanhal
I’ve got GTO questions (don’t we all?).
If I read him/her right, pucmo says he/she might play tight or loose in early streets and then resort to GTO. I’ll call that partial GTO.
Does that make sense?
Can you start GTO play mid round? Is it reasonable to claim that if you don’t play GTO on early streets there might be game states that occur which would not occur with total GTO? GTO play won’t change the cards but could change bet sizes and folds. Seems to me that a partial GTO cannot have all the properties of a full-time GTO, e.g., unexploitability.
OTOH, maybe it can -- for given a mid-street game situation, what necessary conditions for nash equilibrium are violated for GTO from a mid-point on? If partial GTO is possible starting in the middle of the round, then what distinguishes the two starting points?
Comments?
Awesome question!
My opinion is yes, but an important corollary question is "why".
There will be states that exist that would not happen if previous decisions were gto. For instance, I think gto on the flop tries to avoid being capped, even if it means making a check-raise with value as weak as second pair. Gto would rather try to win the hand right there than proceed oop and capped.
But as to is it possible, I think it is because the tree branches are solved from the end of the hand and computed backwards, which is why it takes so much computing power.
Chess, on the other hand, is solved from the existing board and calculated forward with all the possible moves "deep". This is one of several reasons why solving every possible runout in poker takes longer than chess even though chess has lots more possible combinations of boards.
Anyway, if I am correct, you can take any decision in poker, and calculate it backwards into a gto solution, so long as you can account for every possible runout, or you "bin" it together with other very similar hands-boards combos and use a predetermined solution.
So, the closer to the end of the hand you are, and especially if you are deciding on river actions knowing the full runout, then you can more easily crunch the numbers and make the best gto decision you are capable of.
In theory, any mistakes you have made on earlier streets will still cost you money, but if your "gto decision" forgets those and treats the rest of the hand as Nash then you should only lose (or gain) what you deviated from Nash on earlier streets. For instance, if the gto solution says you have to fold because you allowed pure profit to opponent by a previous street mistake, then it is still gto to fold and you are finishing the hand as close to Nash as possible.
I would be of the opinion that a very good gto player would have ranges preflop, and bet sizing according to ranges vs flop texture, and then really only go deep in the tank on the turn, crunching some approximate numbers and preparing for the different river scenarios and SPR likely on the river.
I just dont see how a human could ever actually crunch a specific gto solution mathematically on the flop, given the ridiculously impossible math.
It has to be either starting as exploit and finishing gto, or starting gto ranges and playing into spr that allows balanced turn/river decisions.