I have a question: What BB/100 advantage can we get against certain mistakes (deviations) of our opponents, and how to measure it? For example, we play HUNL cash (but for 3-6 max the situation, I think, will be similar) against a regular who, in general, plays equivalent in skill to us, we both trying to play close to GTO, the game takes place in a rake-free environment, which means that ours winrate will be 0 bb/100 as a result. But, suppose that the opponent at some point begins to deviate and overbluff the river in the triple barrel line in 3BP (3B-B-B-Jam 100% pot, line), for example, doing this with all bluffs that in theory are bluffs of at least 10 % of cases (0EV bluff) which in total means that the ratio of his bluffs is, let’s assume, not optimal 1:3 (vs B100%), but 1:2. If I understand correctly how GTO works in these situations, then if he were playing against a GTO opponent, he would not lose EV, but since he is playing against a human, we notice this and begin to adjust, calling all the bluffcatchers, which will obviously lead to loss in EV for him, but approximately how much does the error cost our opponent’s win rate? -0.1bb? -0.5 bb? Can we somehow measure this? What if he makes a mistake preflop and opens the SB 20% less often than the GTO? Will this deviation generate more EV loss or less?
I think that in order to make the situation more understandable and measurable, for simplicity, let's assume that initially both play GTO with high accuracy and both adjust perfectly if they notice each other's mistakes. It will be similar to the situation when two GTO bots with nodelocks play against each other. By “with nodelocks” I mean that if one bot make an error, then the second bot, assuming that this error is not an isolated case, will adjust his strategy as much as possible.
In an attempt to measure this advantage, I decided to look at how often certain situations occur in the game tree and what the average pot size is in each of them, by multiplying the % of situation frequency by the average pot sizу, using which it is possible get closer to the answer. I call it “situation value coefficient”. The note that this coefficient can either be found in your tracker (H2N for example) by using a certain filter and dividing the number of situations by the total number of hands, or by calculating it manually if you know the optimal frequencies of each line being played. The average bank size can simply be determined approximately, in H2N there is no such stat, maybe there is in HM3.
“Situation value coefficient" is something like the Big Blind Volume, which refers to how many big blinds are invested in each individual situation.
Here are some examples of calculating this coefficient in HUNL:
Description of the first line:
Action before Hiro (Situation: Possible Actions)
Donk bets are excluded in each node.
Example 1(preflop):
Example 2 (flop in SRP):
Example 3 (turn in SRP):
Example 4 (flop in 3BP):
But I don’t know how to specifically determine how many bb/100 an opponent will lose if he deviates in a given situation if it is more “valuable” or less “valuable” than others.
What are your thoughts on this case? I would be grateful for any answers and thoughts.
