Quote:
Originally Posted by scylla
The thread can be found here: https://forumserver.twoplustwo.com/1...nersev-155966/
As it actually turns out, in GTO, bet sizing is almost completely irrelevant. In GTO, for the solution for any bet size, a solution will exist for a different bet size that has almost the exact same performance. All bet sizes perform very closely to each other (typically within 1%). You can check this for yourself by creating the same tree, but for different bet sizes. After that, solve the trees and compare their overall EV. This EV can be found below the table in the very first decision for OOP. See the screenshot below. As you will notice, this EV will barely change, regardless of which bet sizes you use; even very large or very small ones. So GTO solutions don't really prefer any bet size, and the approach of using multiple sizes to see which one is preferred doesn't really work.
All bet sizings will have nearly the same performance. As a result, when using multiple bet sizes to see which sizes performs best, you're mostly just looking at static. If you make even a slight change to the ranges or tree then the frequencies may already shift to a different configuration.
Put another way, its not so much that bet sizing doesn't matter, but rather the solver is able to adapt itself to different sizing which has no overall net benefit to EV? The only net result is increased complexity in game tree which makes more difficult for humans to analyses? I'm still struggling to understand the bet sizing doesn't matter in GTO though.... If you only give it a large or tiny sizing, your causing the pot-odds to change, which has a relationship to equity, and this in turn has a relationship to the positive expectation of hands in your range.......... whats missing here....
Another related thought. If bet sizing doesn't matter, giving the solver the option to bet does right? As we would be limiting its strategic options. For example, not letting the solver donk flop would remove a branch from the tree, should we always give it all options, even if we don't think players do? Then re run the solution to see how its changed basis what we think a 'normal' tree should look like.
When we hear someone say something like 'I like a small sizing here so that opponents bottom pair continue" thats an example of someone focusing on the noise? If bet sizing is noise, whats the signal? How would you recommend using the solver to focus on the signal? cheers