Join Date: Jan 2009
Posts: 4,987
Note: I don’t have a solver so keep that in mind when reading this post but since no one answered as of this posting, I’ll try to get things started.
GTO is defined as that strategy played by two or more players such that no one has any incentive to unilaterally deviate. You are unexploitable and so are your GTO playing opponents.
A solver determines a strategy by having the computer simulating play against itself over and over again revising its strategy until the GTO strategy is close to being realized. You set up a game situation such as your hand, the board, opponent range and possible bet sizes. At each decision point, the solver makes a decision that meets some criterion, presumably one that is related to increasing EV. The “opponent” then responds using the same criterion and play continues. So, decision A at a certain point may result in EVa. If at the same point, decision B yields a higher EVb, then that is the one the computer chooses and processing continues with millions of such iterations.
The mathematical measure of Regret is an underlying game theory concept that the computer algorithm may use where Regret is a measure of the loss you experience by not selecting the best of the available options at each decision point. Fictitious play is another algorithmic approach.
The trials repeat with continual refinements until the so-called Nash equilibrium is close to being achieved – usually when a change in strategy results in such a small improvement that the current strategy is accepted. In poker, that will be when your EV is almost zero assuming no rake and standard position rotation. Typically, the solver will come up with a mixed strategy for a specific situation such as X% of the time bet x, Y% bet y and Z% of the time fold.
