A way of computing that makes the program play itself again and again, cutting the choices that lost.
Short for Counterfactual Regret Minimization. It adds up "how much better another action would have done" (regret), and picks actions with large regret more often next time. Repeated over and over, it gets close to GTO. This is what runs inside a solver. Example. If Fold did better than Call many times in a spot, Fold is chosen more often from then on.