Optimistic Regret Minimization for Extensive-Form Games via Dilated Distance-Generating Functions

Finally we show that when the goal isminimizing regret, rather than computing a Nash equilibrium, our optimistic methods can outperformCFR+,evenindeepgametrees.