Computing Offensive Strategies of American Football via Counterfactual Regret Minimization

Abstract

In this research, we compute offensive strategies of American Football via Counterfactual Regret Minimization (CFR). American Football is one of the highest strategic sports that offense and defense need to anticipate what the other think. We thus analyze their strategies in game-theoretical framework assuming they are rational players. On the other hand, CFR is the well-known algorithm for computing approximate Nash equilibria in imperfect information extensive-form games. Our goal is that we construct CFR algorithm to compute offense's equilibrium strategies according to National Football League play-by-play data. Our simulation suggests that the proposed algorithm finds offensive strategies that are observed in the data.