Serial correlation in National Football League play calling and its effects on outcomes

We investigate the strategic behavior of highly informed agents playing zero-sum games under highly incentivized conditions. We examine data from 3455 National Football League (NFL) games from the 2000 season through the 2012 season, and categorize each play as "rush" or a "pass." We find that the pass-rush mix exhibits negative serial correlation: play types alternate more frequently than an independent stochastic process. This is an exploitable strategy, and we find that this serial correlation negatively affects play efficacy. Our analysis suggests that teams could profit from more clustered play selections, which switch play type less frequently. Our results are consistent with the explanation that teams excessively switch play types in an effort to be perceived as unpredictable.