Decision Making in American Football: Evidence from 7 Years of NFL Data

Do team managers and coaches make rational decisions? In the era of big data one would expect coaching decisions to always be rational with respect to the maximization of the expected points scored. In order to examine this hypothesis we use data from the National Football League for the past 7 seasons to analyze two specific discrete decisions/choices that coaches face several times during a game. In particular, we analyze (i) the Point(s) After Touchdown (PAT) and (ii) the fourth down decisions. We use mean-field approximations and find that despite the statistical evidence available to them through years worth of data, coaches tend to make the status quo decisions, which in the majority of the cases does not lead to point maximization. One of the possible explanations that we put forward for this behavior is the fact that their coaching objective might be to minimize the variance of the expected points scored.