Fourth Down's Critical Role in a Winning Strategy for American Football Games

Abstract

Data analytics is booming, and this boom has started to creep into sports – originating in baseball and moving its way across other popular sports leagues. American Football is one of the last sports to embrace the analytics revolution; however, a noticeable use of data analytics in recent years within the sport of American Football is the decision on whether to punt or “go for it” on any fourth down during a game. There have been studies on this dating back to the 1970’s, yet most of these studies revolved around the discrete concept of points, as opposed to a continuous concept such as win probability. Regardless of the variable of interest, studies have shown that teams are being too conservative on their fourth downs and punting the ball to their opponent far too often. This study uses ten years of historical play-level data from the National Football League coupled with machine learning methods such as random forest algorithms and nearest neighbor matching along with a Poisson regression to evaluate team’s decision-making on fourth down; the study will then evaluation the relationship between decision-making and the number of wins the team has experienced from 2010 to 2019. This study agrees with those before it that teams are not making the “optimal” win probability maximizing choice on fourth down as often as one would expect; however, the frequency of which teams are maximizing their win probability does not have a direct effect on the number of games a team actually won over the course of the seasons included in this study. This could be due to reasons discussed in the discussion session of the study; though we must remember that maximizing the probability of an event occurring does not necessarily mean that said event will occur.