Bringing the Heat: Predicting the Pass Rush and Quantifying Pressure in NFL Football

Abstract

Pass protection and the pass rush are frequently overlooked elements of American football. While the players involved in these matchups are typically the largest on the field, they also tend to be the most anonymous and may not receive the credit they deserve for performing their mental and physical responsibilities. In this paper, the pre-snap strategy and post-snap execution of pass protection and the pass rush are studied. First, models are designed to predict defensive play calling, specifically whether or not the defense chooses to blitz. Several different machine learning models are used, with various strengths and weaknesses, and while a number of interesting relationships arise, the overarching conclusion is that coaches are successfully unpredictable in choosing when to be aggressive with a blitz call. Second, a new metric is proposed for measuring the amount of pressure the quarterback experiences from the moment of the snap as a continuous quantity. The metric shows promise in ranking teams’ pass blocking performances in accordance with general consensus and has potentially interesting applications in pocket dynamics and visualization.