Modeling safe passing regions in American football: integrating ball trajectories and player coverage

Abstract Spatiotemporal analysis of professional sports like American football is a research area that has grown in recent years with the emergence of reliable tracking technologies for both players as well as the ball. Although established research focuses on analyzing which team controls which parts of the playing field, usually the ball is not taken into account in this type of analysis. We propose a nuanced quantification of passing target regions by which parts of the field can be reached by each player before the ball arrives when a pass is thrown. A safe target area in American football is characterized as a part of the field that can be reached by a receiving offensive player but no defender. A physics-inspired method and a data-driven method for estimating the area each player can cover are presented. Each is implemented using two approaches for discretizations, one based on squares and one based on hexagons. This analysis makes the spatial dynamics in passing plays explicit and thereby provides coaches, academics, and fans a more precise understanding and a novel perspective. The utility of the resulting areas is evaluated using a publicly available dataset of passing plays from the 2018 NFL season. Passes to a safe target area as identified by the physics-inspired method using hexagons with 0.1 yard long sides are completed 89% of the time compared to 59% if the pass was not considered safe, and only 2 out of 410 interceptions occurred when the pass was considered safe. Furthermore, passes that are considered safe according to the data-driven method, using 0.1 yard by 0.1 yard squares, result in 2.02 additional yards per attempt and 0.25 more expected points added per attempt. The identified safe passes are less likely to result in bad outcomes and on average lead to more productive results.