Mixing Matching and Sabermetrics: Combining Advanced Analytics and the Generalized Matching Lawin NFL Football Play-Calling

摘要

The fields of advanced analytics in sports and quantitative analysis of behavior as it applies to sports have developed independently over the last several decades. Both fields share the common goal of using a quantitative approach to describe and predict behavior within sports beyond the common traditional verbal accounts. To date, the two fields have not directly intersected. The current study provides an overview of advanced analytics and quantitative analysis of behavior in sports, demonstrates how the two fields can be combined to better account for the behavioral processes involved in decisionmaking in sports, and identifies several possible ways the two fields can be combined in future research. The Generalized Matching Equation (GME) from quantitative analysis of behavior has been successfully used to account for NFL play-calling behavior in previous studies when yards-gained on a play was the measure of reinforcement. The current study compares GME models using yards-gained as the measure of reinforcement with GME models using Success Rate---a seminal advanced analytics metric in football assessing the value of a play---as the measure of reinforcement. The GME models using Success Rate as the measure of reinforcement were largely found to account for more variance in play-calling behavior and provide better generalized matching outcomes, thus demonstrating the potential value of combining advanced analytics in sports with the quantitative analysis of behavior to improve the description and prediction of choice behavior in elite sports.