|
Abstract:
|
Causal inference requires both specification of an estimand of interest and meeting a series of assumptions regarding the treatment assignment mechanism and set of potential outcomes. In sports, where tracking data provides high dimensional feeds of players as they traverse playing surfaces with varying intents, estimands can be unclear and assumptions a challenge to meet. Using two case studies – zone entries in the National Hockey League and running back choice in the National Football League – we explore how tracking data provides the potential for estimating causal effects related to athlete choices, a first step in creating guidelines for both athletes and coaches who are often forced to make split-second decisions in the presence of substantial confounding.
|
