Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 183 - Movement in Sports and Medicine
Type: Contributed
Date/Time: Tuesday, August 4, 2020 : 10:00 AM to 2:00 PM
Sponsor: Section on Statistics in Sports
Abstract #313746
Title: NFL Injury Regression to the Mean and Its Causal Effect on Team Performance
Author(s): Zachary Binney* and Gregory Matthews
Companies: Oxford College of Emory University and Loyola University Chicago
Keywords: sports; injuries; NFL; Bayesian statistics; health; causal inference
Abstract:

Unusually healthy or injured NFL teams are often projected to decline or improve in their subsequent season. This entails 2 assumptions: 1.) injury outliers regress towards the NFL mean, 2.) such regression impacts team performance. The first assumption contrasts with prior work showing some teams with consistently high or low injury burdens. We tested each assumption by combining injury burden (modified adjusted games lost, mAGL; 1 mAGL is 1 starter missing 1 game) and preseason betting market data for 2009-18. We found that of 32 NFL teams, the 4 healthiest each year had a median mAGL rank next season of 9 (IQR 5-14); lower ranks are better. The corresponding median rank of the 4 most injured teams was 21.5 (IQR 15-29). Using a Bayesian model of mAGL changes vs. win changes adjusted for skill changes via preseason betting win lines, we found losing ~44 more starter-weeks to injury was associated with 1 less win; the median yearly changes in mAGL for the most and least healthy teams were +33.6 and -32.1. Thus we found 1.) while injury burden partly regresses towards the NFL mean, team effects are a countervailing force; 2.) the median effect of injury regression may be ~0.75 wins.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program