Online Program Home
  My Program

Abstract Details

Activity Number: 515 - Statistical Modeling for Sports Science and Applications
Type: Contributed
Date/Time: Wednesday, August 2, 2017 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistics in Sports
Abstract #322939 View Presentation
Title: Properties of Logistic Regression Estimates of Team Strengths in Sports
Author(s): Richard Auer* and Jennifer Roem
Companies: Loyola University Maryland and Johns Hopkins Bloomberg School of Public Health
Keywords: Major League Baseball ; The Logistic Regression Model ; Unbalanced Schedules

Using win-loss percentages as measures of sports team strength leads to misjudgments when comparing the quality of the teams in a league. Most of the problems arise due to an imbalance in the relative strengths of the opposition that the teams face. Fitting simple logistic regression models, where dummy variables identify the teams playing in individual games, leads to team effects that are much truer measures of team strength. In Major League Baseball, these team effects account for the specific division a team plays in and which league is faced in interleague games. Win-loss data was taken for every team in every season from 1998 to 2010 accounting for 31000 games in total. In the National Football League, teams only play 16 games per season with even more varying levels of opposing team strengths than in baseball. Using NFL data from the 2010 season, similar logistic regression models were fit and the size and distribution of the team effects were compared between the two sports. The potential fit of normality is considered for the distributions of each sport's set of team effects. And various experiments were tried to study how team effects changed under various conditions.

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

Back to the full JSM 2017 program

Copyright © American Statistical Association