This is a review of the negative binomial regression model research on Heisman voting conducted Nolan Kopkin. The Heisman Trophy is the top award given to a college football player every year. However, many speculate that there is a bias in the voting process. The country is split into six regions, Northeast, Mid-Atlantic, South, Southwest, Mid-west, and Far West, with each region having 143 votes to distribute among its media. Former Heisman Trophy winners are also given a vote. The public is given an aggregate vote via online polling.
Theoretically, it is expected that a regional bias exists when the finalist and voters are from the same region and that finalists will receive more votes from nearby regions than finalists based further away. This impact would decrease if a region has multiple finalists, which would split the vote. The location of the finalist’s opponents also plays a factor, as well as national media coverage.
Data was collected from 1990-2016 and analyzed to produce summary statistics. Each piece of data was weighted by the inverse of the number of finalists in that particular year so that all years are weighted equally. The statistical significance of each difference between in-region and out-of-region samples was tested using chi-squared tests. The stats demonstrate that voters do favor players who are based in their region as well as players who play more games against regional opponents.
Boxplots clearly show the bias regarding in-region players. When voting for out-of-region finalists, voters from the Mid-Atlantic are more likely to vote for players from the Northeast than the other four regions. They are then more likely to vote for players from the South than the remaining three regions. The Midwest is less likely to vote for finalists from the South and Far West voters are more likely to vote for Midwest finalists than those from other regions.
A negative binomial regression model is used to analyze regional vote tallies in order to handle the skewed, discrete data. Results indicate that finalists receive more points from their home region than from other regions. Games played in region lead to a higher tally count as well. Increased national media coverage leads to an overall decrease in regional bias.
Results from a sensitivity analysis show that the fact that players receive more votes from their own region is standard across all six regions. The idea that finalists from the Far West receive fewer votes than those from other regions is shown to be false and that, in actuality, finalists from the Northeast, South, and Southwest tend to receive fewer votes. The Northeast demonstrates the strongest in-region bias.
Analysis demonstrates that bias is definitely prevalent within the Heisman Trophy voting. The fact that national media coverage decreases the bias leads to the question of whether there are additional methods that would aide in decreasing the bias. This would be a worthwhile endeavor for analysts to research as the ultimate purpose of the Heisman Trophy is to award the best player overall.
Analytics Methods Used: Weighting, Chi-Square, Boxplots, Negative Binomial Regression, Sensitivity Analysis
Find out how Sports Analytics Expert Victor Holman can give your team the competitive advantage.
How mature is your team’s analytics program? Take the Sports Analytics Maturity Assessment.
Discover the Groundbreaking Sports Analytics Application and Framework coaches and sports analysts are talking about!
Learn all about sports analytics in Victor Holman’s Sports Analytics Blog.