Ranking Rankings: Alternate Ways to Evaluate NCAA Football Teams

Each winter, American college football becomes the center of debate in the sporting world. Coaches, pundits and casual fans alike argue over topics such as which teams are the best in the nation or who is going to win major game. While different official outlets generate top-25 rankings that are used to determine which teams get into the college football playoffs, these are hardly controversy-free. In this talk, we will discuss the opportunities and challenges of predicting college football games and examine five models used to estimate the ability of NCAA football teams. Pythagorean wins and random forests seek to make predictions for future games based on comparisons of prior box-score results. Bradley-Terry and eigenvector methods consider results as well as strength of schedule to evaluate performance and determine ability. After exploring each model’s individual performance, we will combine these four models into an ensemble with the aim of better predicting the outcome of games, with special focus on bowl games