Racking college football teams: A neural network approach

During the past decade, almost all NCAA Division 1-A college football seasons have ended in controversy. This occurs because it is the only collegiate sport in which a national champion is not determined on the playing field. At present, two bias-ridden opinion polls are used to name the champion, and existing computer-based models provide questionable results or use methods that are not publicly known (or both!). The paper has developed an improved method of determining a fair ranking of college football teams using neural networks. The recent 1993 football season serves as an example. The analysis shows that Florida State and Nebraska were properly matched in the Orange Bowl to determine the national champion in 1993 and that Notre Dame had no valid claim for being named the national champion, and it illustrates flaws in existing computer-based ranking models.