Predicting the Outcomes of College Football Bowl Games

I'm also interested in the possibility that statistical modeling can be used to predict the outcomes of college football bowl games, and in this paper published in the Journal of Quantitative Analysis in Sports, my colleague Madhur Lamsal and I consider a straightforward application of statistical modeling in determining whether team-level variables were able to predict the actual bowl game outcomes in the 2007-2008 bowl season. I also consider applications of the predictions in the development of ratings for college football teams, based on a round-robin playoff scenario.

Results dating back to 2008 can be found below.

2008-2009 Bowls: Predictions and Results (58.8% accuracy)
2009-2010 Bowls: Predictions, Results and Ratings (55.9% accuracy)
2010-2011 Bowls: Predictions and Results (62.9% accuracy)
2011-2012 Bowls: Predictions and Results (62.9% accuracy)
2012-2013 Bowls: Predictions and Results (77.1% accuracy)

Articles referencing the method have appeared in the New York Times, the Ann Arbor News, and the Kansas City Star.

Constructive comments and feedback are more than welcome. Please keep in mind that I do all of this for a hobby, for fun. I do not get paid by anyone to produce these ratings, and I do not have the time to look at every possible predictor of success! I'm always open to advice about data resources where additional (and more informative) team-level statistics can be found. All of these models are certainly in their infancy, and some of the predictions may definitely look odd (of course I don't truly believe that Missouri was the third-best football team in the nation in 2008...I was purely reporting predictions based on my very young and under-developed model). I simply ask that people read in to the general methods that I've proposed before making personal attacks of any kind. Thanks!

Last modified 9/5/16 by Brady T. West