NFL Prediction Using Neural Networks
Student Researchers: Michael Janning and Saif Ahmad
Faculty Advisor: R. Drew Pasteur and John David (Mathematics)
Our research analyzes the ability of a neural network model to predict the outcome of regular season NFL games. This model uses only readily available statistics, such as passing yards, rushing yards, and fumbles lost. A key component of this model is the use of differentials where, for example, the passing yards of one team are compared to the defensive passing yards of the other team. By using principal component analysis and derivative based analysis, we determined which statistics influence our model the most.
We assessed the performance of the model by comparing its predictions to those of media members and the Las Vegas odds makers. We also consider the absolute error in predicting the margin of each game. In both total wins correctly predicted and point spread error, our model performs similarly to the Las Vegas line. Using the second half of the season for our predictions, we obtained an average accuracy prediction of 65.8% for 2006, 72.2% for 2007, 75.8% for 2008, and 68.2% for 2009 over 10 different realizations of the model. The standard deviation for each year was less than 1%.