Parity and Predictability in the National Football League Final Write-up Siddharth Patel

Predicting the outcome of National Football League games is a staple of American sports culture and a significant component of the multi-billion dollar sports betting industry. The NFL prides itself on the perception of parity, the idea that any team can beat any other on ”any given Sunday.” In this paper, I apply machine learning techniques to team performance statistics in order to predict the outcome of games. I seek to determine with what accuracy the outcome can be predicted and how those predictions would perform against sports book betting odds. After tuning a random forest classification tree on the last ten year’s worth of NFL games, I find that I can predict the outcomes of games in the 2013 season with 60% accuracy. I cannot consistently make money against Las Vegas sports book odds.