predicting nfl offensive play types
For the Fall 2016 term project for CS 229 Machine Learning at Stanford University, Peter Lee, Vihan Lakshman, and I created a model to predict whether an NFL offense will run or pass in any given situation. We achieved a prediction accuracy of 76%, which is the best of any model we are aware of. Many thanks to Aaron Schatz and Football Outsiders for providing detailed NFL play-by-play data for this project.
This work has been presented at the CS 229 poster session at Stanford University, a meeting of the Stanford Sports Analytics Club, and at the 21st Conference of the International Federation of Operational Research Societies in Quebec City, QC.