A Practical Application of Machine Learning to Real-Time Sports Gambling on Football

Abstract

The recent Supreme Court decision in Murphy v. NCAA has made sports betting a states’ right rather than a federal law, creating an opportunity for new businesses that could easily be distributed using the ubiquity of handheld computing devices. This paper aims to prove the capability to create a real-time sports gambling application as a practical use of machine learning techniques on football. This is accomplished by first creating an accurate predictive model of play results, then creating a simulated betting environment to test the capability of the predictive model to offer odds to bettors. The result is an adaptive, predictive model that can be optimized to allow a bookmaker to offer microgambling, giving players options to place small bets throughout the game on every single offensive play. The findings of the paper show the viability of a real-time sports gambling business on play outcomes using a newly formulated predictive model with a state of the art 77% prediction accuracy. In total, this thesis details the enormous opportunity for a business to create an actual application for real-time sports gambling using the methodologies described.