Predicting Results of March Madness Using Three Different Methods

Abstract

Three methods are used to predict the results for two years of the Men’s NCAA Division1 March Madness Basketball Tournament. These methods include using the machine-learning method of the support vector machine, the data mining method of the random forest, and a newly developed Bayesian model using the property of probability self-consistency as an extension of Shen et al. (2015). The random forest method and the support vector machine method are found to possibly do slightly better than the Bayes model, although the results vary. Possible ideas as to how to extend the Bayes model are given.