
Predicting NCAA March Madness Games Using Bayesian Logistic Regression Techniques
Date of Creation
5-1-2020
Document Type
Departmental Honors Thesis - Restricted Access
Department
Mathematics
First Advisor
Eric Ruggieri
Abstract
Warren Buffett, otherwise known as the oracle of Omaha, made headlines in 2014 when he offered one billion dollars to whomever could correctly pick every game of the NCAA Men’s basketball tournament, also known as March Madness. There is no question that the March Madness Tournament captivates the nation for an entire month, and because of its significance the tournament also captivates thousands of statisticians and mathematicians every year. We are among those statisticians that will try to create models to predict the most recent March Madness tournament. We use mathematical concepts such as logistic regression, regression trees and Bayesian statistics to create multiple models based on the data we collected from the last 10 seasons of college basketball seasons. The results of each model will then be compared to various baseline models, including Ken Pomeroy’s model, a leading statistician’s model in the field of college basketball. Our models will not only compete with Ken Pomeroy’s model but also on a national stage, placing in the 99th percentile out of millions of models in the ESPN Bracket Challenge.
Recommended Citation
Pogorzelski, Piotr, "Predicting NCAA March Madness Games Using Bayesian Logistic Regression Techniques" (2020). Math and Computer Science Honors Theses. 62.
https://crossworks.holycross.edu/math_honor/62
Comments
Reader: David B. Damiano