Skip navigation
Please use this identifier to cite or link to this item:
Title: Does the House Always Win? Using Machine Learning to Identify Profitable Opportunities in Sports Wagering
Authors: Roberts, Jack
Advisors: Fan, Jianqing
Department: Operations Research and Financial Engineering
Class Year: 2022
Abstract: The main goal of my thesis was to develop algorithms that can identify profitable sports betting opportunities. Specifically, I trained various models to identify profitable betting opportunities in the NBA sports wagering market. Given this goal, my metric of success is the profit earned (or lost) by betting on the opportunities identified. The definition of a "profitable opportunity" will be discussed later in more detail. Before doing any work of my own, I reviewed similar works done by other researchers. One key difference between my work and many others is that I did not simply predict the results of games, but instead searched for games that offered a good betting opportunity. Thus some other projects used prediction accuracy for their metric of success. After reviewing the relevant literature, I began searching for appropriate data sources. The first type of data required was NBA game results, team points per game, team points allowed, etc. This is the data that trained my algorithm to predict the expected win percentages of teams. The second type of data required was the betting odds offered by casinos. The specific type I used was called "Money Line" odds, but I initially searched for sources of various betting odds. Money Lines were most useful because they imply winning percentages for each team, allowing me to compare my algorithm predictions to the implied winning percentages. With both these data sets prepared, I trained algorithms using the first type of data and compared their predictions to the second set of data. For any given game, if the predicted winning percentage of a team was greater than the winning percentage implied by the casino odds, this was classified as a profitable betting opportunity. After storing these opportunities, I calculated the profitability of betting on these games and compared the performance of these algorithms across a few different scenarios. As will be seen, there are some key trends around betting on the home vs away team, as well as a confusing relationship between predictive accuracy and profit.
Type of Material: Princeton University Senior Theses
Language: en
Appears in Collections:Operations Research and Financial Engineering, 2000-2022

Files in This Item:
File Description SizeFormat 
ROBERTS-JACK-THESIS.pdf3.59 MBAdobe PDF    Request a copy

Items in Dataspace are protected by copyright, with all rights reserved, unless otherwise indicated.