Beating the Vegas Spread: Using Multiple Regression to Predict the Outcome of College Basketball Games

Open Access
- Author:
- Grabowski, Connor
- Area of Honors:
- Actuarial Science
- Degree:
- Bachelor of Science
- Document Type:
- Thesis
- Thesis Supervisors:
- Steven L Putterman, Thesis Supervisor
Andrew John Wiesner, Thesis Honors Advisor - Keywords:
- actuarial science
sports
gambling
statistics
regression
actuarial
actuary
basketball - Abstract:
- One of the most popular vices in today’s society is gambling. Many would argue that no thrill rivals the reveal of a perfect blackjack, when a chosen horse crosses the finish line in first position, or when a roulette spin lands right on what a bettor chose. In recent years, however, there has been an industry-wide shift towards a new style of betting: sports. With legislation passing through the United States Congress early in 2019, sports gambling has become the sexiest avenue for casinos, applications and bookmakers alike. Now with well over $10 billion in total dollar value wagered, gamblers find themselves losing larger and larger sums of money. This study aims to add automation and a mathematical approach to sports gambling through the lens of multiple regression and common programming methods. In the four major American sports of baseball, basketball, football, and hockey, the three most popular ways to gamble on games are moneyline (predicting the winner), spread (predicting the margin of victory), and over/under (predicting the total number of points, goals, or runs). This study will focus on point spread but explore ways to implement the same mathematical approach to moneyline and over/under. The hypothesis this study aims to test is if a number of ranking metrics that rate teams from first to last can be combined into a single database to predict a single margin of victory for a specific game. This study will focus on the NCAA Men’s Basketball tournament from 2014-2019 because of its wide syndication of gamblers as well as college basketball’s long list of common rating systems. The study will conclude by testing the database against the results of the 2019 and 2020 NCAA tournaments to determine overall profitability. If this study proves successful, further avenues for implementation in other sports will be explored.