With the NBA growing into an international sensation and sports gambling recently becoming mainstream, research on betting models is limited. Within recent years, the NBA has begun collecting an abundance of data and making it available through its website while sportsbooks have been developing advanced models, using machine learning and statistical analysis, to predict game winners. In this paper we explore the analytical world of basketball and created three different models, to mirror the statistical eras in the NBA, the Dark Ages, the Box Score Era, and Data Ball Era, in order to maximize winnings for NBA sports betting. We took a top down approach by betting on teams based on their recent performance, then based on a team’s basic statistics and lastly creating a model based on the individual player level, achieved through machine learning and statistical analysis. With these three models, we also created a final model which only bets when they all agree. With odds data from the 2018-2021 season, our best model turned out to be our combined one. With this combined model, throughout the three seasons, a bettor would be able to get a return on investment of 12%.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Shanaj, Andrew
• Rago, Philip
• Hatton, Evan

Publisher

• Worcester Polytechnic Institute

Identifier

• E-project-121521-130843
• 43256

Keyword

• NBA
• Data Science
• Computer Science
• Machine Learning

Advisor

• Wills, Craig E.
• Harrison, Lane T.

Year

• 2021

Date created

• 2021-12-15

Resource type

• Major Qualifying Project

Major

• Data Science
• Computer Science

Rights statement

• In Copyright