NFL Gambling

Sunday. January 10, 2021 - 1 min

NFL, Machine Learning, and Gambling

Sports betting is not a new phenomenon in the United States, but it is experiencing a surge due to recent legislation. With 25 states now allowing their residents to wager on sporting events in varying capacities, consumers are searching for tools to assist in making predictions or, instead, to gain an edge in their selection process. Resources are available, but often they are not validated, are high priced, and do not provide any real guidance. With the work conducted under this project, the hope was to create a model the average gambler could utilize in making informed selections through data science.

By analyzing ten NFL seasons consisting of 256 unique games each season, classification models are constructed to see if accuracy rates exceeding 52.4% can consistently be processed. Once considered taboo and sinful, sports betting has grown into an actual corporate enterprise. Given this overall popularity, the field's exploration is just beginning, and the need for data scientists to engage and discover insights will continue to increase. This paper illustrates that machine learning can successfully predict winning and losing teams (approximate accuracy of 65%) and provide an edge in predicting if a team will cover a spread or if a game will be under or over an estimated total score.

GitHub repository can be found here: GitHub
Final results can be found here: Final Results
Final report can be found here: Reports
Presentation can be found here: Presentation

Doug Marcum

Data Science | Machine Learning | Storytelling

Tweet Share