Player Performance Prediction Automation for DraftKings

Katz, Gabriel A.

Student Work

Player Performance Prediction Automation for DraftKings

Public

In this Major Qualifying Project, we worked with the Boston company DraftKings, an online fantasy sports and sportsbook provider, to build models that predict NBA players' performances in an individual game basis (number of points, assists, and rebounds scored in a single game). We also created models that predict individual game outcomes in the form of a sportsbook closed line. We built linear regression, Lasso regression, random forest, and neural network models which utilized data directly from DraftKings and reputable third-party sources to predict these performance metrics. Our most successful models reduced the prediction error by up to 11.46% from the baseline model predictions. When compared with current sportsbook information and newly developed DraftKings data science models, our results were found to be on par with or surpassing industry standards.

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