Skip to main content
Predictive Sports Analytics Using an Exponential Power Function
Northeast Business and Economics Association (NBEA) 2019
  • Robert Kissell, Molloy College
Document Type
Conference Proceeding
Publication Date
Publisher's PDF
Publisher's Statement
PUBLISHED BY THE NORTHEAST BUSINESS & ECONOMICS ASSOCIATION © 2019 The Northeast Business & Economics Association reserves the right to publish the Proceedings in both print and electronic formats. The individual authors retain the copyright over their own articles.

This paper introduces a predictive sports analytics model that can be used to i) rank teams, ii) estimate win probability, and iii) calculate the expected winning margin (spread) of a game. The model is based on an exponential distribution function and is solved using a log-loss function and maximum likelihood estimates (MLE). The most appealing aspect of this quantitative approach is that the model is objective, transparent, and testable. Readers can verify all calculations and results as the model does not employ any black box method. We apply this model to the 2018-2019 NFL season to rank teams, estimate scores, and make predictions across all pairs of teams. The model predicted more than 70% of the NFL games correct, and has 𝑅2 = 26% when comparing estimated spreads to actual spreads. This model can also serve as an objective approach to assist committees determine which teams should be selected for college post season tournaments such as the NCAA basketball tournament and the College Football Playoffs.

Citation Information
Robert Kissell. "Predictive Sports Analytics Using an Exponential Power Function" Northeast Business and Economics Association (NBEA) 2019 (2019)
Available at: