Expected Points Above Average: A Novel NBA Player Metric Based on Bayesian Hierarchical Modeling

TL;DR

This paper introduces Bayesian hierarchical models to create new NBA metrics, 'expected points' and 'expected points above average,' for improved team and player evaluation, demonstrated through top shot takers and a web app.

Contribution

The paper presents two novel Bayesian-based basketball metrics, providing a new approach for player and team evaluation beyond traditional methods.

Findings

EPAA correlates with established metrics like PER and BPM.

The web app enables interactive team and player comparisons.

Metrics effectively differentiate player shooting abilities.

Abstract

In this paper, we propose two novel basketball metrics: ``expected points'' for team-based comparisons and ``expected points above average (EPAA)'' as a player-evaluation tool. Established within the Bayesian hierarchical model framework, teams and players are clustered based on their shooting propensities and abilities using posterior predictive distributions. We illustrate the concepts for the top 100 shot takers over the last decade and offer our metric as an additional metric for evaluating players. We compare our metrics to two traditional NBA player evaluation metrics: player efficiency rating and box plus/minus. Finally, we develop a Shiny web application that allows interested readers to make additional team and player comparisons.