Basketball · Research

Statistically-Grounded All-NBA Defensive Team Selection Model

2024-25Updated May 30, 2025Archived
Project Overview

A statistical framework for predicting All-NBA Defensive Teams and Defensive Player of the Year by modeling possession-level matchup data rather than media narrative or box-score totals.

Problem & Objective

All-NBA Defensive voting heavily relies on box-score totals (steals/blocks), team record, and media narrative bias, misvaluing elite defenders on poor teams or versatile switching defenders.

Analytical Approach

Processed possession-level matchup tracking data to isolate defender vs. ball-handler execution, normalizing for position and role opportunity, and applied SHAP explainability to attribute individual defensive impact.

Key Result & Impact

Formulated a reproducible defensive rating metric that accurately identifies elite individual defenders independent of team win totals or voter reputation bias.

Tech Stack & Tools
Pythonpandasscikit-learnSHAPnba_api