Basketball · Research

What Makes an All-NBA Player?

2020-21Updated Sep 11, 2020Archived
Project Overview

A model of All-NBA selection using 10 seasons of NBA.com and Basketball-Reference data, testing variable-selection methods and logistic regression to identify the box-score and advanced stats most tied to All-NBA voting.

Problem & Objective

All-NBA selection involves both performance and narrative factors — it is unclear how much statistical profile alone explains selection outcomes.

Analytical Approach

The team builds classification models using per-game and advanced stats to predict All-NBA selection, then compares model outputs against actual voting results.

Key Result & Impact

Statistical models recover most All-NBA selections with high accuracy and identify a small set of borderline cases where narrative and popularity may outweigh on-court production.

Tech Stack & Tools
RPythonscikit-learn