← Back to all projects
Classifying NBA Player's Position from Game Stats using k-NN

Classifying NBA Player's Position from Game Stats using k-NN

Dennis Dominic Diego, Jason Dolorso, Cymon Marcaida, Matthew Romero

Asian Institute of Management

Executive Summary

From the pre-2000s big-man era full of post-ups, to fast-paced offense strategies like the Suns’ 7-seconds-or-less, and even today’s 3-point-driven basketball, the game has evolved throughout the decades — and along with it, how each player’s position is played. In recent years, pick-and-roll has become the cornerstone of NBA games, giving birth to positionless basketball, where each defender can switch to guard someone regardless of position.

With today’s games emphasizing switching and floor spacing, we wanted to know if we could still determine an NBA player’s position from their stats using a K-Nearest Neighbors (KNN) classifier.

Key Highlights

  • Players’ positions can be identified based on their season average statistics.
  • Accuracy improves when classifying positions as Frontcourt or Backcourt.
  • Assists, defensive rebounds, and 3-pointers are the most important features.

Full text article and source codes can be provided upon request.