Repositioning the Game: Traditional Positions vs Tracking-Based Archetypes in NBA Performance Models

Driven by the rise of advanced analytics and player tracking technologies, the NBA has transitioned away from traditional positional roles and toward more fluid player archetypes. This investigation uses principal component analysis and k-means clustering to group players based on season-long tracking data, creating new pseudo-positions that more accurately reflect modern playing styles. Predictive models were then built using both the classic position system and the newly generated clusters to forecast player scoring performance. Across every model comparison, both in terms of fit and predictive accuracy, the cluster-based system significantly outperformed the traditional position-based model. These results reinforce the idea that player roles in today’s NBA are more complex than the traditional guard, forward, and center labels suggest, and that adopting a more modern, data-driven view can offer significant advantages in coaching decisions, roster building, and sports betting.