Traditional position labels flatten how players actually play — two "forwards" can be statistically nothing alike.
02Build
A Python pipeline that cleans season stats, engineers per-possession features, and runs K-Means to surface natural groupings by playing style rather than nominal position.
03Result
Archetypes that describe role and style more honestly than guard/forward/center, presented through an interactive published write-up.
Product Surface
Technical Specification
Stack
Python
Pandas
Scikit-learn
RoleData & ML
Year2024
Highlights
Per-possession feature engineering over season statistics
K-Means clustering with elbow-method tuning and archetype labeling
