Player Archetypes
NetStats classifies every player across three independent dimensions. A player's compound profile — e.g. Mid+Paint / Ball-Dominant / Hub — tells you how they score, how they use the ball, and how they fit the passing network.
Three Independent Dimensions
Each player gets one label per dimension, assigned independently by KMeans clustering on real play-by-play and tracking data. The three labels together form their compound profile. A player can be any combination — there is no constraint forcing a "big man" label in all three.
Scoring Types
Clustered on shot zone distribution (rim %, paint %, mid %, above-break 3 %, corner 3 %) plus zone efficiency. Ordered roughly interior → perimeter.
Usage Types
Clustered on touches per minute, dribbles per touch, time per dribble, time of possession, and front-court touch rate. The key differentiator is time per dribble — high = Orchestrator (reads the game, sits in post); low with high dribbles = Ball-Dominant (creator off the bounce).
Passing Types
Clustered on passes made per minute, passes received per minute, pass ratio (made ÷ total), assists per minute, and assists per touch. Distinguishes true network hubs from pure receivers.
Example Compound Profiles
Methodology
Models are fit with KMeans clustering on four seasons combined (2021-22 through 2024-25), minimum 500 minutes played. Each season's assignments are then read from a pre-fit model for consistency — a player's classification doesn't shift because of who else played that year.
Cluster names are assigned deterministically from cluster centers after fitting: the scoring cluster with the highest rim % becomes "Rim-Only", the usage cluster with the highest dribbles per touch becomes "Ball-Dominant", and so on. This means the same player will always receive the same label for a given season regardless of when the model was run.
The current season (2025-26) uses tracking data for past seasons. Players without sufficient tracking data show "—" for all three dimensions.