Archetype Explorer | CHan Analytics

> tune()

Position All

Restrict the universe before clustering.

Historical seasons

Include past player-seasons (21/22 + 22/23 + 23/24 + 24/25). 21/22, 22/23, and 23/24 statistics are calculated slightly differently due to data availability differences, so there may be slight discrepancies as a result.

Show all (M + W)

Pool every available player into one canvas — current and historical seasons, men's and women's combined. Cross-gender entries get a (W) or (M) tag in the labels.

Include anthros (ht + wt)

Add height & weight to the clustering feature set, so size-driven roles (e.g. 7-foot rim runners vs. small-ball fours) separate even when their stat lines look similar. Players without bio data drop out while this is on.

Portal All

Restrict to transfer-portal players. Applies to current-season entries only — historical seasons drop from view when active.

Clusters (K) 8

More K = finer-grained archetypes.

Min MPG 10

Drop low-minute noise from the sample.

Min BPM none

Drop low-impact players. Slider at the far left = no filter; raise it to focus clusters on rotation-quality and above.

> weights()

How much each axis contributes to "distance" between players. Drag a slider to 0 to ignore that dimension.

Scoring 1.0

Shooting 1.0

Playmaking1.0

Defense 1.0

Rebounding1.0

Re-seed picks fresh random initial centroids. K-means converges quickly but the shape can shift slightly per seed.

> names()

Each cluster's name is editable. Pick a suggestion chip or type your own — saved to this browser by member fingerprint, so the name follows the cluster across re-seeds.

> power_tools()

Heavier analytics — most are off by default to keep the canvas snappy.

Sort cluster members

On = nearest-to-centroid first (archetypal exemplars). Off = BPM desc.

Use pins as cluster seeds

When on, pinned player entries (not teams) become initial cluster centroids. Pin 2–8 anchors to guide where the clusters form.

Conference filter All

Restrict the universe to specific conferences (multi-select).

Re-runs K-means with 10 seeds; tags each player with the % of runs in which they end up in their plurality cluster. Strong = consistent archetype; weak = boundary case.

CSV: per-player cluster + features. Snapshot: save the full view state (URL); restore by clicking a chip.

loading…

⌕

PC1 →

← PC2