NCAA women's basketball — Bayesian projection rebuild
D1 women's hoops projections built on the same hierarchical NB2 architecture as the men's side — independent fits per stat, per-(position × experience) quadratic, fitted to a separate Phase 1 women's substrate. Same experience axis (Fr/So/Jr/Sr) since DOB isn't available in any NCAA source. Same architectural lineage as the WNBA pipeline.
Substrate covers 327,913 player-games across 7,626 players and 17,343 games. Position split G/F/C ≈ 214k / 98k / 16k; experience split Fr/So/Jr/Sr ≈ 66k / 69k / 80k / 113k.
Pipeline status
phase 2 grid runningMethodology
how it worksWhy this isn't 538-style team Elo: a Bayesian per-player pipeline lets us combine prior strength, per-position aging, and live in-season updates the way the WNBA stack does. The pipeline below is the spec being built; subscribe to the changelog in the parent NCAA hub for ship updates.
What's in the model
- 3-level hierarchy:
league → position (G/F/C) → player, with non-centered parameterization on the position and player offsets for clean mixing. - Per-position quadratic:
log_rate = mu_player + beta_exp_pos[k]·(exp - 1.5) - gamma_pos[k]·(exp - peak_exp_pos[k])². Each position carries its own peak experience and curvature. - Sport-specific priors: W
mu_leagueper stat (PTS N(-1.15, 1.0), STL N(-3.35, 1.0), BLK N(-3.7, 1.5), TOV N(-2.55, 1.0)). Same widened sigmas as men's (Exp(1) on position + player) for the wide CBB skill range. - Context features (Phase 4, ready): per-(athlete, season) z-scored pace, team quality, mates' usage, and gravity. v2 folds these into Stan; v1 ships first.
What's not in v1
- No per-player experience tilt (sigma_exp_player not identifiable with NCAA's ~1.7-season average career window). Same call as the men's side and the WNBA pipeline.
- No per-row recency weighting yet (deferred to v2 along with the live in-season blend).