wnba methodology
WNBA Methodology
Built batch by batch with pass/fail gates before any metric ships. Same engine class as NBA — no junior partner.
WNBA built one batch at a time — impact metrics, player archetypes, aging curves, defensive measures — each one locked against pass/fail gates before any fitting. The Auto Value Index fuses the six that survived into a single 0–100 player score, weighted by how stable each input is.
engine
Auto Value Index (AVI) — six-component player score + per-archetype aging
Six independently-validated input metrics — impact, archetype fit, lineup RAPM, NET rating, finishing/defense, and aging — are fused into a single 0–100 player value. Each component had to clear a pass/fail face-validity gate before being included in the fusion.
data
Where the inputs come from
sources
WNBA Stats, Basketball-Reference WNBA, lineup tracking-data-free
training
Multi-season WNBA fit; NCAAW Y1 used for rookie priors
holdout
Each batch was rank-validated against expert/coach consensus before fusion
calibration
Out-of-sample performance
metric
Per-component rank correlation + AVI fusion stability across re-fits
value
Published per batch in release notes
Equity-first ship: WNBA is the same engine class as NBA, on the same release cadence, with the same SSR-friendly tables and the same advanced-metrics page.
key levers
What controls the projection
Pass/fail gates
Every input metric had to pass a face-validity gate before fusion. Components that ranked obviously-wrong players highly got cut.
Per-archetype aging
Forward / guard / center aging curves fit separately. League-wide curve hides position-specific decline patterns.
Stability weighting
In the AVI fusion, each component's weight is tied to its observed stability across re-fits. Volatile metrics get downweighted.
Rookie priors
First-year WNBA projections shrink toward NCAAW Year-1 baselines plus draft-position prior. Heavy shrinkage on small WNBA sample.
faq
Common questions
Is WNBA built on the same engine as NBA?
Same engine class, yes. Equity-first means women's leagues ship with the same fidelity tier as their counterparts — same advanced-metrics page, same SSR-friendly tables, same release cadence.
What is the Auto Value Index (AVI)?
A 0–100 player score that fuses six independently-validated input metrics. Each input had to pass a face-validity gate before being included. The fusion weights reflect each input's stability across re-fits, not its in-sample performance.
How do you handle smaller WNBA sample sizes?
Heavy shrinkage toward archetype priors. With ~40 games per season instead of 82, single-season fits are noisy. The aging curves and archetype priors do more work in shrinkage than they would in NBA.
What's NCAAW Y1?
First-year college baselines used as priors for rookie WNBA projections. We re-fit transfer operators between NCAAW production and WNBA Y1 outcomes to set the rookie shrinkage target.
Do you publish defensive metrics?
Defensive components were among the six that passed the face-validity gate. They're in the AVI fusion and on the /advanced page.
apex framework
For the platform-wide methodology framework — pre-registration policy, data philosophy, bias controls, and honesty notes — see the apex methodology page.