2027 WNBA methodology
What enters the model, how partial mocks are treated, and what the confidence labels mean.
Eligibility and independence
Only public, attributable original pages with a reviewed permission status count. Aggregator and roundup URLs remain private discovery provenance while the public record links to the originating writer and publication. Syndicated copies and older versions of one analyst’s opinion cannot create extra votes. Structured big boards may inform player ranking, but without explicit teams they never influence team-pick consensus; metadata pages, withdrawn pages, and low-confidence parses are excluded.
Partial mocks
A lottery-only or first-round mock contributes only to the selections it covers. Absence beyond its final pick is not treated as evidence that a player went undrafted.
Weighting
Each selection receives a restrained recency and parsing-confidence weight. Multiple mocks from the same publication share a publication-level influence cap so one outlet cannot dominate the board.
Ranking and duplicate prevention
Prospects are ranked by weighted average selection, then source appearances. Because the output is a player ranking rather than independently choosing a player in every row, each prospect can appear only once.
Confidence and disagreement
Confidence combines appearance rate, source depth, and projection spread. Tight ranges score higher; wide or lightly sourced ranges receive weak or highly uncertain labels.
Original summaries and corrections
Every player page includes a source-attributed overview sized to the evidence available, comparing placement, team assignment, range, timing, and available reasoning themes. Missing source commentary is labeled instead of invented, and full article language is not reproduced. Explanations based on fewer than three independent publications are labeled preliminary. Publishers can request a correction or removal at any time.