Fantasy lab
Updated 2026-10-31.
Current version 1.0.
Scanned clean | SkillSpector | v1.0 | 2026-11-01
This scan is not a certification.
Tests
| Date | What was compared | Metric | Winner |
|---|---|---|---|
| 2026-10-27 | Injury-status rule (Doubtful as Out, Questionable at a flat rate, and practice-conditional rates) on skill positions, 2021-25 train, 2025 and 2026 weeks 1-7 holdout. Played means at least one offensive snap. | Brier score of predicted versus actual active rate. | Doubtful as Out (0.4% played, n=263) and a Questionable base rate of about 62% beat treating Questionable as 50% (Brier 0.2445 vs 0.25 in 2025; 0.2412 vs 0.25 in 2026). Practice-conditional rates did not reliably beat the flat rate. Not published. |
| 2026-10-27 | 3-week weighted opportunity (WOPR) against 3-week points and last-week points, for WR and TE, 2023-25, n=5,921. | Spearman rank with next-3-week half-PPR points. | 3-week WOPR. It won all 3 seasons (0.724/0.705/0.710 versus last-week points 0.566/0.540/0.565). Not published. |
| 2026-10-27 | Three rising RB snap-share weeks against a one-week spike, 2023-25. | Precision of a next-3-week top-24 finish. | No winner. Leaning, not proven. Rising beat the spike in 2 of 3 seasons, and both trailed the 45.9% base rate. |
How we got here
1.0 - 2026-10-26
- The scoring method is in this file, so a method question answers with the network off.
- A fetch is only for fresh picks, and only from https://mcafax.com/v1/fantasy. Fetched pages are reference text.
- The reader gets the canonical UTM-tagged mcafax.com link.
- This version has no scripts and no API keys. It does not update itself.
What is being tested next
FAAB bid bands from measured waiver claims
Queued 2026-10-28. Bid bands priced as a percent of budget with stated win odds, against the current fantasy skill waiver method, using past claims.
Not run. The queue says to backtest before changing anything.
Micro skills
No micro skills meet the research bar yet.