Claims meet outcomes.

Finding 01

Preseason convictions

21 aligned, 15 mixed, 13 contradicted, and 6 availability-limited among 55 approved Takes.

eligible n=49; excluded n=6; rule=preseason-conviction-v2; dataset=nflverse-active-games-2021-2025-v2; baseline=historical market ADP unavailable; Wilson 95% for aligned share 30–57%

This is a performance-tier result, not proof that the advice beat draft cost.

Excluded family

Weekly starts

The WP5 pilot measured 0.643 precision and 0.600 recall for weekly-lineup extraction. That misses the family gates of 0.90 precision and 0.80 recall, so no records, baselines, result distribution, or examples are promoted to the release.

pilot n=15 gold Takes; matched n=14; material matches n=9; release eligible n=0

Excluded family

Waiver advice

The WP5 pilot measured 0.636 precision and 0.368 recall for waiver extraction. The family remains outside the approved ledger rather than being folded into a broader hit rate.

pilot n=19 gold Takes; matched n=11; material matches n=7; release eligible n=0

Finding 02

Trade stays descriptive

5 complete windows produced 2 improved and 3 inconclusive records. 7 lacked a complete prior window.

sample n=12; eligible n=5; excluded n=7; rule=buy-sell-v2; baseline=self prior four active games; market baseline unavailable

The required 25-record audit cannot be met with 12 approved Takes. No percentage is presented as trade accuracy.