Does the board the manager was handed carry a row for rookies who actually played, and is the gap the fantasy-rostership gate on forecast.py's rookie-prior and second-opinion fallbacks? And were any of the missed men decisions?

Owner lane: decision

Question

Does the board the manager was handed carry a row for rookies who actually played, and is the gap the fantasy-rostership gate on forecast.py's rookie-prior and second-opinion fallbacks? And were any of the missed men decisions?

Prediction

Prereg 2026-09-23T1756Z: played rookies weeks 1-2 30-55% covered, non-rookies >= 95%; rostered rookies >= 95% covered (the gate is the mechanism); E4 2-8 decision-relevant uncovered men per week, majority rookies; consensus has a number for > 60% of uncovered men.

Finding and verdict

TESTED -- PREMISE CONFIRMED, DECISION COST NIL IN WEEKS 1-2. The rostership gate is the whole mechanism: every not-rostered rookie who played had no row (0/84), and rookie coverage was 17% and 11% against 96-97% for everyone else. But the gap has not cost a decision yet: no uncovered rookie scored a flex-starting week (best 14.4 half-PPR vs a 17.7 threshold), and the one decision-relevant hole was a veteran QB with no 2025 played row (H-116's path, Watson). The board cannot rank a waiver rookie against a rostered starter; so far that has hidden 172.6 half-PPR points spread over 86 men, none a start. Two rostered rookies fell off the week-2 board, which is H-246's week-2 window defect, not the gate.

Reasoning

The decision a hole could cost is a waiver add (start/sit only concerns rostered men, who are covered 100% apart from 2 rookies in wk2). Two weeks is small: a breakout rookie (a depth-3 man promoted) is exactly the case the gate hides, and it did not happen in these two weeks.

Next action

Re-run weekly through week 6 with the same runner (cheap, stored boards). The consensus column player_week.proj_fantasypros has no rookie rows in wk1 and no rows at all in wk2 -- a free number the decision lane must grade beside cannot be read; noted for the data path, not measured here.