Is the D-76 passing-TD market anchor (playermodel.lines_for ~:1698, _reconcile(qb, _pt, 'pass_tds')) discarded for teams with a designated QB, and where in the chain?

Owner lane: Not recorded

Question

Is the D-76 passing-TD market anchor (playermodel.lines_for ~:1698, _reconcile(qb, _pt, 'pass_tds')) discarded for teams with a designated QB, and where in the chain?

Prediction

Inside playermodel.lines_for the pass-TD anchor holds for every team regardless of designation (code: no designation branch). On the published forecast table, control team-weeks (no Out/Doubtful/Questionable QB) read retained = 1.000 within 0.002 at every k>=1. Teams with an OUT-designated QB read retained pass_tds well below 1 at k=1 (predict 0.30-0.70) and recover toward 0.9-1.0 by k>=3. Questionable-only teams read within 0.05 of 1 at every k. So H-130 as worded (silently discarded for ANY team with a designated QB) is predicted FALSE on the graded lines_for path and TRUE-IN-PART on the forecast.build future-week path, where the discard is k- and designation-dependent.

Finding and verdict

H-130 as worded is NOT OBSERVED on the published object: 0 of 308 designated-QB team-weeks across 3 generations lose any anchor mass beyond 4-dp rounding. But MEASURED NOTHING about the claimed mechanism: no designated QB reached the rebuild, so the test never exercised it. My preregistered prediction (Out teams 0.30-0.70 at k=1) FAILED -- the premise it rested on (designations arrive) was false. By code the mixed-world discard is reachable in forecast.build once H-227/H-245 are fixed; on the graded lines_for path the anchor is unconditional. Next test is WORK, not time: inject designations in memory into the production latest_rows rows and read forecast.build's own output on a scratch snapshot with forecastkey.preserve stubbed (it appends to data/land, which is why this run did not call build).