27 Sep 2026 · 10:06 AM MTUpdated…
Authored research record

Research output

← All research outputs

forecastFinding recorded

On the board path, which factor makes a line's raw score <= MIN_RAW_FOR_SCALE -- team volume (volume_for), share, or efficiency (efficiency priors)? And what stack value does the guard in stacked_line discard for those men?

Owner lane: forecast

Question

On the board path, which factor makes a line's raw score <= MIN_RAW_FOR_SCALE -- team volume (volume_for), share, or efficiency (efficiency priors)? And what stack value does the guard in stacked_line discard for those men?

Prediction

Prereg research/scientist/prereg/2026-09-23T2158Z-h114-zero-factor.md: every M row is SHARE, zero VOLUME, zero EFF (falsifier: any EFF/VOLUME or touches >= 0.5). F2: >= half of M rows discard v >= 1.0.

Finding and verdict

MEASURED NOTHING FOR THE FORECAST LAYER, AND H-114'S RESCUE CLAIM IS VACUOUS. 22/22 raw<=MIN_RAW rows are share-stage zeros (touches <= 0.0082); volume and efficiency zero none, and the efficiency negative fixture (rates forced to 0) classes 469/469 as EFF, so the classifier could have said so. The stack the guard discards is 0.54 pts in total (max 0.04 per man): a later stage 'cannot rescue' a zero share because the trailing points it would rescue with are also zero. The points H-114's graded zero men scored are information absent from every trailing input -- elevations (H-259), QB regime (H-258), a stale locked build (H-260) -- not a volume, efficiency or stacking defect.

Reasoning

H-114 asserts no later stage can rescue a zero share; the forecast lane owns the stages after share (volume x efficiency x stack). Needed to know whether any of them zeroes or could un-zero a line.

Next action

Nothing further in the forecast layer for H-114. Its remaining work is the share model's handling of ~0-usage men who then play (6 model-stage men here) and the information rows already filed (H-258/259/260). Replaying at the locked wk2 sha a312a9e2 was not done: efficiency.py, MIN_RAW_FOR_SCALE and stacked_line are unchanged since then, and volume_for only lost its market term, so the factor answer should carry, but it is not measured on that build.

Source provenance and publication scope

Owned research record: research/scientist/experiments/2026-09-23T2200Z-h114-zero-factor.json

This public reading view includes authored question, finding, review, reasoning and next action fields. Raw measurements, commands, logs and local paths are withheld. The record ID clock is not proof of completion or deployment.