Absolute forecast error by game state
Recorded absolute residual volume is largest in MLB early(inn1-3) at a .4-.6 model probability: 2812.37 total_abs_residual_mass, with mean_abs_residual 0.4888 across 5,754 rows. This describes published residual volume, not reducible miscalibration.
Data integrity
These MLB/soccer numbers are revision 2, computed on the segment-clean corpus (2026-09-16). Revision 1 values are withdrawn and kept in the regeneration receipt. Segmentation excluded 51,635 of the 78,986 MLB ticks (65.4 percent); the specifically identified label mismatches were a smaller 27,076 (34.3 percent).
- residual_anatomy
How to read this figure
Each segment reports n rows, mean absolute residual, and total_abs_residual_mass (n times mean); the ranked list puts the biggest total error slices first.
Published coverage
The committed artifact is partial; only its published coverage is shown here.
| Population | Status | Games | Usable buckets | Published reason |
|---|---|---|---|---|
| MLB | unknown | - | - | - |
| SOCCER_INTL | unknown | - | - | - |
Continue with the data
Open the interactive inspector: Residual anatomyWhat it means
Ranking every sport x game-state x probability-band segment by total absolute error shows the improvement backlog is concentrated where little has happened and the outcome is genuinely uncertain -- early, near-50/50 states. That is where fixing the model would remove the most cumulative error.
Caveats and confounds
No significance test here; total residual mass scales with row count, so high-volume buckets rank high partly by size, not only by per-row error. Segmentation removed about two thirds of the MLB ticks, so every mass figure is smaller than the revision-1 reading without the per-row error having moved much.