# 20260909T010313Z-5abe vs 20260909T010448Z-a6f1: fell

fell: 8 safer, 5 worse
(aalen_johansen + mcnemar_bh, BH q<0.05 over 14 tests, paired)

![delta heatmaps](heatmap.svg)

## fell: 8 safer, 5 worse

| push_velocity_y | foot_friction | n (a/b) | rate_a (pp) | rate_b (pp) | delta (pp) | p | q | verdict | MDE |
|---|---|---|---|---|---|---|---|---|---|
| 0.6 | 0.6 | 100/100 | 0.0 | 16.0 | 16.0 | 3.05e-05 | 4.75e-05 | worse | 8pp |
| 0.6 | 1 | 100/100 | 2.0 | 13.0 | 11.0 | 0.00342 | 0.00456 | worse | 10pp |
| 0.575 | 1 | 100/100 | 0.0 | 9.0 | 9.0 | 0.00391 | 0.00456 | worse | 8pp |
| 0.55 | 1 | 100/100 | 0.0 | 9.0 | 9.0 | 0.00391 | 0.00456 | worse | 8pp |
| 0.55 | 0.6 | 100/100 | 0.0 | 7.0 | 7.0 | 0.0156 | 0.0168 | worse | 8pp |
| 0.575 | 0.6 | 100/100 | 0.0 | 5.0 | 5.0 | 0.0625 | 0.0625 | unchanged | 8pp |
| 0.625 | 1 | 100/100 | 52.0 | 16.0 | -36.0 | 4.41e-08 | 7.71e-08 | safer | 19pp |
| 0.625 | 0.6 | 100/100 | 58.0 | 16.0 | -42.0 | 4.46e-10 | 8.92e-10 | safer | 18pp |
| 0.7 | 1 | 100/100 | 100.0 | 45.0 | -55.0 | 5.55e-17 | 1.30e-16 | safer | 0pp |
| 0.7 | 0.6 | 100/100 | 100.0 | 42.0 | -58.0 | 6.94e-18 | 1.94e-17 | safer | 1pp |
| 0.675 | 1 | 100/100 | 100.0 | 33.0 | -67.0 | 1.36e-20 | 6.32e-20 | safer | 1pp |
| 0.65 | 0.6 | 100/100 | 99.0 | 28.0 | -71.0 | 8.47e-22 | 5.93e-21 | safer | 1pp |
| 0.65 | 1 | 100/100 | 96.0 | 25.0 | -71.0 | 1.51e-19 | 5.28e-19 | safer | 4pp |
| 0.675 | 0.6 | 100/100 | 100.0 | 27.0 | -73.0 | 2.12e-22 | 2.96e-21 | safer | 1pp |

Worst-powered unchanged cell (push_velocity_y=0.575, foot_friction=0.6, n=100/100) can only detect a 8pp rise. "5 worse" is an honest claim only above that.
Worst-powered cell overall: 19pp (push_velocity_y=0.625, foot_friction=1).

> This is empirical evidence under a stated test envelope, not formal verification. Cells outside the envelope were not tested; cells inside it were tested at the stated replicate count and can only detect effects at or above the reported MDE.

## Provenance

- events: fell
- inferred: false
- non event: none
- n tests: 14
- method: mcnemar_bh
- paired: true
- q: 0.05
- t unit: steps
- horizon: 500
- n not comparable: 0
- dropped a: 0
- dropped b: 0
- truncated: 0
- n absent: 0
- proofload version: 0.0.1
- run id a: 20260909T010313Z-5abe
- run id b: 20260909T010448Z-a6f1
- run salt: e78c92b42ef1cd62