Pre-registered in tasks/chain13_criteria.md (D1).
Primary driver = shoulder-frame reach ‖p_e − p_shoulder‖ (geometric
extension margin). The σ witness is listed for sanity only — κ = σ₁/σ₆
makes ‘σ₆ low at κ peaks’ true by construction, so it can never CONFIRM
the mechanism. Cliff’s δ: peak windows vs matched control windows, +1 =
driver higher at every peak than every control, −1 the reverse.
| run | peaks | reach δ | reach @peaks [m] | reach @ctrl [m] | sj δ (witness) |
|---|---|---|---|---|---|
| r0_6dof_path | 10 | +0.560 ✓ | 0.281 | 0.269 | -1.000 |
| r0_6dof_minus2pct | 10 | -0.060 | 0.269 | 0.272 | -1.000 |
| r0_6dof_plus2pct | 10 | +0.040 | 0.276 | 0.273 | -1.000 |
| r1_7dof_path | 10 | -0.440 | 0.196 | 0.255 | -1.000 |
| r1_7dof_minus2pct | 10 | +0.160 | 0.191 | 0.201 | -1.000 |
| r1_7dof_plus2pct | 10 | +0.880 ✓ | 0.420 | 0.239 | -1.000 |
| r2_7dof_reactive | 10 | +0.260 | 0.372 | 0.280 | -1.000 |
| r3_7dof_anchor | 10 | -0.160 | 0.207 | 0.245 | -1.000 |
Verdict: H3 OPEN — primary driver δ ≥ +0.5 on 2/8 runs (bar: ≥ 6/8, recorded before the data).
regen: /Users/antoniahoffman/miniforge3/envs/new-pin-env/bin/python validation/chain13_h3_closure.py