# E2E W16.5 R2 Replay

**Date**: 2026-06-03  
**Scope**: Anderson BOUNDARY physical perturbation landing loop  
**Status**: CONSTRAINT / NO PROMOTION

## Question

Does the stricter isotropic W16.5 replay from 2026-05-29 reproduce with the
same E2E seed family, size ladder and null pressure?

## Run

Artifact:
`tools/data/landing/boundary_anderson_w165_closure_confirm_20260603_logic_r2_replay.json`

Command shape:

```text
python3 tools/exp_boundary_anderson_w165_closure_map.py \
  --sizes 5,6,7,8 \
  --reps 10 \
  --seeds 2026052901,2026052902 \
  --tilts=-0.070,-0.0675,-0.065,-0.0625,-0.060,-0.0575,-0.055 \
  --anisotropic-hopping 1,1,1 \
  --scramble-trials 160 \
  --scramble-seed 202605292001
```

## Result

The replay reproduces the 2026-05-29 R2 result:

```json
{
  "variant_count": 15,
  "w165_closure_count": 1,
  "w165_global_only_count": 0,
  "w165_state_counts": {
    "local_global": 1,
    "local_only": 8,
    "neither": 6
  },
  "w20_preservation_ratio": 1.0
}
```

The only local/global W16.5 closure remains:

```json
{
  "name": "tilt_-0.062",
  "hopping": "1,1,1",
  "onsite_tilt": -0.0625,
  "W16.5": {
    "state": "local_global",
    "min_graph_bridge_frequency": 1.0,
    "classical_states_seen": ["classical_intermediate"],
    "size_states": {
      "5": "stable_graph_bridge+classical_intermediate",
      "6": "stable_graph_bridge+classical_intermediate",
      "7": "stable_graph_bridge+classical_intermediate",
      "8": "stable_graph_bridge+classical_intermediate"
    }
  }
}
```

## Interpretation

This strengthens the narrow isotropic landing coordinate:

- W20 remains the robust local/global reference.
- W16.5 closes only in a very narrow isotropic negative-tilt band.
- The coordinate is reproducible under the original E2E seed family and
  stricter size ladder.
- It is still not a promoted physical law because the 2026-05-29 R3 stress with
  independent seeds plus mild anisotropy removed the full local/global closure.

## Boundary

No law promotion. No public discovery claim.

Operational meaning: the next useful move is not another derived metric. It is
to decide whether `onsite_tilt=-0.0625` is a weak real basin, a symmetry gate,
or a finite artifact by comparing controlled seed families and anisotropy axes.

## Verification

`python3 tools/lab_surface_invariant_check.py`

Result:

```text
lab_surface_invariant_check: PASS
failures: []
warnings: []
```