The calendar-window migration: what moved

supporting-measurementsupporting

Question

What changed in the calendar-window panel migration, and whether Gate A's headline result moved.

Verdict

Daily panel: 37 columns identical, 8 moved, 4 added; 2h panel: 42 identical, 0 moved, 4 added (must be zero, and is). Gate A's headline re-runs to -0.530% with 40.7% of items above zero -- unchanged. Movement is tilted outside the tradeable tier on all 8 moved columns but NOT confined there.

Population: Full daily panel and full 2h panel (all columns).
Resolution: 1d and 2h panels, column-level diff.
Cutoff: Development span only.
Baseline: Pre-migration panel columns/values, diffed against the post-migration calendar-window panel; Gate A's headline re-run as the standing check.

Limitations

Post-hoc measurement -- ADR-0006's migration decision was made before this measurement existed; movement is tilted outside the tradeable tier on all 8 moved columns but not confined there.

Notes

The post-hoc measurement behind ADR-0006's decision (which the ADR is explicit was decided before this measurement existed, and which this measurement only confirms).

Source

Citation: po-research:panel-migration:4c31d197bafc0e72
Write-up: docs/research/panel-migration.md
Data: docs/research/panel-migration.json
Artifact: research-index.json v1 — sha256:cb3d976779a5cbcb4188039a8d26d5e558a2d844030a638c2f16fe5a8eb0a2fe
Build data cutoff: 2026-08-12
Issues: #32, #39, #56, #67