Is the 2h drift residual's excess selection on spread width, or a defect? (#84)

resultsupporting

Question

Whether the 2h drift residual's excess capture is explained by selection on wider fill spreads, or is a defect.

Verdict

Completed cycles select NARROWER, not wider, BUY-fill spreads (median fill-vs-item spread delta -0.403pp at 1d, -1.126pp at 2h); no evidence of an inverted book side, mismatched item, or entry/exit join defect.

Population: Tradeable tier, completed cycles.
Resolution: 1d and 2h panels.
Cutoff: Development span only.
Baseline: Each item's own all-bar spread distribution (fill-vs-item spread delta).

Limitations

Rules out the wide-spread-selection explanation; does not itself identify the surviving explanation (recorded in ADR-0009, not here).

Notes

The bug-hunt evidence behind ADR-0009's instrument change. Rules out the 'wide-spread selection' explanation for the drift-decomposition surprise; the surviving explanation (weighted-quote geometry) is recorded in ADR-0009 itself, not in this document.

Source

Citation: po-research:spread-selection:5e2a65f745373cff
Write-up: docs/research/spread-selection.md
Artifact: research-index.json v1 — sha256:cb3d976779a5cbcb4188039a8d26d5e558a2d844030a638c2f16fe5a8eb0a2fe
Build data cutoff: 2026-08-12
Issues: #63, #65, #83, #84, #89