RN-001
93.8% of a Cross-Asset Signal's Gross Edge Was Consumed by Execution Costs
A controlled falsification study of a cross-asset ranking signal that looked statistically useful before market frictions were introduced.
Read research note ↗Archive / RN-001 to RN-005
Research
A compact public archive of empirical work. Notes are permanent, versioned and explicit about their limits. There is no requirement to publish every week.
RN-002
Same-Source Parity Is Not Provider Portability
Recovering the provenance of a historical spread field, reproducing it from the original source, then testing what survives a provider change.
Read research note ↗RN-003
Clock Alignment Is Not Feature Transfer
A three-hour source offset could be corrected. The spread structure still did not transfer. This distinction matters.
Read research note ↗RN-004
Separating Historical Evidence From Prospective Validation
A research protocol for freezing model evidence without silently turning already-seen observations into prospective proof.
Read research note ↗RN-005
From Model Skill to Execution Reality
Why a prediction metric, an executable decision and realized PnL are different objects-and why missing fill data breaks the chain.
Read research note ↗