Overnight and Daytime Returns in Commodity ETFs: Do Session Anomalies Survive Trading Costs and Multiple Testing?

We ask whether session-return patterns in commodity exchange-traded funds (ETFs) survive trading costs, multiple testing and implementation constraints. For nine U.S.-listed products over 2007–2025, we evaluate 25 session-specific combinations of Cash, Long, Short, Momentum and Reversal with a self-financing ledger charging actual traded notional. Overnight means are positive for eight funds. The overnight-minus-daytime difference survives a nine-fund Holm adjustment for DBA and SLV, but only DBA survives joint adjustment across 27 session-mean tests. Same-session regressions show no reliable predictability. Of 207 active strategy–fund combinations, 68 beat buy-and-hold before costs, 24 at 1 basis point and four at 2 basis points. Two of these four lose money; the other two, DBE and SLV, show no significant Sharpe-ratio improvement. SPA and stepwise tests reject in none of 27 fund–cost families. Remaining advantages are sensitive to borrow spreads and retrospective outlier treatments. Illustrative participation-rate calculations also indicate limited scale for the least-liquid funds. We find no statistically robust evidence of net outperformance within the tested sample and rule family. This nonrejection does not establish market efficiency or rule out other strategies.