Adding prior overnight gaps raised ridge error for next-session open-to-close returns in SPY, QQQ and IWM in 2016–2025

We tested whether the previous overnight return improved next-session open-to-close return forecasts for SPY, QQQ and IWM. A fixed ridge model used past returns and volatility. Annual fits used only earlier years from 2001 onward. The test covered 2,514 sessions per fund in 2016–2025. Adding the prior overnight gap increased pooled mean squared error by 0.36%. The paired 95% block interval for error reduction was -1.19% to +0.17%. Adding the current opening gap as a forecast diagnostic also failed to improve the pooled score. We did not simulate same-open fills for this diagnostic. With entry and exit costs of 2 basis points each, the prior-gap model earned 2.63% a year in an equal-fund long-or-cash simulation, with interval -1.88% to +7.35%. The model without the gap earned 4.65%, with interval -0.03% to +9.51%. Gains depended strongly on a few large days. These results do not establish reliable profit or rule out other overnight signals. The data are one retrieved historical price vintage, not archived point-in-time vintages. The package preserves calculated model inputs, all forecasts, fits, paper trades, uncertainty draws, source hashes, code and replay instructions. Raw price records remain external under provider terms.

ETF · return forecasts · overnight returns · chronological evaluation · trading costs · negative result · ridge regression