Recent return signs reduced daily Nasdaq risk forecast loss with dated inputs in 2015–2025

We tested whether recent return signs improved daily Nasdaq Composite risk forecasts. Each main forecast used an ALFRED data version cut two calendar days before the target session. Monthly fits used only data then available. The test had 2,689 forecasts in 2015–2025. A fixed model used past squared returns over 5, 22 and 252 sessions. Adding the share of recent squared returns due to negative returns reduced mean QLIKE loss by 0.0402. A paired 95% block bootstrap interval was 0.0251 to 0.0578. The gain was positive in each of the 11 years and remained positive without 2020. With five calendar days of input delay, the gain fell to 0.0228. Variance mean squared error fell by 1.24%, but its interval included zero. The source outage caused 78 omitted sessions. Independent source and scalar checks passed. A second run matched all 19 core output files. We preserve code, data provenance, forecasts, model fits, uncertainty draws, figures and negative findings. Raw source records remain outside the public package under provider terms. This result supports a small risk forecast gain for this index, model and period. It does not establish return-direction skill, a causal mechanism or a net trading gain.

Nasdaq Composite · risk forecasts · return signs · point-in-time data · volatility · QLIKE · chronological evaluation