A fixed volatility mix cut daily Nasdaq risk loss against a selected 20-session scale and gave no clear gain over 0.94 EWMA in 2018–2025

We tested a fixed mix of three daily volatility scales on the Nasdaq Composite. Each input used a dated ALFRED record from before entry and prices from session t-2 or earlier. We selected one half-life on 754 eligible rows from 2015–2017. The chosen value was 20 sessions. All models then used annual calibration from earlier dated forecasts and released labels. In 2,011 test sessions in 2018–2025, an equal-weight mix of 5-, 20- and 60-session scales cut mean QLIKE loss by 0.03831 against the selected scale. The paired 95% interval for the loss change was -0.06760 to -0.01569. The gain remained without 2020 and with first-release outcome labels. Against a standard 0.94 EWMA, the change was +0.01347, with interval -0.01077 to +0.03786. This gives no clear gain over that reference. The five-session component also had lower point loss than the mix. The later four-year gain was uncertain. We preserve the fixed plan, all forecasts, fit states, source records, code, figures, negative findings and uncertainty draws. Independent checks passed. Eleven numerical files matched in a separate execution. This test supports a limited second-moment forecast comparison. It does not establish return-direction skill or trading profit.

volatility forecasting · point-in-time data · forecast combination · EWMA · Nasdaq Composite · chronological test · negative results