Past yen gains gave no clear added gain in daily Nasdaq risk forecasts in a 2018–2025 archive test

We tested whether past yen gains improve daily Nasdaq Composite forecasts after past index return sizes and signs, and yen volatility, are known. Each input used an ALFRED version dated before the target started. Monthly models used fixed historical features and only available labels. The main test had 1,930 forecasts in 2018–2025. Adding yen direction gave mean QLIKE gain of -0.00076, with a paired 95% 20-origin block interval from -0.00843 to +0.00668. The planned clear-gain test did not pass. Five-origin and 60-origin blocks, fresher inputs, later labels and period checks also gave no clear gain. The fitted risk response to yen gains was positive in every main fit. This association did not give useful test evidence. A separate mean-return model had out-of-sample R-squared of -0.273% against the past mean, with interval -0.736% to +0.223%. Its sign accuracy was 53.73%, versus 55.34% for always positive. The weekly rate release makes this a test of delayed public yen information. It does not reject value from newer currency prices. No trading strategy was tested. The package preserves public domain yen data, source hashes, code, all 2,432 forecast rows, 117 fitted states, all planned score and uncertainty files, figures, negative results and repeat instructions. Independent checks passed. A full repeat and its hashes are preserved. Restricted Nasdaq labels and features are retained separately under source terms. The result supports a scoped negative forecast benchmark, not reliable return prediction or trading profit.

Nasdaq Composite · Japanese yen · risk forecasts · point-in-time data · chronological evaluation · QLIKE · negative result · exchange rates