Empirical path bounds gave no clear score gain over normal path bounds in 385 Nasdaq archive tests in 2018–2025

We tested a 90% bound for all five daily closing-price moves after a Nasdaq Composite origin. Inputs used prior-date ALFRED archives. Calibration used only complete path labels already released and each past origin's issued volatility scale. The final test has 385 fresh, nonoverlapping five-session outcomes in 2018–2025. The main empirical bound used the last 104 available paths. It had 4.94% more mean quantile loss than a normal path bound with the same scale. The paired 95% eight-origin block interval for error reduction was -11.45% to +1.36%. Thus it gave no clear score gain. Its full-path coverage was 90.65%, versus 87.01% for the normal path rule. A bound calibrated only on the final close had 92.21% final-close coverage but only 77.14% full-path coverage. Empirical path coverage was 83.17% in low-scale cases and 93.31% in high-scale cases. Planned delay, history, block and label checks gave no clear empirical gain. The source has a long update gap; 16 stale test slots remain in a separate case. This is a historical risk benchmark on shared market data. It does not show conditional coverage, return-direction skill or trading profit. The package preserves code, 6572 forecasts, 2756 calibration records, all scores and uncertainty cases, source requests and hashes, negative results, figures and repeat instructions. Independent checks passed. A full repeat matched ten public and three restricted files byte for byte. Raw provider prices and individual realized returns stay outside the public package under source terms.

stock market · path risk · quantile forecast · point-in-time data · chronological test · negative result · Nasdaq Composite · forecast calibration