Later data gave no clear daily risk score gain for an exponential Nasdaq rule in a 2015–2025 ALFRED replay

We tested how later data change a daily Nasdaq Composite risk test. We held three fixed rules and all target labels constant. Three data arms used dated ALFRED records, later numeric values on the same dates, or the complete later history through each old cutoff. All inputs preceded the entry session. The test has 2,766 target returns in 2015–2025. For the primary exponential rule with decay 0.94, dated loss minus complete-history loss was -0.000754 QLIKE units. The paired 95% interval from 20-session blocks was -0.002255 to +0.000595. Later data gave no clear gain. Changed values and added observation support both had uncertain effects. Later data reduced mean loss for a 252-square rule but raised it for a 20-square rule. The exponential rule had the lowest mean loss in every arm. A known feed delay produced 76 stale-input targets and about 57% of the net primary score change. The test concerns one archive channel; other services could have supplied timely prices. The package preserves all 24,894 forecasts, data-date audits, source hashes and requests, code, 15,000 paired resampling rows, figures, checks, and negative findings. Raw prices and outcome ledgers remain separately retained under source terms. This is a further analysis of a known sample, not an independent replication. It establishes no return direction skill or trading profit.

Nasdaq Composite · point-in-time data · data revisions · forecast evaluation · volatility · ALFRED · QLIKE · feed delay · negative result