An earlier 31-day curve predicts later HD 86728 velocities in two time tests with Gaussian noise

We test whether an earlier circular velocity curve predicts later public NEID data for HD 86728. We use 220 exposures on 162 nights from 2021 to 2025. We carry period, amplitude, phase, and extra-scatter uncertainty into two time tests. A 58-night source predicts 104 later nights across the 2022 thermal cycle. A 152-night source predicts ten nights from October 2024 to February 2025, beyond the discovery data span. We remove a separate mean for each instrument era. The period prior spans 20 to 40 days. Under independent Gaussian noise, the earlier curve reduces later RMS error from 1.90 to 1.34 m/s and from 1.83 to 1.22 m/s. Natural-log predictive density ratios over a flat later curve are 43.133 and 5.668. Ratios over a later curve with independent amplitude and phase are 4.491 and 3.913. All stated prior, era, and fixed correlated-noise controls retain positive ratios. Removing any one later night also retains positive ratios. Direct Gaussian checks agree with the calculation to better than 0.000000001 in log density. The tests share data and are retrospective. Their ratios are not independent evidence or planet probabilities. The result supports persistence of the known signal under the tested circular and Gaussian models. It does not establish a new planet or exclude all stellar and instrument effects. The package preserves public inputs, full data provenance, code, posterior grids, figures, and reproduction instructions.

HD 86728 · NEID · radial velocity · time prediction · public data · Gaussian noise · Bayesian model comparison