Adding a 647.6-day signal to HARPS fits raises error on 65 corrected ESPRESSO nights for HD 20794 at fixed periods

We tested whether a 647.6-day velocity curve fitted to HARPS reduces error on public ESPRESSO data for HD 20794. We fitted 746 HARPS nights with two circular inner signals, then added a circular or eccentric outer signal. Periods were fixed at 18.3140, 89.68, and 647.6 days. The ESPRESSO test used 65 nights from 2019 to 2023, a fixed Gaia motion correction, and one fitted constant offset per model. All orbital coefficients were fixed before the test. Adding the circular signal raised test root mean square (RMS) error from 0.691 to 0.837 m/s. The eccentric signal gave 0.803 m/s. Mean squared error gains were -0.2229 and -0.1668 square metres per square second. Both additions improved the HARPS fit. Matched DRS data, two noise-floor controls, and local period fits also gave higher test error. With four ESPRESSO season offsets, the gains changed to +0.0074 and -0.0146 square metres per square second. Those offsets removed 83.3% and 81.4% of the predicted change variance. The result applies to the stated products, correction, and model. It does not refute the planet. Published periods and upstream data processing use the full record, so this is not a blind forecast. The package preserves source tables, provenance, code, fits, every prediction, scores, figures, and numerical checks.

HD 20794 · radial velocity · HARPS · ESPRESSO · prediction error · perspective acceleration · public data