Annual terms and season offsets remove 64–74% of 348.6-day signal information in linear GJ 163 HARPS fits

We test the reported 348.6-day signal in public HARPS-RVBank velocities for GJ 163. Quality selection gives 203 spectra and 201 night means over 18.15 years. The baseline linear model has separate instrument offsets and free sine and cosine pairs near four reported periods. We use fixed independent-noise weights with 2 m/s extra scatter per night. Free season offsets and annual sine and cosine terms retain 26.3–35.8% of baseline signal information at 348.6 days. These bounds include every signal phase and use only observation times and errors. The fitted amplitude is 0.502 m/s in the baseline model and 1.378 m/s with season offsets. It is 1.248 m/s with both season and annual terms. Adding annual terms to the season model reduces the chi-square improvement from 16.58 to 5.36. The joint model has a nominal 95% amplitude upper bound of 2.589 m/s. This bound is above the earlier central value of 2.47 m/s. Bounded fits of the other periods and harmonic controls also show dependence on the nuisance model. We preserve the raw data snapshot, source records, code, figures, result tables, and simulation checks. This exploratory result applies to the stated linear templates and noise weights. It does not confirm or reject a planet or identify a physical cause of an annual effect.

GJ 163 · radial velocity · HARPS · HARPS-RVBank · annual signal · season offsets · signal information · planet candidate · reproducible analysis