An early 20-day fit reduces later Tau Ceti ESPRESSO error after season means are removed

We test whether a reported 20-day velocity signal transfers between observing seasons of Tau Ceti. We fit a fixed 20.00-day circular signal to the first 50 nights of public ESPRESSO19 data. We predict the last 59 nights and remove each season mean from both the data and the prediction. For full template matching with telluric correction, the later root mean square error falls from 0.60046 to 0.57088 m/s. The gain is 0.02958 m/s. It remains positive in eight reductions under the main covariance, and in eight noise models for the main reduction. Every tested period in the published 19.99–20.02 day interval gives a positive gain. Under the main fixed Gaussian noise model, 208 of 50,000 draws without a coherent signal give a larger gain. With a coherent signal of semi-amplitude 0.49 m/s, 11.23% give this gain or less. This fraction varies from 1.86% to 21.69% across the noise checks. The result shows limited transfer of a velocity component. It does not establish a planetary source. We use existing public data. The noise scales come from an earlier study of the full data. The package preserves source files, provenance, a plan written before velocity inspection, results for each night, figures, code, and reproduction instructions.

Tau Ceti · HD 10700 · ESPRESSO · radial velocity · planet candidate · stellar activity · prediction test · public data · reproducible research