Four fixed-period signals improve predictions in five omitted HD 32564 seasons in two HARPS reductions

Do four previously selected HD 32564 velocity signals help predict omitted observing seasons? We test circular signals at 2.5566, 11.0624, 23.6067, and 49.2110 days. The inputs are 205 public HARPS spectra on 193 nights. Each fit omits one season and uses all other nights. Five test seasons contain 184 nights. We compare a complete model with four models that each remove one signal. All four signals improve every test season in both SERVAL and DRS reductions. The pooled weighted prediction errors are 1.817 and 2.357 m/s. Fitting nearby periods from the training velocities preserves all positive gains. Linear activity terms give one failed season for the shortest DRS signal. A 122.4-day nuisance term gives one failed season for the 23.6067-day DRS signal. The 2.5566-day model has lower pooled error than four tested daily aliases in both reductions, with fixed or fitted periods. The earlier study selected the periods from the full data set. The two reductions use the same spectra, and later nights remain in training. Thus these are conditional prediction tests, not blind forecasts or independent planet confirmations. The package preserves the code, input data, source hashes, main test predictions, tables, figures, and reproduction steps.

HD 32564 · HARPS · radial velocity · season prediction · daily alias · stellar activity · reproducible research