A fixed Gliese 12 b signal lowers error in two of three instrument tests with 2 m/s extra noise

We test whether the transit clock of the known planet Gliese 12 b predicts public radial velocities in an omitted instrument. We use 199 retained measurements from HARPS-N, ESPRESSO, and CARMENES. The period is fixed at 12.761417 days. The phase is fixed from a published fit to transit data. We fit a signed circular amplitude to two instruments and apply it to the third. The test instrument supplies only a constant offset. With quoted variance plus (2 m/s)^2, the mean squared error gains are -0.01536, +0.25793, and +0.25118 (m/s)^2. The mean across instruments is +0.16458 (m/s)^2. A common fit gives 0.6221 +/- 0.2566 m/s under this fixed covariance. The mean gain stays positive in 68 specified calculations that change noise, activity terms, reductions, source flags, clock values, or orbital shape. These calculations are related tests, not independent confirmations. An activity regressor reduces the gain. The result supports a small signal at the known transit clock within the tested model set. It does not give a new planet discovery or a new precise mass. The package preserves exact source data, row selections, all predictions and scores, figures, code, and reproduction instructions.

Gliese 12 b · radial velocities · instrument prediction · transit clock · fixed noise covariance · reproducible analysis