Kepler fits leave less than 0.02 m/s RMS in 39 GJ 1061 orbit tests at 210 public times with stated errors

Can the public observing times for GJ 1061 separate weak planet interactions from three free Kepler orbits? We integrate 39 stated orbit states at 184 public HARPS times and 26 published ESPRESSO times. The states use published best values, marginal medians, three chosen epochs, 32 random phase sets, and one coordinate control. We assign 1.41 m/s errors to HARPS and the published errors to ESPRESSO. We fit 15 Kepler parameters and three offsets to each simulated curve. Every main test leaves less than 0.02 m/s residual root mean square. In the first test, the scatter relative to the initial Kepler orbits is 2.49 m/s. The fit reduces the residual RMS to 0.01296 m/s and the squared Gaussian distance to 0.02418. For fixed known Gaussian errors, a test with at most 5% false positive probability at every Kepler orbit then has at most 6.82% power against this interacting state. Six further controls change eccentricity or mass. Numerical checks use a second integrator and analytic one-planet curves. The result is conditional on the stated orbit states, epochs, errors, and model. It does not fit measured velocities or reassess the published mass posterior. The package preserves public source data, hashes, all orbit inputs, paired velocity curves, fitted parameters, numerical checks, figures, and repeat instructions.

GJ 1061 · radial velocity · planet interactions · Kepler orbits · detection power · public data · reproducible benchmark