Local stellar rotation terms did not reduce prediction error for 593 later Kepler-1520 b transits in a fixed linear test

We tested whether a measured stellar rotation wave adds information about Kepler-1520 b transits beyond recent transit history. We used 17 public Kepler DR25 light curves. Local fits measured a sine term and cosine term at a period of 22.9 days from flux outside transit. The target orbit was excluded from their fit and normalization. The stellar fit used flux before and after the event. Thus this is a conditional test, not a blind forecast. We fitted two linear rules on 525 events in quarters 1 through 8. Their coefficients stayed fixed for 593 events in quarters 9 through 17. Adding the stellar terms increased mean squared error by 0.190%. Root mean squared error rose from 606.033 to 606.608 parts per million. A conditional 95% range for the error reduction was -1.965% to +1.423% with 35-orbit blocks. Ranges from longer blocks and separate early and late resampling also included zero. Seven well-sampled cases gave no detected gain. A strict quality check had only 16 fit events and was unstable. A constant-transit control gave a 10.93 parts per million RMS difference between smooth and flat stellar curves. These results limit the stated fixed linear rule. They do not exclude nonlinear or changing stellar effects. Earlier work already tested rotation modulation, local flux and past-only harmonics. This contribution supplies a distinct two-sided conditional test and its measured harmonic ledger. The package preserves raw data, provenance, source code, all predictions, controls, figures and offline reproduction instructions.

Kepler-1520 b · KIC 12557548 · disintegrating planets · stellar rotation · conditional prediction · transit variability · reproducible astronomy