A companion depth correction with gain one increased the mean Q10–Q16 error ratio for 22 selected Kepler targets

We tested whether one planet can supply a transit depth correction for another planet at the same Kepler target. Fixed precision and coverage rules select 22 pairs with confirmed archive labels. Q2–Q9 sets each planet's constant and four season means. Q10–Q16 supplies 136 paired quarter measurements. The correction adds the companion's later depth residual with gain one. Against the recipient season baseline, the geometric mean RMSE ratio is 1.282. Its conditional 95% target bootstrap interval is 1.095–1.502. Only 6 of 22 targets have less error. The matched constant comparison gives 1.289, with an interval of 1.110–1.508. Every planned alternative rule gives a mean ratio above one, although some intervals include one. Simulations show that the same code removes a strong shared change and adds noise when that change is absent. An independent parser checks all 69,914 depth records against the original catalog. The package saves inputs, source hashes, the local plan, code, selection records, predictions, checks, and figures. The test uses full-mission catalog templates, flags, and current planet labels. It measures later catalog estimation error in a selected sample. It does not establish a physical cause or a new planet.

Kepler · transit depth · multiple planet systems · calibration · seasonal variation · reproducible benchmark