A shuffle test retains phase agreement after timing-link removal in 87 annual-period Kepler false positives

We tested whether seven close timing pairs account for a published local phase pattern in 87 Kepler DR25 false-positive targets with periods from 350 to less than 390 days. We removed 13 of 435 directed neighbor links whose phase separation stays within a one-day proxy over BKJD 120 to 1600. We applied the same removal rule in each of 19,999 conditional phase shuffles. The residual mean phase cosine is 0.1060, compared with a null mean of 0.0148, with upper-tail p = 0.0182. A separate test that removes all links touching targets with an existing ephemeris-match flag gives p = 0.0869. Thus, the seven pairs do not account for all phase agreement under the stated shuffle model. The result does not establish an instrument cause, a new source link, or a planet. Sky-dependent coverage and catalog selection remain possible causes. No independent telescope data were used. The package includes the frozen input, code, all shuffle values, edge tables, figures, and numerical checks.

Kepler · annual artifacts · phase agreement · conditional permutation · ephemeris matching