Later measured transits reduce the best periodic box signal for two of 80 selected Kepler planets in an equal white noise model

We test whether later measured transits can reduce the largest expected signal-to-noise ratio of a periodic box filter. We use public Holczer et al. (2016) transit times for 80 selected confirmed Kepler planets with periods of at least 20 days and significant source timing variations. The model has equal depths and durations, complete independent event records, a known baseline, equal white noise, and fixed cycle labels. Early means the first floor(N/2) retained events; full means all N events. We optimize period, phase, and width by complete contact enumeration with exact integer arithmetic. In units of ideal single-event signal-to-noise, the full/early ratios are 0.859662 for Kepler-603 b and 0.901951 for Kepler-30 b. The other 78 planets gain signal. The same two cases decline with width fixed to the true duration. Both declines remain when the assumed duration changes by 25%. A conservative test with independent center shifts of up to three printed timing errors plus rounding gives upper bounds of 1.038265 and 1.024360 on the full/early ratios. Thus it does not certify either decline throughout those intervals. The result is a benchmark at the printed measured centers, not a detection test in flux data or a new planet claim. The package preserves the source numerical tables, provenance, all 320 primary optima, sensitivity results, exact witnesses, independent checks, figures, and reproduction instructions.

Kepler · transit timing variations · periodic transit search · signal-to-noise · exact optimization · benchmark · public astronomical data