Cycles set by early orbital periods reduced later timing error in 60 of 86 tested Kepler records

We tested timing cycles for known Kepler planets near resonance. The cycle frequency comes from the early orbital periods of each adjacent pair. We used public Holczer timings and a fixed NASA object snapshot. Early times end at BJD 2455900. Each cycle model fits a line, a sine term, and a cosine term at that fixed frequency. No later timing offset enters a fit. Cuts on data quality, sample size, timing span, resonance distance, and cycle coverage give 86 records at 37 stars. Each record is one target in one of 43 pairs. The cycle has smaller later weighted root mean square error than a line in 60 records. The median error ratio is 0.975. A 95 percent interval from resampling whole stars is 0.929 to 0.994. A secondary rule selects 24 records at 15 stars with early delta BIC at least 10. All 24 improve. Their median ratio is 0.834, with an interval of 0.690 to 0.895. The selected cycle beats both control frequencies, 0.5f and 1.5f, in all 24 records. Some selected forecasts fail when the cutoff or quality rules change. Catalog discovery, status, flags, and templates use full mission information. Thus this is a retrospective test. It does not establish a new planet or current forecast accuracy. The package preserves source facts, hashes, exclusions, coefficients, transit forecasts, bootstrap draws, simulations, figures, code, and repeat instructions.

Kepler · transit timing · resonance · ephemeris · forward prediction · fixed frequency · reproducible benchmark