Early orbital beat models reduced later timing error for 17 of 20 selected Kepler pairs in a 70/30 test

We tested whether early orbital beat periods help predict later planet transit times. We used public Kepler timing data and a saved NASA planet catalog. We formed adjacent confirmed planet pairs near first-order period ratios. Each pair needed at least one predicted beat cycle in its early record. We fitted a line, a quadratic, and a line with one fixed-period sine and cosine. Fits and selection used the first 70% of the pair's time span. Scores used the last 30%. Of 62 eligible pairs, 20 pairs at 19 stars passed the early support rule. The cycle model gave less later error than the line for 17 of these 20 pairs and less error than the quadratic for 18. The median pair error ratios were 0.816 and 0.718. A bootstrap that kept pairs at one star together gave 95% percentile ranges of 0.720 to 0.909 and 0.569 to 0.804. Across all 62 eligible pairs, the median cycle-to-line ratio was 0.988. Changed splits, weights, and quality filters retained the selected-sample result. Independent parser and fit checks passed. The package preserves source tables, provenance, code, predictions, controls, and figures. The result is a finite forecast benchmark within the Kepler record. Source measurements and confirmed status can use full-record information. This test does not establish a new planet, a unique physical cause, or forecast accuracy across a later mission gap.

Kepler · transit timing variations · orbital beat period · chronological validation · ephemeris · reproducible benchmark