Early curvature marked 13 of 20 window failures in 711 selected Kepler planets with fixed durations and bounds of three reported timing errors

Can early transit timing curvature warn of later failures of a constant-period forecast? We tested a fixed rule on public timings for 711 selected confirmed Kepler planets at 623 stars. Early fits used expected days at or before BJD 2455700. A failure requires at least three selected later transits outside full windows twice the fixed transit duration. Each center can move favorably by three reported timing errors. A quadratic coefficient divided by its scatter-scaled standard error gave a warning at a score of at least 3. It marked 13 of 20 failures and 24 planets without this outcome. Failure rates were 35.1 percent with a warning and 1.04 percent without one. A quadratic forecast gave 77 failures, compared with 20 for the line. Using only timings with no source outlier flag left 14 failures among 665 planets; curvature marked 12. These are retrospective results conditional on catalog measurements, fixed event numbers, and fixed symmetric durations. The error bounds are not joint confidence levels. The package preserves source bytes, forecasts, selection ledgers, code, figures, and an offline repeat that reproduced all 22 numerical result files byte for byte. No new planet or physical timing signal is claimed.

Kepler · transit timing · observation windows · ephemeris · forecast validation · timing curvature · reproducible research