Two transit models with quadratic baselines give KELT-9 b mean times 42–53 seconds apart in six TESS sectors

We fit 88 KELT-9 b transits in six public TESS sectors from 2019 to 2024. Each event has a symmetric fit and an asymmetric fit. Both fits use the same data and free quadratic baselines. Center, width, edge scale, curvature, and depth are free. The symmetric model gives earlier sector mean times by 42.48 to 52.50 seconds. The paired time difference has a fitted slope of -0.76 seconds per year. Its conditional 95% Student t interval is -2.76 to +1.25 seconds per year. This interval assumes a linear mean and independent Gaussian sector errors with equal variance. The 2024 mean difference is 3.78 seconds below the 2019 mean difference. A paired event bootstrap gives a 95% interval from -11.86 to +4.28 seconds for this change. The test does not resolve a change in the difference. A control with fixed stellar pulse terms reduces the sector offsets to 26–32 seconds. An SAP-flux control gives a negative end difference under event resampling. Its sector trend interval includes zero. These intervals use different error assumptions. With symmetric synthetic inputs, the nominal 95% sector interval contains zero in 93.83% of finite-pool draws. These results compare two specified empirical profiles. Neither center is assumed to give true orbital conjunction. The package includes model controls, synthetic recovery tests, all source files, provenance, event measurements, code, figures, and repeat instructions. It extends an earlier SCI width study with a paired timing dataset. It does not establish orbital decay or a unique physical cause.

KELT-9 b · TESS · transit timing · transit asymmetry · model sensitivity · reproducible astronomy · public data