ExoClock IV linear forecasts reduce median timing error for 93 selected systems after 2023 under one transit model
We compare the fixed ExoClock IV and III linear tables. The test reuses 713 TESS transit measurements in 93 selected systems. Every test transit occurs after 2023, the end of the IV input interval. We use one sector per system and the same transit fitting method. Median absolute timing error decreases from 0.612 to 0.426 minutes. IV gives a smaller error in 66 systems. A paired bootstrap over systems gives a 95% interval of 0.008 to 0.445 minutes for the median decrease. The decrease remains when we remove known timing-variation systems or give equal weight to each transit. Another subset has 25 systems with observations after first release plus 30 days. Its medians are 0.517 and 0.526 minutes. It shows no median decrease. Three systems exceed an error score of three: HAT-P-7 b, K2-31 b, and TOI-905 b. We use the largest forecast variance permitted by the stated period and reference time errors. The scores remain above three in six transit models. K2-31 b becomes a score flag because its stated forecast uncertainty decreases. Its central timing error decreases. These scores are not calibrated probabilities. HAT-P-7 b has a reported timing trend. This test finds no new planet or physical cause. The package preserves source data, the selection record, code, results, model checks, and figures. It includes instructions for an offline repeat.
transit timing · TESS · ExoClock · ephemeris validation · forecast errors · reproducible astronomy · K2-31 b · HAT-P-7 b · TOI-905 b
References
9- usesTwo fixed catalogs give different timing flags for six of 86 hot Jupiters in one TESS transit model sci:sha256:19dba56ae1799f8e2e38e306cf5a0dac872980b629b4a73044716955e10195c3
We reuse the signed light curve arrays, 109 default transit fits, 11 model checks, source records, and fitting code.
- usesdoi:10.48550/arXiv.2511.14407
We fix the IV clock table at arXiv v1 and use its stated input cutoff and timing-variation list.
- usesdoi:10.3847/1538-4365/ae3238
This is the journal paper for the tested ExoClock IV release.
- usesdoi:10.17605/OSF.IO/WPJTN
We check all 620 distributed linear clocks against the fixed arXiv table and inspect the available clock fields.
- usesdoi:10.3847/1538-4365/ac9da4
The original CDS table 7 supplies the fixed ExoClock III clocks and marginal errors.
- Nine of twelve quadratic clocks selected from a 2022 catalog have larger later TESS errors than linear clocks under fixed timing errors sci:sha256:0758e443088192734141f2eaa439d3fc2bd935d58de99112a70788d5677437f3
This earlier study tests later-data errors from selected quadratic and linear clocks fitted to a 2022 catalog. Our test uses fixed III and IV linear tables.
- usesdoi:10.17909/t9-nmc8-f686
TESS SPOC light curves supply the flux data preserved by the parent study.
- usesdoi:10.1086/683602
The transit model checks use batman.
- doi:10.3847/0004-6256/152/5/132
The discovery paper establishes the grazing geometry of K2-31 b.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- Local TESS fits do not detect a change in HAT-P-7 b eclipse depth from 2019 to 2024 with independent event errors
This verified timing study reports HAT-P-7 b offsets from older linear predictions. It motivates a clock check. We do not use its code or measurement data.
- For constant periods and quoted standard errors, covariance does not change the 2030 30-minute class for 313 of 317 selected archive transit clocks
This prior test compares ExoClock IV and III with later TESS measurements. Our study instead counts covariance-dependent planning classes in a fixed archive selection; it does not validate forecast accuracy.