Local two-planet forecasts with fixed initial orbit planes do not cover the 2025 TOI-216 times in joint 95% regions

We tested whether nonlinear orbit uncertainty explains an earlier TOI-216 time forecast error. We used 60 TESS transit times from 2018–2023 to fit the two known planets. We reserved 16 times in 2025 for a test. Stellar values and initial orbit planes were fixed. We sampled two local orbit regions and two assumed timing-scatter values. A polynomial model proposed samples. Exact orbital integrations supplied likelihood weights and future times. Each setting used two independent ensembles and 8,000 exact draws. With a 1.7 minute scatter and the smaller local bound, 10 of 16 test times lie inside separate 95% predictive intervals. The observed joint quadratic distance is 121.8; its simulated 95% limit is 26.1. The test vector remains outside the joint 95% region with the larger bound and with a 3.4 minute scatter. Nonlinear sampling thus retains the correlated error seen in an earlier linear check. Local boundaries still affect the orbit distribution. This retrospective result applies to the stated local priors, fixed physical values, inherited timing extraction, and independent noise. It does not reject all two-planet models or establish a new planet. The package gives joint draws, code, figures, provenance, and reproduction instructions.

TOI-216 · TESS · transit timing variations · posterior predictive check · nonlinear orbit uncertainty · importance sampling · forecast validation