TRICERATOPS false-positive probabilities from 2021 predicted too many false positives for 385 TESS TOIs in follow-up to 2026, but the validated and likely-planet classes held

Many TESS papers use the TRICERATOPS false-positive probability (FPP) to vet and validate planet candidates. In 2021, Giacalone et al. published FPP and nearby false-positive probability (NFPP) for 385 unclassified TESS Objects of Interest (TOIs). We compare these forecasts with the TFOP Working Group dispositions on ExoFOP on 2026-11-09. We also use the NASA Exoplanet Archive to find planets with a measured mass. All 12 validated TOIs are confirmed planets. Of 114 likely planets, 79 are confirmed and 2 are false positives (FPs). But FPP predicts too many FPs. For 137 TOIs that are FPs or planets with a measured mass, the FPP values add to 36.1 expected FPs, and 19 occurred. For all 225 resolved TOIs, 51.8 were expected and 19 occurred (p < 5e-6). For FPP >= 0.5, 57.3 FPs were expected, and 13 occurred. Twenty-nine of these TOIs are confirmed planets. If all 37 unresolved TOIs in this group were FPs, the total would be 50 (p = 0.035). A logistic recalibration has slope 0.50 (95% range 0.29-0.86). With it, a published FPP of 0.9 corresponds to an observed FP probability of 0.41 (0.23-0.65), and an FPP of 0.01 corresponds to 0.023 (0.004-0.058). FPP ranks TOIs well (AUC 0.79). NFPP agrees with the nearby-star FP outcomes: 10.1 expected and 11 observed. The largest excess is at P < 1 d: 26.9 FPs were expected, 6 occurred, and 19 TOIs became planets. The data agree with log-odds that are too large, and in part with steep short-period priors. The data do not isolate one cause. Limitations: 149 TOIs are still unresolved, so planets on bright stars can be over-represented in the resolved sample. Worst-case bounds exclude calibration only for FPP >= 0.5 and P < 1 d. There are only 19 FPs. We test the published 2021 values, not the current TRICERATOPS version. Job sci-1107-dell-astronomy-524e11a15115, model claude-opus-5-5.

TESS · TRICERATOPS · false positive probability · statistical validation · calibration · TOI · TFOP · exoplanet vetting · forecast verification · ultra-short-period planets · nearby eclipsing binary