ASASSN-14ko flares from 2022 to 2026 peak up to 21 days later than the published constant period-derivative ephemeris; random period changes of about 0.6 days per orbit fit the peak times better

ASASSN-14ko is a repeating nuclear transient with optical flares at a period of approximately 110 days. Payne et al. (2021, 2022) fitted the 2014-2021 peak times with a constant period derivative (Pdot = -0.0026), which is the main evidence for steady orbital decay in this candidate repeating partial tidal disruption. We did an out-of-sample test. We measured 8 flare peaks in ASAS-SN Sky Patrol g-band aperture photometry (2022 February to 2024 October) and 1 peak in TESS Sector 98 full-frame images (2026 January 1). We used two peak methods (template fit and kernel maximum). Our pre-2022 peaks agree with the published peaks (RMS 0.76 d). The flares come progressively later than predicted. The offset is -10 to 0 d in 2022 and +7 to +13 d in 2023-2024. The TESS peak (MJD 61041.3 +/- 0.5) is 21.3 d after the published prediction. The difference image puts the flare on the nucleus. Constant-Pdot models trained on the pre-2022 peaks fail as forecasts: chi-square 26-67 for 9 flares (p = 2e-10 to 0.002). With uncertainties scaled to the published widths, p is 7e-6 for template peaks and 0.05 for kernel peaks. A Kalman model in which the period gets a random change of approximately 0.6 d at each orbit forecasts the 2022-2026 peaks better (log score better by 4.6-10.0) and is consistent with the kernel peaks (p = 0.085). The pre-2022 peaks alone already prefer these kicks (p = 0.04). For all 30 peaks, kicks are preferred (p = 0.009-0.045), and a steady period decrease is then not required (p = 0.10-0.16; Pdot 95% interval -0.0043 to +0.0009). The period became shorter from approximately 113 d to 108 d, but not at a constant rate. The g-band flare amplitude shows no monotonic decrease. We predict the next peaks for 2026 November 10-16 and 2027 February 23 to March 2 (published ephemeris: 2026 October 18 and 2027 January 30). Limitations: aperture photometry on unsubtracted images, no public data after 2025 January except one TESS sector, peak times of structured flares depend on the method by up to 8 d, and 30 peaks cannot fully separate random kicks from smooth higher-order changes.

ASASSN-14ko · repeating nuclear transient · partial tidal disruption event · periodic flares · period derivative · timing noise · out-of-sample forecast · ASAS-SN · TESS · Kalman filter · ESO 253-G003