Reported cycles increase 2017–2026 log-flux forecast error against OU noise in fixed-index data for 10 of 11 selected gamma-ray sources

We tested whether reported cycles improve later gamma-ray predictions for all 11 high-confidence candidates in Peñil et al. (2020), Table 1. We used public NASA Fermi LAT monthly photon flux in the 0.1–100 GeV band. We compared a constant mean with Ornstein–Uhlenbeck (OU) covariance against the same model family plus a sine curve at the reported period. We fitted each model to bins ending by 2017-11-01 UTC. We scored frozen forecasts on valid bins after that cut and ending by 2026-10-01 UTC. The spectral products used the same valid times. In fixed-index data, the cycle increased log-flux mean squared error for 10 of 11 sources. The mean relative increase was 19.3%, with a conditional 95% time-block interval of 14.7–28.2%. With free-index data, all 11 errors increased; the mean increase was 14.3%. Early period refits and linear-flux scores also gave a mean error increase. The package preserves raw responses, source matches, hashes, quality decisions, code, predictions, and figures. An independent Kalman calculation checks all 110 fitted models. This retrospective test uses current archive fits with some catalog information from after the early-data cut. The result applies to the stated stable sine prediction rule and selected detections. It does not exclude transient cycles or binary black holes.

Fermi LAT · gamma rays · active galactic nuclei · reported periods · prediction · Ornstein–Uhlenbeck model · correlated noise · public data · reproducible research