Published gamma-ray periods of 54 Fermi-LAT blazars do not predict flux maxima in 2017-2026 LAT data better than chance; at most 22-36% of candidates can have a coherent period

Many studies report 1-5 year periods in blazar gamma-ray light curves as evidence for binary black holes. A coherent period must predict later flux maxima. We tested published periods out of sample with Fermi-LAT Light Curve Repository 30-day fluxes to October 2026. Set B: 59 dated forecasts for 30 blazars (Rico et al. 2025; Penil et al. 2025), made with data to 2020-12-10 and evaluated by Penil et al. 2026. In 18 of 48 strictly scorable forecasts (0.38), the one-cycle maximum was within 0.2 period of the predicted date. A phase-shift null on the same light curves gives 0.40 (90% range 0.27-0.52); a coherent period with the training amplitude gives 0.64. Set A: sinusoids at the 24 Penil et al. 2020 periods, fitted to 2008-2017 data, hit 31 of 87 later maxima (0.36; chance 0.37, red-noise simulation 0.39, coherent 0.71). Control: the best training period of 180 bright blazars hit 204 of 546 (0.37; chance 0.38), and training periodogram power did not predict forecast skill. Forecast correlations (mean r +0.002 and +0.043) and train-test phase differences agree with random phases. A mixture model limits the fraction of candidates with a coherent period at the training amplitude to 0.22 (Set A) and 0.36 (Set B) at 95% (0.36 and 0.56 at half amplitude). Red noise gives hit rates near 30-40% with a 0.2-period window, so the 32% agreement reported by Penil et al. 2026 is at the chance level; their rule as written always passes. PG 1553+113 and S5 1044+71 agree with their forecasts (2/2 and 1/1 hits) but are not significant after selection, and the 2008-2017 sinusoid of PG 1553+113 is 0.21 cycle out of phase in 2018-2026. Limitations: LCR 0.1-100 GeV fluxes differ from some original analyses; few cycles per source; H1 depends on the assumed amplitude.

blazars · gamma-ray periodicity · quasi-periodic oscillations · Fermi-LAT · Light Curve Repository · binary supermassive black holes · red noise · out-of-sample forecast · PG 1553+113 · S5 1044+71 · null result