Periods of 727 ASAS-3 Mira variables changed by 1.1 percent rms from 2000-2009 to 2013-2025 as random wander; early trends did not continue, and only NSV 08001 exceeded the wander

Question: do Mira variables show period changes over two decades that are larger than random period wander, as expected after a thermal pulse? We measured the periods of the 2,896 ACVS Mira stars in independent time windows of ASAS-3 (V, 2000-2009) and ASAS-SN Sky Patrol v2 (V and g, 2012-2025), with a cycle-jackknife error, and used Gaia DR3 long-period-variable periods as a third check. For 727 stars that pass strict quality and saturation cuts, the period change over 15.7 years has a robust scatter of 1.23 percent with a median error of 0.47 percent; a Gaussian excess of 1.19 percent rms fits the data (likelihood ratio 680), and the mean change is -0.10 percent. Inside ASAS-SN, adjacent windows 3.1-3.6 years apart agree with the errors, but windows 6.7 years apart show a 0.76 percent excess, so the periods wander with time. The 2001-2007 trend does not predict the 2015-2021 trend (322 stars, Spearman 0.05, p = 0.35), and the AAVSO rates of Templeton et al. (2005) do not predict the new changes for the 8 weak-case stars that pass our cuts (1 of 8 with the same sign). Against the wander model, 8 stars have |z| > 3 (2.0 expected); after Bonferroni correction only NSV 08001 (265.2 to 248.3 d, -6.6 percent, Gaia 252 +/- 10 d) remains as a fast-change candidate, and one other outlier (V1596 Sgr) is an ASAS-3 artifact. Four stars with consistent 3-5 percent changes (RY Hyi, V1055 Oph, NSV 09906, ASAS J054812-3200.1) are consistent with the wander. Limitations: southern sky only, 15.7-year baseline, saturated bright Miras (R Aql, R Hya) excluded, and residual wrong ASAS-3 periods for faint bulge stars.

Mira variables · period change · thermal pulse · AGB stars · ASAS-3 · ASAS-SN · Gaia DR3 · long-period variables · forecast test · random walk