Do tides only tip critical faults?

We took every M≥2.5 earthquake in the USGS ComCat catalogue for Oklahoma (2009–2024) and California–Nevada, and removed aftershocks within one day and 10 km of a larger event. For each origin time we computed the equilibrium solid-earth tide from the Moon and Sun, band-passed it to the diurnal and semidiurnal band, and read its phase from the analytic signal. A Rayleigh test then asks whether origin times cluster in tidal phase. We tested three things the claim requires: phase locking in 100-event blocks through the Oklahoma injection cycle; the same in California as a tectonic control; and phase concentration in the 90 days before versus after each M≥4.5 event.

• Weak locking, not none. The median block concentration is R = 0.099, below the 0.125 expected by chance for 100 events, and 10.5 per cent of blocks reach p < 0.05. Some windows do lock; most do not.
• No rise with pore pressure. Mean R is 0.108 in 2009–2015, while injection was rising, and 0.086 in 2017–2024, after the Oklahoma Corporation Commission cut back disposal volumes. The difference is in the predicted direction but not significant (Mann–Whitney p = 0.071), and the time trend is flat (Spearman ρ = −0.14, p = 0.23).
• No precursor. Across 185 M≥4.5 events, mean phase concentration is 0.0724 in the preceding 90 days and 0.0723 in the following 90 days (Wilcoxon p = 0.34). Tidal sensitivity does not announce the event.

The claim makes tidal sensitivity a state variable that tracks distance to failure. In the one place where pore pressure changed by a known amount over a decade — Oklahoma — the signal does not follow it, and it does not rise before rupture. What locking exists is a property of particular windows, not a precursor. The result is a null, and a useful one: a free catalogue-derived precursor would have been valuable, and this catalogue does not contain it.