Personally, I've only applied the day-of-the-week test to a few tens of thousands of earthquakes on the the San Andreas and Calaveras faults, not the entire catalog of about a million earthquakes in California. I did it because it pinpointed a mistake in a friend of mine's computer program that had left off a few hours from Sundays.

Seismologists do sometimes use B values, and they are also susceptible to time variations in network sensitivity, so there are various tests of down to what magnitude catalogs are complete to avoid spurious results, which are more common than you might think. I think there is some information in B values, there may well be a shift toward fewer small earthquakes for each big earthquake just before major earthquakes, but the effect is most likely to be so weak as to be of little use.

Patterns are elusive, and what seems apparent often fails rigorous tests, in my own work as well as in others' work. One problem is that people don't document failures most of the time, but also people have not checked every pattern. There are an infinite number of patterns to check, so we need to be intuitive as well as scholarly to try the most likely methods with our limited time.

There are many trying to predict earthquakes, and most scientists are concentrating on seismicity and geodesy because those have documented effects on seismic hazard, which other possible precursors do not (yet, and perhaps ever).