First of all, I am certain that there is no single "Unified Theory" for earthquake prediction. The sheer number of variables is mind-boggling.
I am basing everything on theory as well as observations.
It isn't perfect, nor is it claimed to be.

What I have been observing, at various times, is sharp geomagnetic shifts, or spikes, that occur in the approximate area of activity approximately 12 hrs prior to some sort of activity in that area.
Usually, it shows up in one of 2 ways.
1. Usually the F or H field strength spikes, usually positive for north of the site, negative for south of the site. This is also moderated by magnetic lines of force as per the IGF mag charts.
2. If all 4 have a sharp dip simultaneously, then it is usually very near the geomag site.

I have also observed that the sharpest spikes (70-100 nT) within a 1-10 minute period are usually followed by some activity, usually in less than a 12 hour time period.

I am not certain of the exact mechanism which causes the local or regional magnetic field to behave in this manner, or as to why there is the 12 hour delay.
As a hypothesis, I could could postulate the following:
The crust and mantle is theorized to consist of Iron ( a magnetic conductor) and Silicon (an electrical semiconductor) (as well as other elements).
Motion of the fluid below the crust, as well as motion of the semisolid crustal area at depth could theoretically create a minor change in the regional or localized flow of magnetism (due to electrical and magnetic flows being altered by the fluid changes) which could be picked up with a geomagnetic instrument.
Since electricity and magnetism are mutually exclusive, all of the necessary elements are present (A magnetic and electrical semiconducting fluid).
Kind of short, but...