Reading Cal's book, Jim discusses many, many possible ways to predict earthquakes.

I don't think it is a stretch to guess he simply took the one with the best track record in the past, without testing whether the apparent success might have appeared at random, and has applied it since without success.