Roger:

About a 1 in 20 isn't too shabby (as a first approximation).

A more accurate way (probably) is to use the list of potential epicenters with the 65 km radius to generate a list of all quakes > M5.2 (or M4.9) (with same window). Then you would purge the list of duplicate events (resulting from the area overlap...if any). This would give a better historical probability number.

Of course, this number is all that (some) can use as of the moment to approx. the liklihood of an event. At least until some other theory emerges which might change understanding in this area. It may be that No connection exists between historical probability and (future) probability.

Just $.02 worth