EQF,

Check this link for calculating probability:

http://faculty.vassar.edu/lowry/binom_stats.html

Plugging in 17, 10, .25, .75, to be conservative, which would be the odds for 10/17 falling in the range +/- 45 degrees, which is 1/4 of the possible angles.

The result is likely to happen at random less than one time in a thousand.

To be scientific, you should generate some random times, plug them into your calculator to make sure they don't predict an equally unlikely distribution (that checks whether your code is working and other ways you might have made a mistake).

Also, you should repeat the experiment for events M7 and bigger, since a true correlation should appear not just for the very biggest events.

For the others posting, this result would not be of great use in predicting earthquakes, as any area would be on alert a quarter of the time, but it is a way higher correlation than previous studies of its kind in the literature.

I suspect there is an error in the analysis somewhere, but one can always hope some there's not.

John