Of COURSE I'm doing that!

The program looks at each quake, finds the window it's in and counts that window as a hit.

Once all quakes are "windowed" the windows with hits are counted. I do not count how many are in each window; it doesn't matter.

As to Martin't 3 hits,

5377 has a probability of 31.7%. I used a 15 day window, 1973 - 2000. There were 682 windows, 216 had quakes in them (436 quakes total). A hit of moderate significance.

5240 has a probability of 1.4%. I used a 9 day window, same timespan. There were 1137 windows, 17 quakes, 16 hit windows. 16/1137 = 1.4% Very significant hit.

5017 has a probability of 97.1%. I used a 12 day window, same timespan. There were 853 windows, 7888 quakes, 828 hit windows. 828/853 = 97.1%
A hit of no value at all (nearly).

But this is not the whole story. To establish credibility, ALL his predictions must be evaluated to get his overall success rate. You can't just point at the hits and claim some ability, it's the total picture that counts. And he's missed far more than he has hit.

Roger