Brian;

It's the Jones Observed Probability calculation.

You take a time interval, usually many years, divide the days into consecutive windows of the prediction's specification and sort the quakes into the proper cells.

Then you count the number of cells and the number containing one or more quakes.

The odds are then hit cells/total cells.

That's the observed probability that a cell chosen at random will contain a quake.

Easy to compute and mathematically sound.

Choosing the right interval is problematical. A long time gives the best average but that may not fit a period of increased or decreased seismicity.

Roger