That's the way I did it when running the program for the USGS in Golden. Each prediction got a point score depending on how much information was supplied. That was adjusted for probability based on the seismicity of the area. Then the results were compared to a series of random predictions.

It requires a series of predictions to establish the overall credibility. It's difficult to do it this way here, so all we can do is determine the probability of the predictions.

Roger