[kqr]

You're missing one important thing: confidence level. You can probably make a 7+ magnitude prediction every day as long as your confidence in it is -- I guess, not knowing much about the Pacific Rim -- less than 1 %.

It's the 10 %, 20 %, 50 % confidence level predictions you have to be a bit more sparing with!

[cjensen]

I completely agree that a more advanced model would use confidence levels too.

And I forgot to mention this isn't my idea at all. This framing of what is a prediction was invented at USGS. Mea Culpa.