We are at opposite ends of the spectrum on this. But, I do offer you this possible solution. I liked parts of your idea and I liked parts of Roger Hunters idea. The neat thing about Roger's concept is that it seamed to rate predictors against each other, sort of like the old scoring curves in chool. The neat thing about yours is that it allows for somethnig other a yes/no point on a paramter. I'd suggest this:

A prediction is defined by:

1. A single point in time.
2. A single point in space.
3. A single magnitude.

A prediction is scored by how close the EQ occurs to these defined points. Each of these parameters has a maximum 100 points (sound familiar).

So If I predict a 3.0 under your house (a little humor there) and an EQ occurs right under your feet, at the exact time I predicted, at the exact magnitude, I would get 300 points (sounds like bowling!).

The point systems would trail off from 100 points to 0 for each parameter over some agreed time, distance, and magnitude from the predicted EQ. The point that these parameters would hit zeros should probably be based on the predicted magnitude, and a formula should be defined to set these zero points for each paramter for any given predicted EQ magnitude. This would solve my Disneyland complaint I stated in an earlier post.

So, for example:

A 3.0 would go to zero after say 10 hours, 10 Km, and .5 in mag.
A 4.0 would go to zero after say 50 hours, 100km, and 1.0 mag.
A 5.0 would go to zero after say 250 hours, 1000km, and 1.5 mag.

Something along those lines. The zero point for each paramter should be linear for any predicted magnitude, and that is something a formula would provide, based of course on a predicted magnitude.

Whadda ya think?