Chris;

> Not sure what I'm missing, but I agree with Amit, again. I agreed with him a few weeks ago when he said that multiple hits on one day should be counted (I think they should be subtracted from the total expected quakes).

I'm scoring predictions, not quakes. If he predicts the number of quakes, I'll allow it but the odds will change too.

> I think Roger's statistics are assuming that M6 quakes are not clustered, and if they do cluster, then the evaluation is not correct.

No, the method is designed to eliminate clustering as much as possible. It's easier than declustering the catalog.

> Roger never answered my question about Amits example of a few weeks ago that he could have 100% of his predictions correct, but if some came on the same dates, Rogers evaluation at that time would give him a lower score.

It isn't true. The score is determined by the odds on a hit, not the number of quakes involved.

> And I totally don't get how beating 92% of random predictions is below chance. Is Roger expecting 97% significance? Isn't one standard deviation 67% or something like that?

No, 99% I set the bar high because I don't want to credit a lucky guess. Remember, even at 99% there's a 1% chance it's just luck.

> Unlike some here, I don't think a prediction has to be useful for evacuation or whatever to be useful for understanding earthquakes.

Agreed but do you want to give prediction credit to someone who's only right half the time?

> But, scientists are never going to Amit seriously because the methods he uses make no physical sense at all.

Exactly!! Would you expect success with a phony method?

Roger