I'm convinced the definition of "close but not a hit" will never be agreed upon.

I think the best we can do is each make arguements for/against the performance of a predictor, and base that on whatever criteria and methods each of us believes in. Of course a solid consistent database of predictions is a must.

Michael