I do have a way to do this that works for me. It requires the convergence of 6 different factors, all objectively measurable, which I think is a reasonably small number, so I don't think I've just been curve-fitting, in the bad sense, all this time. As of now, I've got data on about 700...