Cathryn;

It makes sense all right but I doubt the predictors would accept it.

I would prefer to evaluate on a "hit or miss" basis for actual predictions. I feel Near miss or best fit evaluation should only be for training purposes.

Roger