Agreed but I wonder how many predictors will understand that?

Past predictions should not be entered, no way to prove accuracy.

Evaluation should be separate but using this as input. Possible?

Roger