That is why I have always criticized multi regression stats... and I minored in them in college.... basically they try and find the most common things that move in a direction and try and predict the likelihood that something will move in a certain direction. Problem is stats dont predict causality only that there is a likely relationship, Eliminating other less likely relationships. Also they are less predictive when not enough data is available and in this case Josh is/was a case study in both less data and elimination of other potential factors including the ability to improve... something exceedingly hard to quantify. Which is why eye test and interviews on personality were so important to the Bills decision to draft him so high.