The problem I saw with the training project was that it was only able to consider one factor for each skill (at least the study associated with the now-defunct USA offsite blog). That is, it could only say that each skill was affected mainly by one other skill. From BB posts, I'm pretty sure that all skills affect all other skills, but all to a different degree.
It still came to the conclusion that certain skills affected the rest more than others. The big takeaway was that handling was super important and passing and SB was less important (i.e.more skills are pulled up by high HA then by high PS/SB).