I've always thought that the 100 point scale for attributes was stupid. It's almost entirely arbitrary to assign a value that is as specific as, say, Jason Witten's 88 "Toughness" rating. It makes even less sense when the rating is compared to other players, like Romo or Columbo, who have a 97 and 80 "Toughness" rating, respectively. Is Romo really 10% tougher than Witten? How in the world do the developers think that they can accurately quantify that?
The way that the rating system should work is the same way that fans base their evaluations. If we had a discussion on the forum trying to rank the NFL QBs, there would be too many opinions to come up with one set list of "This person is #1, this person is #2". Instead, we would come up with a grouping, or tier, system that allowed us to put comparable QBs together.
The best way to do this would be to compile all the data of actual compeltion percentages, make minor adjustments based on data from a place like F.O. that tracks dropped passes, and then do a linear regression analysis on the data. We could then "grade" all the data on a bell curve based on average, giving a grouping that placed only a few players at the top of the spectrum.
All players with "A" rated accuracy would have comparable skill and would show a marked improvement over a "B" rated player, who would have more skill than a "C", etc. I commend EA for trying to lower the ratings of their players, as most guys in the NFL are just average. However, their actual rating system still makes very little sense.