The thing is, there is no ‘complete puzzle’. A player isn't the average of X number of statistical models. In particular, all the models have trouble quantifying individual defensive ability, and it doesn't get any better if you average them together: they are not just missing the target in different directions, they don't even agree what the target is or how much it should be weighted in ranking a player.
In circumstances like that, you're better off just ignoring the least accurate model instead of trying to juggle it with all the others.
__________________
WARNING: The preceding message may not have been processed in a sarcasm-free facility.
|