Pre-computer era elite GMs and computer-era elite GMs have similar error rates[1], with the caveat that we presume disagreeing with the computer's top choice to be an error. They mostly agreed with the computer's assessment something like 50-57% of the time.
As far as I know Magnus' games haven't been subjected to this kind of analysis, and I recall 2 pretty bad moves in the recent match with Anand, so I'd be interested to see it. I suspect he may be even slightly above 60%.
Should it be something like high 70%'s though I think it would be time to start a conversation about it. It's how Ivanov was busted: overwhelmingly playing one of the top 2 engine moves, which is basically impossible. [2]
1 - http://en.chessbase.com/post/computers-choose-who-was-the-st...
I understand that Belichick has a doghouse for backs that fumble. The point seems to be that his doghouse is likely to be rarely occupied since 2007, and whether that is due to coaching, drafting backs with good hands, or something less wholesome.
I appreciate the points on Maroney and Ridley.
"Your explanation isn't supported by the data": I wasn't trying to give an explanation supported by the data. I was trying to give an explanation for the data that makes sense and is supported by observing the Patriots play.
It's like saying Magnus Carlsen must cheat at chess, because the chance of him making so few errors is effectively zero assuming he makes errors at the rate that other grandmasters do.