Nate Silver is not "wrong" in how his models reach their conclusions, they're just built on assumptions that could be wrong. Just like this guy's conclusions.
What is consistent in almost all these polls:
Obama has a slim lead where party ID leans Dem;
Romney has a larger lead (sometimes drifting into the "huge" category) among Ind;
Obama scores slightly to dramatically higher than Romney on relatability issues;
Romney scores slightly to dramatically higher than Obama on the economic issues.
If Romney maintains his high-single to double digit advantage with Inds, and GOP turnout equals Dem turnout, Obama is toast. If Dem turnout exceeds GOP turnout by enough to "eat up" the advantage Romney has with Inds then Romney, also obviously, is toast.
The underlying dynamics (fewer early votes for Obama than four years ago, greater enthusiasm for GOP, etc.) would seem to favor a turnout model that is closer to parity (as Gallup has found). But even with the above facts in the data it remains a complete and total question mark whether Obama can blunt these perceived advantages that Romney carries into Tuesday and equal what the polls are presuming will be the case.
Everything until that answer is mostly noise, unless there is some new event that casts one of the candidates in a terrible light.