the egghead election
Above are the results (some of which could still change a bit, but probably not much) plotted against the corresponding RCP averages from the fifteen most contested (and most polled) states. The dotted line is the line of identity (x = y). Data points above the dotted line are where Obama or Romney outperformed their poll numbers, and below where they underperformed.
It should be said that the general tendency is going to be towards underestimating either candidate's share of the votes, because the polls include undecideds and vote tallies, obviously, do not. Still, in only one case was Mr. Obama's share of the vote overestimated, and that was in AZ, where perhaps not coincidentally only two polls figured into the RCP average (there were 7 for OH, by contrast).
There's a fair amount that could be said about this, but for the moment I think I'll just let the numbers speak for themselves.