Nate Silver's GOP bias
Not only did Nate Silver correctly predict the outcome in all the states but his vote share projection was incredibly accurate. For the swing states, where his model really mattered and where he had the most data, he was off by an average of only .3 per projection. (standard deviation of the difference between projected and actual was .9) But his model did not, contrary to some claims, have a Democratic bias. He underprojected 8 of 11 swing states for Dems - underprojecting .2 on average. He overprojected on 8 of 11 for the GOP, overprojecting by .8 on average. See the spreadsheet at: link.