So did everyone elses. He was still underweight Clinton compared to the market and others attempting to do what he did.
His call on the 3rd party candidates making the election tougher to predict was obviously prescient.
If you woke up on Tuesday morning and said "Damn Nate Silver has Trump at 71% which corresponds to -250 and the betting markets have him at +500" then you printed money.
Diasagree.
Nate is basically a Bayes-theorem practitioner, which is all about theoretical probability and nothing about certainty. So Silver is basically incapable of evaluating the inputs to his calculations. When your whole purpose is to use "other factors" to interpret the polls and beat them in accuracy, you have to be able to read those said factors outside of the numbers themselves. A good example would be taking oversampled polls and discarding them.
It is worth noting the most accurate poll of 2012 (Investor’s Business Daily) was also the most accurate this cycle - an outlier in Silver's flawed model.
Garbage in, garbage out... if the polls are wrong ("rigged" as Trump would say) so is Silver - and he was.