Damn...
I'd heard people bandy about "Nate Silver"... I just found out who he is.
Wow... that's really something, innit? 100% accurate EC and close popular predictions?
Nerd.
I don't think Nate Silver isn't particularly proud of getting every state right. He said about the 2008 elections that it's easy to know who will win on the day before election day, there is an abundance of polling at that point. I imagine he'd feel the same way about these elections Just going with a simple average of mainstream pollsters would give you every state but Florida and an even chance to get Florida depending on which pollsters you included. Nate's real work was much more impressive:
-He successfully predicted the variance of the results compared to his predictions. He had not just predictions for every states vote but accurate margins of error. And these margins of error seem to be pretty close. For instance I've seen it reported that if you look at his 80% confidence interval, 84% of the state results were in this interval. A good model would have those numbers close together, if too many states were in then it means the margin of error is too large, if too few then the margin of error is too small. This margin of error is useful because it not just tells us who's more likely to win but how much more likely to win they are.
-He was accurately depicting the election months in advance. Nate used race "fundamentals" in addition to the polls to predict the election way in advance. These fundamentals were calibrated off historical data and seemed to predict the future changes in the race very accurately. He was able to use data back when it was still scarce to give good estimates going forward. And these estimates seem accurate, there is no jarring change in his predictions when the more plentiful data came pouring in. This tells us that he not only had a good estimate of the margin of error at the end but had a good estimate all along, which is freaking amazing.
-His model also did a good job ignoring the short term swings of the race. After the democratic debate everyone said it was over but Nate Silver said Romney still had a shot. After Denver everyone said Romney was ahead but Silver said Obama was still the favorite. His model understood in both cases that a short term surge is usually followed by a candidates numbers falling back to earth. The million campaign events that the media fed to us were ignored by Nate's model, only getting the stuff that really mattered.
Being impressed that Nate Silver went 50/50 is like being impressed because Steven Hawking predicted a coin flip (and after having guessed wrong on the coin flip 4 years ago, which is when Indiana was the 50-50 tossup but Silver incorrectly had it leaning just barely towards McCain.).