Predicting election results in the 50 states is actually much more easy than most people think. West Coast and East Coast are democrat, Midwest, Texas etc. are mostly republican (the Midwest becoming more republican because the population is aging due to brain drain by young, smart people - mostly democrats). So indeed the task is not about correctly predicting results for the 50 states, but simply predicting results for the 5 states where prediction is not straightforward.
If your predictions for these 5 states were totally random, your chance to be correct for all these five states would be 3% (1 in 32, to be exact). Now if 200 people make random predictions for these five states, a few of them are bound to get it right. Just interview one of these lucky winners, and here you go: you've got what seems, at first glance, to be a great story.
Now I'm sure Nate did more than just making random predictions for these 5 states, which increases his chance of being correct on all these 5 states (as well as the remaining 45 states where even a kid could make correct predictions) from 3% to maybe 40%. In short, nothing spectacular - much much easier than winning $1,000 when playing lottery.
I'm sure some plenty of other folks made the right predictions, including some who made purely random guesses.
Anyway, here's the story:
The New York Times election statistician Nate Silver perfectly predicted all 50 states last night for President Obama, while every single major pundit was wrong – some comically wrong. Despite being derided by TV talking heads as a liberal hack, Silver definitively proved that geeks with mathematical models were superior to the gut feelings and pseudo-statistics of so-called political experts. The big question is, will the overwhelming success of statistical models make pundit forecasting obsolete, or will producers stubbornly keep them on the air?
Silver’s analysis, and statistical models generally, factor in more data points than even the most knowledgeable political insider could possibly juggle in their working memory. His model incorporates the size, quality, and recency of all polls, and weights them based on the polling firm’s past predictive success (among other more advanced statistical procedures).
Read full story at http://techcrunch.com/2012/11/07/pundit-forecasts-all-wrong-silver-...