The infotainment business is unlikely to jettison the gut-feel/insider sources/wishful thinking approach to pre-election analysis -- and there can be value in some of it when it's done well -- but maybe they'll add few stat geeks to the mix for accuracy. 

Not really right for the last link to label the poll-crunchers as "pundits," which probably speaks to the resiliency of the old mindset. And, yes, duh, poll-crunchers rely on polls, by definition, which is one of the things (another being the use of the phrase "magic formula for predictions") that makes this the worst analysis of Silver and his ilk I've seen today.

Comments

[Andrew Brod]

I saw the Engber piece in Slate earlier today and was struck by how clueless he was. It reminded me of the equally clueless tweet by a Politico writer: "Nate Silver admits he's simply averaging public polls and there is no secret sauce." Not only was it hardly a secret that Silver was averaging polls, but there actually is some secret sauce. Silver doesn't reveal how he does the averaging and weighting of the polling data.

On the other hand, I don't think it's quite right to say that Silver predicted 332 electoral votes, per Slate's pundit scorecard. His actual final forecast was 313 electoral votes.

Yes, Silver predicted that Obama would win enough states to generate 332 electoral votes, but not all states were equally likely wins for the president. It looks like Silver weighted his state forecasts by the probability of each state being won by Obama, in the same way that the expected value of a coin flip is 1/2 rather than 1 (heads) or 0 (tails). Florida was so close in the 538 model that the forecast hedged its bet. And the delayed results in Florida bear that out -- it's still very close!

Even so, it's more accurate to say that Silver predicted 313 electoral votes. If Obama wins Florida, Silver will be a little low; if Romney wins Florida, Silver will be a little high.