Statistics are like a sewer pipe. The output is dependent on the input. Or as Nate Silver has recently discovered (H/T, TOM),

Nov. 2: For Romney to Win, State Polls Must Be Statistically Biased

Well, duh.

Polls are susceptible to all sorts of bias caused by poor design. Sometimes the error is an honest mistake. Liberty magazine predicted that FDR would lose in 1936 based on the results of a telephone poll. Their sample was skewed because a significant percentage of voters couldn’t afford a phone. Other polls aren’t really honest but are designed to show momentum for a candidate.

The polls this year simply don’t appear to reflect the real world sentiments one hears around the office coffee pot or in the checkout line at WalMart.

I suspect that Mr. Silver’s model is a bit buggy. We will see how well it does on Tuesday. On Wednesday, I expect that I’ll be able to say:

You smell that? That charred smell? That’s Nate Silver, son. I love the smell burnt out Nate Silver in the morning. It’s the smell … of victory.