Silver’s probably doing what the big banks did with the mortgage-backed derivatives... selling them to their customers but shorting them in their own trading accounts.
Hold your right hand high in the air. Wave it and yell loudly “Look here! 80%!”
And under the table with your left hand you quietly bet on the 20% outcome...
You know, I’m so sick of hearing about Nate Silver. The only reason he carries any weight is that he predicted all but one state correctly in 2008. I heard nothing at all about him prior to that.
The real truth is that Nate Silver is nothing more than a manifestation of what statisticians call the lottery principal. Basically, that principal states that if you do something enough, events of low probability will occur simply by chance. In this case, Nate Silver was certainly NOT the only person trying to make predictions about the 2008 election. There were probably hundreds of bloggers out there trying to predict the state-by-state results.
So what are the odds of correctly predicting all but one state by chance alone? First of all, realize that there typically are only a handful of states where there’s an actual prediction to be made. Most results are pretty clear. For the sake of argument assume that there were 8 states where the results were uncertain and that these were true tossups. Given those assumptions, the probability of flipping a coin and predicting at least 7 of them right is about 0.4%. Pretty unlikely, but here’s where the lottery principal comes in. If you had 20 people trying to flip coins to predict the election, the probability of SOMEBODY doing as well as Silver is 35.6%. Obviously, the more people you had doing the predicting, the better the odds that SOMEONE would get it right or nearly so. The fact is that Silver was just this lucky someone last time around. If he does equally well this year, I would give him more credence, but predicting a single election well means nothing.
BTW, it’s much the same for mutual fund managers and stock brokers. There are lots of people trying to predict the stock market; by chance SOMEONE in a given period of time is going to look pretty good. That does not mean you should trust them to invest your money for you, especially if their advice doesn’t pass the “smell test” which is how I would describe Silver’s prediction this time around.