Foxy Nate Silver and why old-media hedgehogs could soon be old news

People don't "wait and see" because they want to know now what the likelihood of the future is. That's hardly unique or surprising. This is an important subject.

But I don't follow your evidence at all. The primary data output from 538 is a single number -- a chance of winning the election. How is that refuted by the inability to call every single state race? No one, certainly not Silver, claims to be able to do that. But numerate folks still understand that those probabilities can be combined to give a meaningful answer.

So let's flip it around: why don't you believe the (currently) 83% figure? If your answer is because some of those states will come out differently than their median expectation, you're completely missing the point.

The Daily Caller analysis misses an important point - there is a much larger volume of polling data for presidential elections than for congressional elections. Drawing conclusion from one about the other is largely meaningless.

The certainty of Romney's loss is bad for page views and viewer numbers for one. Pundits seem to have an interest in perpetuating uncertainty and turning the election into a narrative of a dead-heat horse race.

Duh. Base a model on polling data and you are only as good as your polling data. If the election is swung by cheating this time (as evidence suggests has happened plenty of times in the past), polls will be spectacularly wrong. Incorporate everything else you can think of and you get only a bit more accurate.

However the question is this. Do you have anything more accurate? I've had the experience on NPR of hearing pundits going on at length about what states matter. And it was less informative than a glance at http://www.electoral-vote.com/ is.

So no, it is not perfect. Nobody, particularly Nate, says that it is. But we don't have anything better.

Sports fans are constantly arguing about who they think is going to win such-and-such game. It's not important... but people enjoy it.

This isn't much different.

For both the games and the election, we will wait and see what actually happens. In the mean time, we enjoy discussing and forecasting.

Scarborough's the one saying it's a toss-up, so Silver's proposing a bet at Joe's odds. If Joe really believes his own odds, the bet would be fair.

I'd rather get 1/1 on a bet bookmakers are laying 1/4 on than make my confidence more visible :)

That's true of his model, but in his writing he's definitely taken a position.

If you believe that bookies are consistently giving the odds wrong, then you should start consistently taking their money.

Yes, I'm aware that this way of thinking is odd to people who do not understand markets, gambling, and so on. But it is still true. Bookies try really, really hard to give correct odds. And if you look at 20 bets which had 80% odds, you would expect to find several examples where the expected winner lost. That's why the odds were set at 80% and not 99%.

I think that goes to show more about the narrowness of the bet versus real life. Sports outcomes have scores and winners, case law has subtleties.

Technically, the individual mandate was illegal under the commerce clause according to the ruling. This was the main area of debate and clause that people assumed it would fall under.

The surprise was the expansion of the taxing power given in the ruling allowing the individual mandate to be considered a tax. It also was small enough not to be considered punitive, thus still legal under the "expanded" taxing power.

They were right, there were probably an 80% chance that they would declare Obamacare unconstitutional. The thing with 20% chance to happen happened, like it does once in a while.

If there's 80% chance of something happening, it doesn't mean that the 20% alternative won't happen or that you was wrong if it happened.

Citing statistical models based on massive samples of polling data is not biased reporting; it's based on fact, on what's actually happening. What the the media should not be saying is that the race is close when it isn't. It has no basis in reality.

One purpose of predicting the outcome is to take the narrative back from the pundits whose obvious motives are to maintain the perception of a close race. That's media bias. The FivetThirtyEight is a response to that.

I can give you another reason with a much bigger population base.

You have a candidate you are backing, with your time and/or money. But you are not made of time and money. What are you doing this weekend, working in the yard, or at the phone-bank? Signing checks to your candidate, or putting that money back into your business? Predictions help a lot here - they're useful for time and money allocations other than just at the "high-level campaign manager" level.