Oh, I fully agree with you, and, as I mentioned and as others mentioned, both flavors of this basic method work so well it’s difficult to distinguish their effectiveness in an objective sense.
I will say this, though: Wang called the Republican Primary for Trump back when Silver was saying that Trump was likely to lose, but since Silver himself has owned to that mistake, I don’t really want to harp on it here. (And it isn’t even a mistake in Silver’s methodology, since his numbers were essentially similar to Wang’s; his mistake was in not taking the numbers seriously and trying to be a pundit, as he himself said.)
But, assuming they both end up fairly accurate, you’ve got one person making simple assumptions that produce a fairly consistent prediction early on. You’ve got another that has lots of special sauce adjustments that gives rather large swings in it’s predictions through the election season. Simple, consistent and accurate trumps complicated, erratic and accurate. imho.
But there is only one data point at the end. Without checking the agreement between mid august guesses and final outcomes over a large number of runs, I can’t know whether a volatile model is correct. I could have posted on June 30 2015 that Clinton had a 80% chance of winning, and never budged. If she wins does that mean that I’m a better prognosticator than Wang, since my prediction was very simple and consistent?
Also take a look atthis. Now an election isn’t a football game, but it is certainly possible that the probability of winning could vary substantially over time. If Trump had aced the debate, and his leaked tax forms had been replaced by leaked emails in which Clinton pushes her subordinates to down play the terrorist angle in Benghazi, I don’t think that her probability of victory would remain unchanged any more than a teams chance of winning will stay steady after scoring a fourth quarter touchdown.
I was also a Nevadan for about 15 years. It’s definitely vaa like “vat”, not *vah *like Mrs. Howell’s priceless Ming vaaahse.
Here’s hoping he visits Missouri soon. As a former Missourian I can say that whether you pronounce it Mizz-urr-ee or Mizz-urr-uh you’re going to piss off about half the residents. It’s a must-lose situation. Don’s a natural fit for that. “I’ll take Kobayashi Maru gambits for $200, Alex.”
Well, yes frankly. If you consistently managed to be right. Remember I started the post with “assuming they both end up fairly accurate”. To jump to a different sports analogy, imagine baseball batters waiting on a fastball. Who’s the better hitter, the one who thinks it’s going top left when it leaves the pitcher’s hand or the one who figures it out 2 feet from the plate?
The odds of one side’s success do in fact fluctuate day to day. It’s not like a pitch where the point of eventual plate-crossing is fixed by physics when it leaves the pitcher’s hand.
To be sure, answering the question “Who would win if the election were today?” is different from “Who will win in Nov based on what we’ve seen so far plus what we expect to see between now and Nov?”
The first question should have a bunch of volatility in a hard-fought punch-counterpunch race. Especially one played for ratings by the media. The second one should also have volatility. But how much? If everyone was a hard-over partisan, everybody always voted, and there were no wildcards like 3rd parties, Big news events or candidate gaffes, and no election day snowstorms or voter suppression, intimidation, or election board incompetence, this’d be easy.
But there *are *independents, undecideds, and maybe-I’ll-vote-maybe-I-wont’s. And all of those groups do in fact fluctuate over time. And with different dynamics in different states. A projection that does not capture those fluctuations is wrong in principle. Even if, like a blind pig, it finds an acorn now and again.
Unless the race is poised on a razor-edge, an ideal prediction should be less volatile than the daily opinion polls. But how much less?
@LSLGuy:
Ok it’s a bogus analogy. So instead imagine one election predictor who pretty accurately predicts early and another who eventually settles on an accurate prediction late. “But a whole bunch of stuff happened in the middle so you can’t blame the guy for only being accurate late”. Sure, but one guy didn’t agree all that middle stuff made much difference and another did. Who’s better at predicting?
Because I have close family who are gonna be in the middle of this storm, I keep checking the radar and satellite imagery, and it looks to me like this storm has taken a tack to a direct westerly heading since it got to Freeport.
So that got me to wondering: what if the hurricane hits Palm Beach and obliterates Mr. Trump’s Mar-A-Lago? How might that affect the [del]campaigns[/del]election?
Would the voting public take it as evidence that God did not care for Mr. Trump?
Or would people think it was Obama using secret government technology to control the weather to decimate his political ally’s foe while making it seem like God does not like Mr. Trump?
