Nate Silver's forecast wasn't that bad

Post Reply New Topic

Looking at Nate Silver's forecast, in the final forecast he gave Trump a 29% chance of winning. That is, 29 times out of 100, Trump wins. The chances he gave Trump were much higher than many others:

http://fivethirtyeight.com/features/why ... yone-else/

This is like with the weather forecast, if there's a 30% chance of rain, then 3 times out 10 it's gonna rain, but a lot of people like to call meteorologists stupid or incompetent during the times it does rain as though a probability less than 50% is supposed to never happen. You judge the 30% predictions by looking at all of them. If, analyzing them, you see it rained 3 times out of 10, then the model is doing what it's supposed to.

He was wrong for 1.5 years, every step of the way.

Possibly, the most wrong person out there.

EDIT: on election night it was amusing to see many of his subscribers on his site complaining that he sucks

And yet Scott Adams was consistently predicting a Trump win, maybe a year ago at least, with explanations and examples all along the way.

http://blog.dilbert.com/

This makes zero sense. The people who gave the lower probabilities were wronger than him.

A 29% chance is actually a pretty decent chance, and if you held stock by his forecast the Trump win would not have been that surprising.

Personally, I actually wasn't really all that surprised. As soon as I saw that Hillary's numbers were lower than expected, I realized Trump might just win this thing.

Another pollster/researcher who picked Trump very early on was Anne Sorock of the Frontier Lab:

Research Guru saw Trump phenomenon coming before anyone else

For most people, both Donald Trump’s campaign and success came out of nowhere.

But not for Anne Sorock of The Frontier Lab....

At the Frontier Lab, Anne has applied deep values methodology to numerous political topics, including why people decide to become politically active, Occupy movement participants’ motivations, and why Republicans won’t call themselves Republican, among others.

I saw Anne at CPAC 2015, and in the course of our discussions, I asked Anne who she liked among the many rumored presidential candidates. She said Donald Trump.

I was like, what is that all about? He’s never going to run, he always teases, and anyway, Trump? Seriously? She was serious. She said, look, he’s the one. She was insistent not only that Trump would run but that he’d win. It seemed totally incredible.

http://legalinsurrection.com/2016/11/re ... yone-else/

Jacoby posted a graphic of the numerous times Silver said Trump was finished.

Well, here's one example, from 10/20:



http://fivethirtyeight.com/features/cli ... ast-night/

It's pretty clear Nate Silver at the time considered a Trump win to be well within the realm of possibility.

It was pretty bad, Silver is simply a pundit that aggregates polls and makes his interpretation of them. The polls were almost entirely wrong, take a look at the final election polls and tell me they got it right. Silver's interpretations and understanding of the electorate was wrong, he's been consistently wrong from a long time now. Giving him any credit is like a weatherman predicting a 20% chance of rain and getting credit either way. The guy had no clue, I personally had a much better understanding than him and was able to decipher thru the BS polls

http://www.realclearpolitics.com/epolls ... -5952.html

Statistical methods like the models weather forecasters use are scientifically valid and can be verified by looking at the performance of an aggregate of forecasts over time.

I have not, however, seen any scientific research on 538.

By all means take Silver's writings with a pinch of salt, although I'm not sure where you expect to find better ones. Even the articles posted by Jacoby showing that Silver was "wrong" contained a large sample in which he made accurate predictions.

His models have a very good track record. They've been right for three elections in a row now.

Understanding the models requires an understanding of prediction generally which frankly most people do not possess. Unless you are willing to put the effort in to understand them then your criticisms will usually be groundless and uninformed; this goes equally for the people on the left who feel disappointed and for the people on the right who feel smug. If you feel there is some variable that Silver is missing, then design your own model and test it against his long-term (or latch onto an existing alternative model). That's the only way you'll be able to show that you know better.

The really funny part was watching the rest of the liberal blogosphere rip into Silver in the week leading up to the election for giving Trump too much of a chance; he may have been wrong, but he wasn't dead wrong. Also pretty illustrative of that liberal tendency to eat their own when they break with the party line.

How have they been right 3 elections in a row? Pure nonsense. His predictions were wrong and 'probability' is a way of saving face. Silver got famous for CORRECTLY picking the winners in pre-2014 elections, he has rarely been right since. There is nothing he can add, he isn't even a pollster, he rides on their coattails making his own little interpretation. You should listen to me now as I have been more correct than Nate Silver for years now.

I don't care about Nate Silver bogus 'data driven' approach or his punditry, I find it hilarious that people are trying to defend their polling guru instead of coming to obvious realization that there is NOTHING scientific about political polling and it's agenda and narrative driven as the person administrering the poll can make it look like whatever they want... Like Hillary Clinton winning the presidency! Don't believe me, you'll just continue to be wrong in the future however.

Lets see, what next will they get wrong? Europe has some interesting elections coming up...

They were right in 2008, 2012 (when WrongPlanet's conservatives derided him for saying Obama had a 95%+ chance of winning when the polls were 49-48), and 2016.

Probability isn't "a way of saving face", it's the only sensible way of doing things because people can't see the future!

Silver deliberately doesn't say things like "Clinton will be the next President", because he's acutely aware of the degree of uncertainty. He can't predict which way undecided voters will fall or whether it will rain in Michigan or whether the FBI will announce an ungrounded last-minute investigation into a presidential candidate, for example. It's entirely sensible to present a range of possibilities. It's entirely sensible to run millions of simulations with variables slightly changed and use them to determine probabilistic outcomes.

You just say that the things you want to happen, will happen. On this occasion, they did happen. It doesn't make you a master of prediction any more than Kraichgauer or AuntBlabby were master predictors in 2012, it makes you a fortunate ideologue. Your own lack of understanding of the fundamentals of prediction is not a feather in your cap. Frankly you're undoing any credit you may have earned for "predicting" the way that Pennsylvania flipped by showing your arrogant post-factual anti-intellectual understanding of the way the world works.

Nope, you're wrong. Nate didn't predict the winner as he did in the past, he has been consistently wrong in his assertions for a long time now. In fact, part of the reason you like Silver and other did as well was because he was 'progressive' partisan posing as a numbers guy and he was able to call Obama beating McCain and Romney, if I made a prediction in this years it would of been the same because it was clear McCain and Romney were goin to lose from the beginning.

I actually understand polls unlike most, they're not magical or unquestionable and in fact more are just BS. I also understand America and the part of the country that decided this election, you could never ever know so you can only depend on pundits who are wrong almost all the time so you should listen to me on these specific things because I have a better idea than most.

I don't have to respect Nate Silver and I think this saving face is pretty funny, everybody is laughing at him. Maybe he should stick with baseball or get a real job.

No, he predicted a close race that had a 1 in 3 chance of going to Trump if lots of things went his way. You know, that thing that happened

If he'd predicted a 97% chance of a Clinton win then yeah laugh in his face (although of course even 97% is not 100%).

Imagine you have a three sided spinner, with two sides saying A and one side saying B. It is to be spun once.

Jacoby: It will definitely land on B! Unless it's rigged!
Nate: There is a 1 in 3 chance of it landing on B, and a 2 in 3 chance of it landing on A.

The spinner is spun and lands on B.

Jacoby: HAHA! I was right!
Nate: I knew this was possible.
Jacoby: You know nothing, you were hedging! B! B! B!

Silver isn't wrong if the outcome he says is most likely doesn't happen; he's no more "wrong" about Trump winning than he would have been if Clinton had won exactly the states he forecast, or even if she won 400 electoral votes. He's wrong if something the model declared impossible actually happened (DC voting Republican, Clinton winning Idaho), or if the model assigned the probabilities incorrectly.