• Quick note - the problem with Youtube videos not embedding on the forum appears to have been fixed, thanks to ZiprHead. If you do still see problems let me know.

How did Nate Silver get it so wrong?

It wasn't as "wrong" as many other polls that gave Hillary 99% or so chances.
 
Lots of undecideds. I think some polls take voting history into account too, and I would guess a fair share of the rural whites were first-timers.
 
Nate Silver got it more right than other forecasters. His model showed a fairly high degree of uncertainty because of an unusual number of undecideds/third party voters. Because of this, though Hillary had a bigger lead in the polls than Obama did in 2012, his model gave her a smaller chance at victory.

I said his model was crap but looks like I was wrong. The other models were the crap ones.
 
Last edited:
You guys are talking odds based on aggregates of polls. The more appropriate question is why the polls were off. Internal polls on both sides were also wrong.

The theory put forward on NPR earlier in the night was that the White voter (esp. non-college) margins were greater for Trump than expected.
 
This was the point I made on an earlier thread. But he didn't get it wrong. If he keeps getting every single state right all of the time while assigning probabilities otherwise, then he gets it wrong. He was giving Trump basically a 30% chance. And like people pointed out. Once one domino falls odds are more or going to fall.
 
Tony Stark said:
Nate Silver got it more right than other forecasters. His model showed a fairly high degree of uncertainty because of an unusual number of undecideds/third party voters. Because of this, though Hillary had a bigger lead in the polls than Obama did in 2012, his model gave her a smaller chance at victory.

I said his model was crap but looks like I was wrong. The other models were the crap ones.
Click to expand...

I was going to say something along these lines. 538 came out of this with a modicum of credit here and refused to be browbeaten by the baying of the Huffington Post people and the Princeton Election Consortium.
 
Caper said:
This was the point I made on an earlier thread. But he didn't get it wrong. If he keeps getting every single state right all of the time while assigning probabilities otherwise, then he gets it wrong. He was giving Trump basically a 30% chance. And like people pointed out. Once one domino falls odds are more or going to fall.
Click to expand...

Yes, in the Signal and the Noise he explains that a model probably won't get everything completely right. Some models are designed in such a way that they overfit the data which means they can only really work under identical conditions. If anything, I think the 538 people will say they can get a lot of good data from this which will help them improve the models in the future.
 
angrysoba said:
If anything, I think the 538 people will say they can get a lot of good data from this which will help them improve the models in the future past.
Click to expand...

FTFY

It's always possible to tweak a model to fit all historical data. The trick is knowing whether or not such a model will also predict future results. Pollsters have a pretty poor record lately.
 
Yeah, I have seen a lot of people really angry and saying that they'll never trust FiveThirtyEight again. This baffles me. They've consistently said that there was a lot of uncertainty and that while several things had to go right for Trump it was absolutely possible.

The ones that called it for Trump were (at least for the ones I am familiar with) not credible sources. So it's not like I'm going to say "Oh man, I should have listened to the Breitbart online poll!".
 
Steve Schmidt (Repub) said last night on MSNBC that this was an organic groundswell, across several states. Not sure how that means it gets missed by polling, but he had predicted Hillary at 320 electoral votes.
 
Fivethirtyeight nailed it. Have you read the Wang/Cohn and silver back and forth? People have slit their bellies for less than how wrong they were.
 
Anybody who's played a fair amount of poker knows that a 75% chance of winning is nothing to take to the bank.

Essentially, Hillary lost all the toss-ups.
 
Silver also gave Hillary a fairly significant (10.5%) chance of winning the popular vote but losing the electoral college.

His model is solid.
 
Tony Stark said:
Silver also gave Hillary a fairly significant (10.5%) chance of winning the popular vote but losing the electoral college.

His model is solid.
Click to expand...


The huffpost guy apologized. Frankly, they probably need a "diversity is white genocide" level of acknowledgement.
 
Brainster said:
Anybody who's played a fair amount of poker knows that a 75% chance of winning is nothing to take to the bank.

Essentially, Hillary lost all the toss-ups.
Click to expand...

Yeah, Hilary basically did not make any of her Saving Rolls.....
 
dudalb said:
I am all in favor of making bad pollsters eat a little crow,but not with racist statements, thank you.
Click to expand...

It is a reference to last night. Tony Stark lost am avatar bet with skeptic tank. The first proposed avatar was that phrase.
 
You must log in or register to reply here.

Back
Top Bottom