Jump to content

Nate Silver is full of shit


Victor Lazlo

Recommended Posts

2016 he had Hillary 67% on Election Day. Nice call. 

This morning he had Biden 89 out of 100 - or 89% chance the way any rational lay person would read it. 

It doesn’t appear that the race was anywhere near what he said it was.  It should have been a walk in the park for Biden, according to him.

Oh, yeah, blah blah, his methods, his 40,000 simulations. Bullshit.

He doesn’t know dick. 

 

  • Hook 'Em 4
  • Haha 2
Link to comment
Share on other sites

Just now, longhornmatt said:

I eventually bought into this after 2016, but it’s much worse now.   If you act like American politics is a random grab bag, then sure, they’re doing fine.   But we only have two political parties and very narrow potential deviations that we’re dealing with in every election.  For the models to be off by so much in that environment really strains credulity.  

What I'm saying is that we have no evidence that the model is far off at the moment.

If you want to be pissed about polls in Iowa and Texas, be pissed off.  But 538 had dozens of scenarios where Biden lost those states and still won the election.

It's important to read what one criticizes.

Link to comment
Share on other sites

Just now, longhornmatt said:

Consistently getting the probabilities wrong can happen even if you mail the ultimate winner or loser.  Vegas doesn’t get credit for setting a -21 line on a game of the favorite ends up winning by 3.

I'm again going to disagree here.  I don't think 538 was claiming Biden would win with 330 electoral votes.  They were saying there were far more scenarios where Biden wins than he loses.

That is because Biden needed a few swing states and Trump needed almost all of them.

Link to comment
Share on other sites

7 minutes ago, fellside said:

If a guy says someone has an 89% chance of winning and you interpret that to mean it's going to be a blowout, it says more about your understanding of probability than it does the guy himself.

Go look at the 538 forecast and you will see numerous scenarios as close as the one we're seeing unfold.

Also, there's a very decent chance Biden wins several of these states as mail in votes get counted.

House, Senate, and EC are all coming up on the very edge of any reasonable confidence interval, and all in the same direction.  Just coincidental noise, I’m sure.  Quite the probability defying event.  

  • Like 1
Link to comment
Share on other sites

33 minutes ago, Voldemort86 said:

The polling was wrong for the second presidential election in a row. The state polling is where it has been extremely disappointing.
 

Do you think Trumps Nuremberg rallies are a cheat code who hacks the system? I guess the shy trump voter isn’t a myth.

It's nowhere near as disappointing as the Constitutional mechanism by which we select the president.

  • Hook 'Em 2
  • Like 2
Link to comment
Share on other sites

42 minutes ago, BehoId, The Underminer! said:

House, Senate, and EC are all coming up on the very edge of any reasonable confidence interval, and all in the same direction.  Just coincidental noise, I’m sure.  Quite the probability defying event.  

You're speaking very confidently considering that of the remaining ballots we can reasonably expect them to break 60-40 Biden. In Michigan and Pennsylvania we can likely expect it to break 66-33. In Georgia, we're basically just waiting on Atlanta. Expect the score to run up a little more in Biden's favor in the meantime. Wisconsin has already flipped.

Link to comment
Share on other sites

They had Michigan and Wisconsin going to Biden by double digits.  If he manages to win them, it will be by 1 or 2%.  That's well beyond the moe.  The willingness of people to vote for Donald Trump cannot be accurately polled.

Edited by kevwun
  • Hook 'Em 2
Link to comment
Share on other sites

27 minutes ago, kevwun said:

They had Michigan and Wisconsin going to Biden by double digits.  If he manages to win them, it will be by 1 or 2%.  That's well beyond the moe.  The willingness of people to vote for Donald Trump cannot be accurately polled.

If you're going to talk shit at least get your numbers right. They had it at +8 not double digits. They were wrong and they should not have been as wrong as they were but they weren't as wrong as some people in this thread are trying to argue they were.

Link to comment
Share on other sites

Yeah, I read week's worth of "The Silent Trump Voter is Bullshit."  It's way beyond more real than can be possibly imagined.  We're watching 7-12% leads in key states just get absolutely eviscerated.  Biden will be lucky, and I mean LUCKY, to win a few of these states by 1-3%.  The entire polling and political consulting industry needs to be burned to the ground.  How can these fucking idiots, in a day and age of easily accessible digital information...be so fucking wrong?  Landline half-assed polls in the 1980's were literally 2x as accurate as this crap.  

Edited by Lobo
Link to comment
Share on other sites

They missed most of the swing states by more than they did in 2016. All in one direction. Most by almost the MOE if not more.

Their methodologies might be fine but it's probably time to accept that polling humans in 2020 just isn't really possible to do accurately.

  • Hook 'Em 2
Link to comment
Share on other sites

Yeah, they know when I need to order more paper towels almost to the hour.  But based on my political behavior for 25 years given my voter roll and donation history (or lack of both), it's just a total fucking mystery how I might come down on the 2020 POTUS race.  Known only to God, I suppose.  

What a fucking con that industry is.  

Link to comment
Share on other sites

1 hour ago, kevwun said:

They had Michigan and Wisconsin going to Biden by double digits.  If he manages to win them, it will be by 1 or 2%.  That's well beyond the moe.  The willingness of people to vote for Donald Trump cannot be accurately polled.

You’re talking about polls, which 538 does not do. They just give the odds based on polls. 

Link to comment
Share on other sites

6 minutes ago, BradInATX said:

They missed most of the swing states by more than they did in 2016. All in one direction. Most by almost the MOE if not more.

Their methodologies might be fine but it's probably time to accept that polling humans in 2020 just isn't really possible to do accurately.

Name a state that fell within the MOE.

Link to comment
Share on other sites

Nate Silver makes predictions on polling data.  He has no control over that data.
If Biden pulls this out, then yes, it will be consistent with Silver's 89% estimation.  It certainly doesn't imply that polling was particularly accurate.
That's an overly simplistic outlook. Okay, they correctly predicted something that they gave a 90% chance. But they missed badly in almost every state that matters, just like in 2016. Outside of margin of error.

If any real data science pro turned in a project with 10 projections and missed outside of MOE on all 10, they wouldn't have a job anymore. That's what 538 just did. They missed the 8-10 most important data points outside of MOE. That's atrocious. I understand it's the polling that's the issue. But we may need to come to grips with elections not really being attainable to model and predict with any real accuracy.
Link to comment
Share on other sites

Here's the story:

#1 America is filled with people who don't care about America. They care about themselves, and their feelings, not their reality. 

#2 Democrats obviously don't understand microtargeting, the whole "Latinos are not all the same" isnt just Latinos, it's blacks, whites, asian, and even within that rural suburban and urban. Further than that it's income, donation history etc.

With modern computing it's not difficult to do the targeting you need to do to get the turnout that you need to get... Trump did it. 2 elections in a row. 

#3 Texans have alienated hispanics and blacks in very bad ways. And that starts locally not nationally. 

 

  • Hook 'Em 3
Link to comment
Share on other sites

17 minutes ago, immamac said:

Here's the story:

#1 America is filled with people who don't care about America. They care about themselves, and their feelings, not their reality. 

#2 Democrats obviously don't understand microtargeting, the whole "Latinos are not all the same" isnt just Latinos, it's blacks, whites, asian, and even within that rural suburban and urban. Further than that it's income, donation history etc.

With modern computing it's not difficult to do the targeting you need to do to get the turnout that you need to get... Trump did it. 2 elections in a row. 

#3 Texans have alienated hispanics and blacks in very bad ways. And that starts locally not nationally. 

The crypt keeper democratic leadership will continue to do things as they have, as is tradition. Microtargeting? A data strategy? What are you, some junior know-nothing from Illinois?

  • Hook 'Em 1
  • Like 1
Link to comment
Share on other sites

The polling is the major issue here. Trumpism and social media are variables that pollsters just can't property account for apparently. 

People are getting their "news" from all corners of the internet and no one knows the fuck what to do about it. The days of reading the newspaper and watching 3 major TV networks made it easier to understand a voter's mind. It's a total circus now. 

 

 

Link to comment
Share on other sites

41 minutes ago, BradInATX said:

That's an overly simplistic outlook. Okay, they correctly predicted something that they gave a 90% chance. But they missed badly in almost every state that matters, just like in 2016. Outside of margin of error.

If any real data science pro turned in a project with 10 projections and missed outside of MOE on all 10, they wouldn't have a job anymore. That's what 538 just did. They missed the 8-10 most important data points outside of MOE. That's atrocious. I understand it's the polling that's the issue. But we may need to come to grips with elections not really being attainable to model and predict with any real accuracy.

Unless it can be proven that the 538 simulation methodology is flawed, it's entirely a polling problem.  I think it's reasonable to say that it may turn out that predictions off polling data (or more correctly, aggregated polling data) are a waste of time.

  • Hook 'Em 2
Link to comment
Share on other sites

3 hours ago, BehoId, The Underminer! said:

House, Senate, and EC are all coming up on the very edge of any reasonable confidence interval, and all in the same direction.  Just coincidental noise, I’m sure.  Quite the probability defying event.  

You expect them to be independent events?

Link to comment
Share on other sites

So if Biden wins Wisconsin, Michigan, Nevada, Pennsylvania, and Georgia and gets to 306 electoral votes, was 538 still wrong?  Just checking.

By the way I have no specific love for Nate Silver.  But the Cloak Room mid election looks just like the football board when Texas is down by 7.  

  • Hook 'Em 4
Link to comment
Share on other sites

12 minutes ago, jimmyjazz said:

Unless it can be proven that the 538 simulation methodology is flawed, it's entirely a polling problem.  I think it's reasonable to say that it may turn out that predictions off polling data (or more correctly, aggregated polling data) are a waste of time.

Yeah, we're saying the same thing there. If I haven't been clear, I'm not blaming Silver. I'm just saying he might be trying to perform a type of science that isn't really possible to perform with any accuracy, by no fault of his.

  • Hook 'Em 2
Link to comment
Share on other sites

 

2 minutes ago, BradInATX said:

Yeah, we're saying the same thing there. If I haven't been clear, I'm not blaming Silver. I'm just saying he might be trying to perform a type of science that isn't really possible to perform with any accuracy, by no fault of his.

He should still beat simple polling averages ala RCP, which he has failed to do in every presidential election.   He somehow manages to make the wrong polls more wrong.

Link to comment
Share on other sites

7 minutes ago, fellside said:

So if Biden wins Wisconsin, Michigan, Nevada, Pennsylvania, and Georgia and gets to 306 electoral votes, was 538 still wrong?  Just checking.

By the way I have no specific love for Nate Silver.  But the Cloak Room mid election looks just like the football board when Texas is down by 7.  

Biden is going to win by decimal points but he should’ve won by five, six, seven, or eight points in some of the states. If all you can get your head around is the black or white win or a loss, feel free to look at the individual senators and representatives that won’t get over the line there. This isn’t exactly a nuanced and obscure point.

Link to comment
Share on other sites

I'm not sure it was as much the polling that was done was flawed as much as the polling that wasn't done or given enough weight.

Everything angled towards to you like or dislike Trump.  I think many people honestly answered that they don't like him, but what wasn't asked was how much a factor that was going to be on their actual vote.

The economy has always been a big vote sway, yet this year I heard next to nothing related to it, but now it's showing up all over the place as a decider.

Link to comment
Share on other sites

I also have a bit of a theory working in my head that the “weighting by education“ is probably not as real or accurate as we pretend it is. I feel like pollsters like to use it because it Feeds their ego. I feel like we like to repeat it because it feeds our ego. My bet is that it’s way way more sophisticated than that, or just a way way smaller factor.

 

there are so many guys that I know that are the exact same person that they were before they went to college. 

Link to comment
Share on other sites

4 hours ago, longhornmatt said:

Consistently getting the probabilities wrong by a material amount can happen even if you nail the ultimate winner or loser.  Vegas doesn’t get credit for setting a -21 line on a game if the favorite ends up winning by 3.  Over and over.

How did they get the probability wrong? 

Link to comment
Share on other sites



×
×
  • Create New...