Total Posts:15|Showing Posts:1-15
Jump to topic:

Why Five Thirty Eight is Failing Now

YYW
Posts: 36,282
Add as Friend
Challenge to a Debate
Send a Message
11/5/2016 9:07:00 PM
Posted: 4 weeks ago
http://www.rawstory.com...

Five Thirty Eight (read: may have one of the most accurate prediction models available) is not representing reality, because it can't. Nate Silver, who built Five Thirty Eight tells us why:

...the wild swings toward Trump are more a sign that an orange-furred monkey wrench has jammed FiveThirtyEight"s soothsaying machine than a candidate who is detested by nearly 60 percent of voters is suddenly floating to the top of the 2016 shitshow.

There are so many moving parts in this election that "likely voters" who are polled do not reflect a representative sample of where the country is. The reason is because Trump has destroyed traditional coalitions.

The main issue is FiveThirtyEight works best when electoral coalitions are well-defined and change slowly as they have been since Bill Clinton was in office. Trump has scrambled all that, however, and analyzing polls cannot account for rapid demographic shifts.

It looks like Trump is leading the working class, but that's not likely to be accurate in the polls. He leads with working class men who are white, and no one else.

Trump"s bastion of working-class white men includes significant numbers of Democrats who"ve crossed over since 2012. ...The focus on Trump"s working-class support also misses one crucial aspect: a majority of the white working class is female, and they began rejecting Trump after the Access Hollywood tape scandal. And the gender gap is wider than ever, with women favoring Clinton over Trump by an average of 16 points in October polls.

Trump has also also fragmented the GOP's popular outreach to virtually every single minority group in the United States... except uneducated WASPS.

The college educated, and the generally highly educated, disdain him.

But he"s alienated typically Republican college-educated whites and spurred Latinos, Asians, and Muslims to mobilize in record numbers to defeat him.

Demographic shifts have changed the game, too. Texas could still go blue, and that's looking more likely because of high early voting turn out, even though representing the shift is hard to do. Georgia may also still be in play. We could see the entire Eastern Seaboard (other than South Carolina) go blue.

Two other significant reasons why FiveThirtyEight overestimates Trump"s chances is early voting and get out the vote. More than 35 million Americans have already voted, which is 28 percent of the total vote in 2012. In nearly every battleground state early voting is outpacing last election, meaning Clinton banked millions of votes when her numbers were peaking. One poll of early voters indicates Clinton is ahead by enough 5 points, which is enough to tack on a 1.3 percent advantage in the final election results.

Bottom line:

FiveThirtyEight"s model worked fantastically well when the sailing was smooth, but it is foundering in the violent seas of 2016. The Cook Political Report, the brainchild of the dean of election forecasting, Charles Cook, observes that despite Clinton"s eleventh-hour woes, "The race has tightened to its "natural resting place" with a 2-4 point lead for Clinton," while Trump"s "path to 270 electoral votes remains decidedly and almost impossibly narrow."
Tsar of DDO
Geographia
Posts: 1,467
Add as Friend
Challenge to a Debate
Send a Message
11/6/2016 2:04:16 AM
Posted: 4 weeks ago
At 11/5/2016 9:07:00 PM, YYW wrote:
http://www.rawstory.com...

Five Thirty Eight (read: may have one of the most accurate prediction models available) is not representing reality, because it can't. Nate Silver, who built Five Thirty Eight tells us why:

...the wild swings toward Trump are more a sign that an orange-furred monkey wrench has jammed FiveThirtyEight"s soothsaying machine than a candidate who is detested by nearly 60 percent of voters is suddenly floating to the top of the 2016 shitshow.

There are so many moving parts in this election that "likely voters" who are polled do not reflect a representative sample of where the country is. The reason is because Trump has destroyed traditional coalitions.

The main issue is FiveThirtyEight works best when electoral coalitions are well-defined and change slowly as they have been since Bill Clinton was in office. Trump has scrambled all that, however, and analyzing polls cannot account for rapid demographic shifts.

It looks like Trump is leading the working class, but that's not likely to be accurate in the polls. He leads with working class men who are white, and no one else.

Trump"s bastion of working-class white men includes significant numbers of Democrats who"ve crossed over since 2012. ...The focus on Trump"s working-class support also misses one crucial aspect: a majority of the white working class is female, and they began rejecting Trump after the Access Hollywood tape scandal. And the gender gap is wider than ever, with women favoring Clinton over Trump by an average of 16 points in October polls.

Trump has also also fragmented the GOP's popular outreach to virtually every single minority group in the United States... except uneducated WASPS.

The college educated, and the generally highly educated, disdain him.

But he"s alienated typically Republican college-educated whites and spurred Latinos, Asians, and Muslims to mobilize in record numbers to defeat him.

Demographic shifts have changed the game, too. Texas could still go blue, and that's looking more likely because of high early voting turn out, even though representing the shift is hard to do. Georgia may also still be in play. We could see the entire Eastern Seaboard (other than South Carolina) go blue.

Two other significant reasons why FiveThirtyEight overestimates Trump"s chances is early voting and get out the vote. More than 35 million Americans have already voted, which is 28 percent of the total vote in 2012. In nearly every battleground state early voting is outpacing last election, meaning Clinton banked millions of votes when her numbers were peaking. One poll of early voters indicates Clinton is ahead by enough 5 points, which is enough to tack on a 1.3 percent advantage in the final election results.

Bottom line:

FiveThirtyEight"s model worked fantastically well when the sailing was smooth, but it is foundering in the violent seas of 2016. The Cook Political Report, the brainchild of the dean of election forecasting, Charles Cook, observes that despite Clinton"s eleventh-hour woes, "The race has tightened to its "natural resting place" with a 2-4 point lead for Clinton," while Trump"s "path to 270 electoral votes remains decidedly and almost impossibly narrow."

What timezone are you in? I want to watch the shitshow with you.
YYW
Posts: 36,282
Add as Friend
Challenge to a Debate
Send a Message
11/6/2016 2:08:45 AM
Posted: 4 weeks ago
At 11/6/2016 2:04:16 AM, Geographia wrote:
At 11/5/2016 9:07:00 PM, YYW wrote:
http://www.rawstory.com...

Five Thirty Eight (read: may have one of the most accurate prediction models available) is not representing reality, because it can't. Nate Silver, who built Five Thirty Eight tells us why:

...the wild swings toward Trump are more a sign that an orange-furred monkey wrench has jammed FiveThirtyEight"s soothsaying machine than a candidate who is detested by nearly 60 percent of voters is suddenly floating to the top of the 2016 shitshow.

There are so many moving parts in this election that "likely voters" who are polled do not reflect a representative sample of where the country is. The reason is because Trump has destroyed traditional coalitions.

The main issue is FiveThirtyEight works best when electoral coalitions are well-defined and change slowly as they have been since Bill Clinton was in office. Trump has scrambled all that, however, and analyzing polls cannot account for rapid demographic shifts.

It looks like Trump is leading the working class, but that's not likely to be accurate in the polls. He leads with working class men who are white, and no one else.

Trump"s bastion of working-class white men includes significant numbers of Democrats who"ve crossed over since 2012. ...The focus on Trump"s working-class support also misses one crucial aspect: a majority of the white working class is female, and they began rejecting Trump after the Access Hollywood tape scandal. And the gender gap is wider than ever, with women favoring Clinton over Trump by an average of 16 points in October polls.

Trump has also also fragmented the GOP's popular outreach to virtually every single minority group in the United States... except uneducated WASPS.

The college educated, and the generally highly educated, disdain him.

But he"s alienated typically Republican college-educated whites and spurred Latinos, Asians, and Muslims to mobilize in record numbers to defeat him.

Demographic shifts have changed the game, too. Texas could still go blue, and that's looking more likely because of high early voting turn out, even though representing the shift is hard to do. Georgia may also still be in play. We could see the entire Eastern Seaboard (other than South Carolina) go blue.

Two other significant reasons why FiveThirtyEight overestimates Trump"s chances is early voting and get out the vote. More than 35 million Americans have already voted, which is 28 percent of the total vote in 2012. In nearly every battleground state early voting is outpacing last election, meaning Clinton banked millions of votes when her numbers were peaking. One poll of early voters indicates Clinton is ahead by enough 5 points, which is enough to tack on a 1.3 percent advantage in the final election results.

Bottom line:

FiveThirtyEight"s model worked fantastically well when the sailing was smooth, but it is foundering in the violent seas of 2016. The Cook Political Report, the brainchild of the dean of election forecasting, Charles Cook, observes that despite Clinton"s eleventh-hour woes, "The race has tightened to its "natural resting place" with a 2-4 point lead for Clinton," while Trump"s "path to 270 electoral votes remains decidedly and almost impossibly narrow."

What timezone are you in? I want to watch the shitshow with you.

Central
Tsar of DDO
YYW
Posts: 36,282
Add as Friend
Challenge to a Debate
Send a Message
11/6/2016 2:42:10 AM
Posted: 4 weeks ago
At 11/6/2016 2:19:23 AM, FourTrouble wrote:
Polls in general aren't very helpful. Are you familiar with civisanalytics? They're in Chicago.

Never heard of them.
Tsar of DDO
dylancatlow
Posts: 12,244
Add as Friend
Challenge to a Debate
Send a Message
11/6/2016 5:40:58 PM
Posted: 4 weeks ago
The recent anti-Nate Silver article put out by Huffington Post is one of the most retarded I've seen. At the end, the author actually claims that the debate between Nate, who's more optimistic than most about Trump's chances, and analyses like Huffington Post's which puts his chances at 2 percent, will be "settled" once the election results come in. The implication is that if Clinton wins those who said she had a 98 percent chance to win will be vindicated, while people like Nate will be put to shame. That would be somewhat plausible if Nate predicted a Clinton loss, but he doesn't. He has Clinton in the lead by a reasonable margin, so a Clinton win would only somewhat raise the probability that he was overly optimistic about Trump. The rest of the article is not much better.
dylancatlow
Posts: 12,244
Add as Friend
Challenge to a Debate
Send a Message
11/6/2016 5:46:19 PM
Posted: 4 weeks ago
I'm not saying that Nate is perfect, who seems to have been somewhat traumatized by his poor prediction performance in the primaries, but a lot of the critiques against him don't seem to be based on anything except "he says X, but I don't want X to be true".
NHN
Posts: 624
Add as Friend
Challenge to a Debate
Send a Message
11/6/2016 7:10:48 PM
Posted: 4 weeks ago
At 11/5/2016 9:07:00 PM, YYW wrote:
TL;DR
The blog 538 is ESPN's contribution to election-day excitement. Nate Silver pulled the same stunt in 2012, making it appear as a close election.

If you want an accurate forecast, based on Bayesian probability, go here instead (http://election.princeton.edu...).

Enjoy (if you'll allow yourself).
dylancatlow
Posts: 12,244
Add as Friend
Challenge to a Debate
Send a Message
11/6/2016 8:08:21 PM
Posted: 4 weeks ago
At 11/6/2016 7:10:48 PM, NHN wrote:
At 11/5/2016 9:07:00 PM, YYW wrote:
TL;DR
The blog 538 is ESPN's contribution to election-day excitement. Nate Silver pulled the same stunt in 2012, making it appear as a close election.

If you want an accurate forecast, based on Bayesian probability, go here instead (http://election.princeton.edu...).

Enjoy (if you'll allow yourself).

Three days before election day Nate Silver gave Romney a 9 percent chance to win. I wouldn't call that close. He did have Romney up to 37 at one point.
dylancatlow
Posts: 12,244
Add as Friend
Challenge to a Debate
Send a Message
11/6/2016 8:25:20 PM
Posted: 4 weeks ago
At 11/6/2016 7:10:48 PM, NHN wrote:
At 11/5/2016 9:07:00 PM, YYW wrote:
TL;DR
The blog 538 is ESPN's contribution to election-day excitement. Nate Silver pulled the same stunt in 2012, making it appear as a close election.

If you want an accurate forecast, based on Bayesian probability, go here instead (http://election.princeton.edu...).

Enjoy (if you'll allow yourself).

Interestingly enough, the disparity between the 538 forecast and Princeton's forecast isn't even captured very well by the current difference. Back in late July Nate silver put Trump's chances at 50.1% and Princeton had BELOW where they do now - like 1/500. One of these forecasts is really, really wrong.

http://election.princeton.edu...
NHN
Posts: 624
Add as Friend
Challenge to a Debate
Send a Message
11/6/2016 8:26:17 PM
Posted: 4 weeks ago
At 11/6/2016 8:08:21 PM, dylancatlow wrote:
At 11/6/2016 7:10:48 PM, NHN wrote:
At 11/5/2016 9:07:00 PM, YYW wrote:
TL;DR
The blog 538 is ESPN's contribution to election-day excitement. Nate Silver pulled the same stunt in 2012, making it appear as a close election.

If you want an accurate forecast, based on Bayesian probability, go here instead (http://election.princeton.edu...).

Enjoy (if you'll allow yourself).
Three days before election day Nate Silver gave Romney a 9 percent chance to win. I wouldn't call that close. He did have Romney up to 37 at one point.
Look-see here (http://fivethirtyeight.blogs.nytimes.com...). Something changed dramatically in the last week, otherwise it remained close in a way that other forecasters did not present.
NHN
Posts: 624
Add as Friend
Challenge to a Debate
Send a Message
11/6/2016 8:48:41 PM
Posted: 4 weeks ago
At 11/6/2016 8:25:20 PM, dylancatlow wrote:
At 11/6/2016 7:10:48 PM, NHN wrote:
At 11/5/2016 9:07:00 PM, YYW wrote:
TL;DR
The blog 538 is ESPN's contribution to election-day excitement. Nate Silver pulled the same stunt in 2012, making it appear as a close election.

If you want an accurate forecast, based on Bayesian probability, go here instead (http://election.princeton.edu...).

Enjoy (if you'll allow yourself).

Interestingly enough, the disparity between the 538 forecast and Princeton's forecast isn't even captured very well by the current difference. Back in late July Nate silver put Trump's chances at 50.1% and Princeton had BELOW where they do now - like 1/500. One of these forecasts is really, really wrong.

http://election.princeton.edu...
PEC changes its election forecast logarithm as the election nears. I have no idea where you got that 1/500 number from.
dylancatlow
Posts: 12,244
Add as Friend
Challenge to a Debate
Send a Message
11/6/2016 8:50:45 PM
Posted: 4 weeks ago
At 11/6/2016 8:48:41 PM, NHN wrote:
At 11/6/2016 8:25:20 PM, dylancatlow wrote:
At 11/6/2016 7:10:48 PM, NHN wrote:
At 11/5/2016 9:07:00 PM, YYW wrote:
TL;DR
The blog 538 is ESPN's contribution to election-day excitement. Nate Silver pulled the same stunt in 2012, making it appear as a close election.

If you want an accurate forecast, based on Bayesian probability, go here instead (http://election.princeton.edu...).

Enjoy (if you'll allow yourself).

Interestingly enough, the disparity between the 538 forecast and Princeton's forecast isn't even captured very well by the current difference. Back in late July Nate silver put Trump's chances at 50.1% and Princeton had BELOW where they do now - like 1/500. One of these forecasts is really, really wrong.

http://election.princeton.edu...
PEC changes its election forecast logarithm as the election nears. I have no idea where you got that 1/500 number from.

There's no published data which actually says 1/500, it's just my best attempt to interpret this graph: http://election.princeton.edu...

95 percent of all election outcomes are presumed to fall within the gray area, so 2.5 percent are below that.
FourTrouble
Posts: 12,757
Add as Friend
Challenge to a Debate
Send a Message
11/6/2016 11:35:31 PM
Posted: 4 weeks ago
At 11/6/2016 2:42:10 AM, YYW wrote:
At 11/6/2016 2:19:23 AM, FourTrouble wrote:
Polls in general aren't very helpful. Are you familiar with civisanalytics? They're in Chicago.

Never heard of them.

Ah, they do awesome stuff, you should check 'em out.
YYW
Posts: 36,282
Add as Friend
Challenge to a Debate
Send a Message
11/7/2016 12:31:46 AM
Posted: 4 weeks ago
At 11/6/2016 11:35:31 PM, FourTrouble wrote:
At 11/6/2016 2:42:10 AM, YYW wrote:
At 11/6/2016 2:19:23 AM, FourTrouble wrote:
Polls in general aren't very helpful. Are you familiar with civisanalytics? They're in Chicago.

Never heard of them.

Ah, they do awesome stuff, you should check 'em out.

I'll have to check them out
Tsar of DDO