• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

PoliGAF 2016 |OT14| Attention NV shoppers, democracy is on sale in aisle 4!

Status
Not open for further replies.
V

Voysa Reezun

Banned
Emarv said:
Again, I think it's a little much to dogpile on Nate when he's giving sound reasoning at a macro level. Litigating some of the details is definitely warranted. But his notion of using a model that correlates the states and takes into account National Polls has not been entirely disputed yet, scientifically.

Wang's notion of a state independent meta-margin is also a defensible path to this whole thing. Our sample size is so small and overall polling is very easy to "call" and defend your method as the correct one.

The problem is that Nate needs to stop being thin-skinned and continue to help educate the public on his model's successes and failings. I bet it must be super frustrating for non-math people to misconstrue you, but that's how math and science works. That's the field he's in.
Click to expand...

I'm fine with focusing on national polls as well as state polls, but I just wish that I understood more about stuff like polls-plus, how different elements like the economy or the sitting president's popularity are weighted and why, why he didn't build his model to consider early voting numbers, and whether or not he would make that change and why, etc.

I feel like he did lay out his argument for endorsements being weighted into the primary numbers, but when that whole model blew up, Bronze sort of stopped explaining what he's actually doing over there.

I've only taken Stats 101, so I am not a statistician by any means, and I don't mind learning! But Bronze throwing hissy-fits online and accusing other people of ignoring data-based evidence is absurd and puts me right off of this dude.

Plus, 538's podcast is a dumpster fire of shitty analysis and stupid jokes. That podcast has done more to cool me off of Nate's work than anything else he's done. If you think the shitty punditry on his Twitter feed is bad, just listen to him hash out the race with Clare Malone and Jody Avirgan. Sweet fuck, they care about narratives as much as a dunce like Mark Halperin does, except that I expect better from 538.
 
H

HanShotFirst

Member
I wonder how many newly registered voters show up in midterms and in future elections. Not saying Democrats win anything in 2018 but will 2018 be better than 2014 for Democrats?
 
B

brianmcdoogle

Member
Maybe I'm just being an ass, but a lot of this tweet storm is "my model is tested, my model uses proof, my model isn't partisan, and everyone else's model is untested, lacking proof, and is partisan."

Like, he's saying to other forecasters "prove your model", and... they are? You can see their data and their models and read their reasons for why they are doing what they've done. And even some of them aren't using any tilt factors like a "polls plus" forecast.

"Your model is untested". Uh, how long and how accurate have these other forecasters been in previous elections, including this one?

And then, even if you do all the stuff he wants a forecaster to do, like have it be tested, use proof, not be partisan, have a track record of doing well in forecasting, and whatnot... if your model is too high in its prediction, then it's wrong anyway?

Eh.

Voysa Reezun said:
Plus, 538's podcast is a dumpster fire of shitty analysis and stupid jokes. That podcast has done more to cool me off of Nate's work than anything else he's done. If you think the shitty punditry on his Twitter feed is bad, just listen to him hash out the race with Clare Malone and Jody Avirgan. Sweet fuck, they care about narratives as much as a dunce like Mark Halperin does, except that I expect better from 538.
Click to expand...

This, too. I thought FiveThirtyEight was founded to not be this type of punditry hackery of "Trump has less of a chance of being the nominee than being in another Home Alone movie"?
 
A

Aaron Strife

Banned
brianmcdoogle said:
Maybe I'm just being an ass, but a lot of this tweet storm is "my model is tested, my model uses proof, my model isn't partisan, and everyone else's model is untested, lacking proof, and is partisan."

Like, he's saying to other forecasters "prove your model", and... they are? You can see their data and their models and read their reasons for why they are doing what they've done. And even some of them aren't using any tilt factors like a "polls plus" forecast.

"Your model is untested". Uh, how long and how accurate have these other forecasters been in previous elections, including this one?

And then, even if you do all the stuff he wants a forecaster to do, like have it be tested, use proof, not be partisan, have a track record of doing well in forecasting, and whatnot... if your model is too high in its prediction, then it's wrong anyway?

Eh.



This, too. I thought FiveThirtyEight was founded to not be this type of punditry hackery of "Trump has less of a chance of being the nominee than being in another Home Alone movie"?
Click to expand...
His model had Obama as a 90% favorite in 2012 when the polls were worse for him than they are for Hillary, I don't get how projecting a more certain outcome should be a knock against anyone else.

Nate is such a child.
 
J

Joe Shlabotnik

Banned
Green Yoshi said:
If voter turnout is high, does this increase the chances of the Democrats to win the majority in the congress? 2016 was a horrible year for the Republican party after all.
Click to expand...

Depends on the makeup of the electorate, but generally high turnout is better for Dems as there is a larger party ID for Democrats.

But also I prefer higher turnout whatever the result because at least it approaches a more honest evaluation of the will of the country. If we're largely comprised of racist fucks, well, we might as well know about it.
 
T

thcsquad

Member
My thoughts on 538's model: not about the unskewing, but about EV and Hispanic turnout (and I guess polls in general).

We all see the polling problems with Hispanic voters. If a model ignores persistent Hispanic polling problems with otherwise good pollsters and blindly uses them in heavily Hispanic states, the model isn't comprehensive enough to be trustworthy.

In the past, Hispanic voters didn't turn out as reliably as other demographics, and most pollsters are assuming this to be true. What we're seeing is that this election is different. We've been saying this is different since Trump became the frontrunner, that he would be the thing that got Hispanic turnout up. How would a model predict this in an objective way?

Well, there are some pollsters who seem to be getting this right, like Saguaro. A model could weight them much more heavily in polls of states like NV/AZ/TX than a pollster who otherwise is solid but doesn't bother with Spanish speakers.

We're seeing strong evidence that the Hispanic turnout will increase. What I'm curious about is if it will be heavily pronounced in states other than NV. If NV is the only state with a large Hispanic population that deviates from the polls a lot, that would be one thing, and maybe you could attribute it to Harry Reid (plus the fact that polls in NV have been off for ages).

But what about other states? if FL turns out to be within one point in either direction, AZ is Trump+4, TX is Trump+9, I think 538's model will be vindicated (at least in this aspect...poll trendlines aside). If we add the Clinton + 5 movement that some early turnout numbers suggest, and she takes AZ and keeps TX within five, Nate has to do some serious thinking about basing a model on polls and pollsters that have such an obvious blind spot.
 
E

Emarv

Member
Voysa Reezun said:
I'm fine with focusing on national polls as well as state polls, but I just wish that I understood more about stuff like polls-plus, how different elements like the economy or the sitting president's popularity are weighted and why, why he didn't build his model to consider early voting numbers, and whether or not he would make that change and why, etc.

I feel like he did lay out his argument for endorsements being weighted into the primary numbers, but when that whole model blew up, Bronze sort of stopped explaining what he's actually doing over there.

I've only taken Stats 101, so I am not a statistician by any means, and I don't mind learning! But Bronze throwing hissy-fits online and accusing other people of ignoring data-based evidence is absurd and puts me right off of this dude.
Click to expand...

I agree that transparency to polls-plus would be better, but 1) He's running a commercial site, so his "secret sauce" is his own to do what he wants and 2) it doesn't explain the difference between his Polls-Only model and that of his peers.

What's separating his Polls-Only from Upshot, Huffington, etc, is the issue. Not his weighting of the economy or Obama's popularity.
 
G

geomon

Member
Stephen Hawking Issues One Perfect Sentence On Donald Trump And His Supporters

So how do you explain Trump’s continued popularity? Perhaps the physicist can shed some light on the situation?

“I can’t,” Stephen Hawking said, after being asked in a interview with CNN affiliate ITV to explain Trump’s unprecedented political ascension.

For someone who has the ability to explain the unexplainable, this observation is a bit disappointing. However, the (arguably) smartest person on Earth did go on to offer his personal take on the man:

“He is a demagogue, who seems to appeal to the lowest common denominator.”


Trump has yet to comment, but don't be surprised if there’s a tweet storm a’ brewin’.
Click to expand...
 
H

HanShotFirst

Member
Manmademan said:
probably not. maybe 2022 once millennials are comfortably middle aged.
Click to expand...

Democrats received 44.9% of the vote in 2012 House elections. In 2014 they received 45.5%. What I'm saying is, will there be another slight improvement? Especially with new voters? Like maybe they get anywhere from 46% to 46.5%?
 
E

Emarv

Member
The Hispanic EV point others have mentioned is interesting.

Personally, I believe there were 2 big polling issues during the primary: 1) Bernie winning Michigan was a polling miss. 2) The big one - Latinos were always a polling miss.

I'm not sure how much Nate worked in a Latino polling miss into his model, but he definitely was aware of it. Like thcsquad said above, this election is definitely different with regard to Latino turnout and skewing that demographic correctly is hard.


But what Nate and Wang are doing is still sound. Focusing on Polls-Only is sound, even to the point of not including EV. We'll never have the data that campaigns have to correctly analyze EV numbers. It'll only ever be reports from states that give that stuff out. Maybe down the line, someone will get that model down pat, but for now, a polls-only model is perfectly defensible and produces solid results.

Like others have said, the point of a good website built around a model is to help contextualize it. Harry's next piece should basically be "Our model says Nevada is Red but here's what it doesn't understand about what we now know" and there should be no shame in that.

The 538 model's major problem is still Nate's aggressive trendlines and reaction to new polls and his relatively high uncertainty weighting despite both (relatively) consistent polling and decreased 3rd party voters.
 
T

thepotatoman

Unconfirmed Member
https://twitter.com/NateSilver538/status/794996132946608128
@NateSilver538 said:
It's wrong to show Clinton with a 6-point lead (as per HuffPo) when **almost no national poll shows that**. Doesn't reflect the data.
Click to expand...

If you want HuffPo to be more reactive to recent movement in polls, you can do that with an simple option change.

54D0qWE.png
Click to expand...

It's fine if you want to only see stuff from the last 2 weeks, but it makes it harder to see the big picture depending on where in the noise you are, and it seems to me like the ones that have tried to filter out the noise with a longer term view have been proven right thus far, with the polls always reverting to that norm.

Maybe now that we're so close to election day, the recency of the polls should be made more important, but I don't know if the evidence really points to that being helpful. It seems to me like 2012 and 2014 results were closer to the long term averages than the shorter term ones.
 
B

brianmcdoogle

Member
Aaron Strife said:
His model had Obama as a 90% favorite in 2012 when the polls were worse for him than they are for Hillary, I don't get how projecting a more certain outcome should be a knock against anyone else.

Nate is such a child.
Click to expand...

Maybe I'm misremembering or misunderstanding a tweet, but I think Silver blocked Wang on twitter a while ago, too.

"No child. No child. You're the child!"
 
M

Manmademan

Member
HanShotFirst said:
Democrats received 44.9% of the vote in 2012 House elections. In 2014 they received 45.5%. What I'm saying is, will there be another slight improvement? Especially with new voters? Like maybe they get anywhere from 46% to 46.5%?
Click to expand...

well yes- younger generations (millennials, gen-x to a lesser extent) trend more democratic than boomers do. as boomers die off and younger generations become older and more engaged, that's going to shift.

Gerrymandering has kind of fucked democrats though, they need to overperform republicans for equal representation. it'll be a bit before midterms stop screwing them.
 
J

Joe Shlabotnik

Banned
Yeah, Pollster's "we got this" to "panic!" range depends entirely on the smoothing options. Though even the lowest smoothing still shows a 2-point Clinton lead, of course.
 
C

cartoon_soldier

Member
Y2Kev said:
She's just countering Trump.
Click to expand...

Sending both top people there? Obviously the internals also show that the lead is soft.

The lack of quality polling is really bad this year.

RNC is also telling reporters that they see Michigan in their path. Internals have caught something.
 
T

Tall4Life

Member
Green Yoshi said:
Is there any chance the Supreme Court can do something about gerrymandering?
Click to expand...
They've ruled previously that each district has to be similar in size and have a compact shape, among others. Redistricting based on partisanship is allowed but it cant be grossly partisan. Some of that might be wrong, but thats what I remember. It protects against some of the more obvious and weird redistricting, like that snake district awhile back, but it still allows for gerrymandering.

If a liberal SC doesn't uphold precedent (most SCs do), then they might be able to change it. Otherwise, it is constitutional.
 
H

HanShotFirst

Member
Manmademan said:
well yes- younger generations (millennials, gen-x to a lesser extent) trend more democratic than boomers do. as boomers die off and younger generations become older and more engaged, that's going to shift.

Gerrymandering has kind of fucked democrats though, they need to overperform republicans for equal representation. it'll be a bit before midterms stop screwing them.
Click to expand...

I know it won't amount to much but I'm hoping for a full percent swing instead of a half percent swing lol. some of these new voters have to stick right?
 
G

Gotchaye

Member
I think 538's model should be taken pretty seriously now. I have big issues with how variable it's been but the obvious problem there was always that it was neglecting the possibility of movement in the time left before the election. Now that there's no time left before the election the only thing that was clearly fishy isn't an issue anymore.
 
Y

Y2Kev

TLG Fan Caretaker Est. 2009
cartoon_soldier said:
Sending both top people there? Obviously the internals also show that the lead is soft.

The lack of quality polling is really bad this year.

RNC is also telling reporters that they see Michigan in their path. Internals have caught something.
Click to expand...

Well I agree the lead is on the softer side if I had to guess, but MI is in their path because is has to be...
 
B

Boke1879

Member
cartoon_soldier said:
Sending both top people there? Obviously the internals also show that the lead is soft.

The lack of quality polling is really bad this year.
Click to expand...

I mean we don't know what it means. If the lead is soft or not I'm glad they are treating it like it is. I want her and her surrogates to campaign until the last possible minute. Meanwhile you have the other candidate scrambling all over.
 
E

Emarv

Member
More from Sam Wang. I think this is totally fair and makes sense.

Sam WangVerified account
‏@SamWangPhD
Been thinking. I think polling-error parameters feeding PEC probability should probably be reset, which would give 91-93%. Below that, nah.
Atticus Goldfinch
‏@AtticusGF
@SamWangPhD @ClaraJeffery Nate seems to be implying that all the other models are overconfident. Thoughts?
Click to expand...
 
M

Maengun1

Member
Nate can defend his model all he wants. My issue with him this whole year has been that he writes 15 articles a day...when Clinton has a 90%+ chance to win even in his model, his articles are like "CLINTON'S LEAD COULD COLLAPSE AT ANY SECOND AHHH
panic3_zpsoerys3kb.gif
" and even when his own model still has her winning 2/3 of the time the tone of his tweets/articles are like "HILLARY DOOMED ALL HAIL PRESIDENT TRUMP AHHH
panic3_zpsoerys3kb.gif
." He follows up with the "well you see, uncertainty and there's some very good reasons my model acts this way blah blah blah" but he always leads with "FREAKOUT!!!" And then "I told ya so" smugness.

I personally have largely tried to ignore the site all cycle but they ARE still seen as the no. 1 for most people which means that every 20 minutes when he posts a new article it gets shared in every corner of the internet for a fresh meltdown. Now he's pissy that people are fed up with him. Tough shit IMO. Enjoy the attention you've been seeking.

Harry eternal <3 though
 
M

mo60

Member
Her lead in MI probably went from the high single digits/low double digits to mid to high single digits. Also, since MI does not have early voting her campaign is trying to get people in MI to vote on November 8th.
 
T

Third Strike

Member
Manmademan said:
well yes- younger generations (millennials, gen-x to a lesser extent) trend more democratic than boomers do. as boomers die off and younger generations become older and more engaged, that's going to shift.

Gerrymandering has kind of fucked democrats though, they need to overperform republicans for equal representation. it'll be a bit before midterms stop screwing them.
Click to expand...

Will be interesting to see how things pan out in Florida, Virginia, and North Carolina. All three had their maps changed recently, and this will be the first election where they take effect. Would be great if it resulted in Democrats at least picking up one extra seat in each state.
 
T

Thunder Monkey

Banned
I've been telling you guys.

The biggest story of the election is going to be the increase in the latino vote. The second is going to be Dems losing their grip on the rustbelt. It's not going to hurt them this election, but it's still going to be true.
 
G

Grief.exe

Member
Thunder Monkey said:
I've been telling you guys.

The biggest story of the election is going to be the increase in the latino vote. The second is going to be Dems losing their grip on the rustbelt. It's not going to hurt them this election, but it's still going to be true.
Click to expand...

The irony is the Democrat party's economics would objectively help that section, but the Republican narrative is so much juicier.
 
M

MassiveNights

Banned
Gotchaye said:
I think 538's model should be taken pretty seriously now. I have big issues with how variable it's been but the obvious problem there was always that it was neglecting the possibility of movement in the time left before the election. Now that there's no time left before the election the only thing that was clearly fishy isn't an issue anymore.
Click to expand...

I mean the model still has Hillary as a 2 to 1 favourite. That's about where my gut is on all this.

The perception of the difference between 538 and other models 90%+ is probably bigger than the reality of the difference.

Some of Nate's articles have been ropey but the team there have provided a pretty consistent message on their website and podcasts.

I'm not sure why people get so angry about it anyway and I'm not sure why so many people want him to have egg on his face on Nov 9th. He's never said he thinks Trump will win.
 
U

User 463088

Banned
Michigan has no early voting. (or almost no early voting). You are totally dependent upon your day of turnout. Demographically, Michigan is better for her than Wisconsin, yet we know Trump has given up on Wisconsin because of early voting. With PA and MI, you have one chance to get this thing right. It makes sense to go there.
 
Status
Not open for further replies.

Similar threads

  • Locked
PoliGAF 2017 |OT6| Made this thread during Harvey because the ratings would be higher
338 339 340 341 342
17K
2M
UselessNapkin replied
  • Locked
PoliGAF 2017 |OT5| The Man In the High Chair
395 396 397 398 399
20K
2M
RDreamer replied
  • Locked
PoliGAF 2017 |OT4| The leaks are coming from inside the white house
396 397 398 399 400
20K
2M
ShironRedshift replied
  • Locked
PoliGAF 2017 |OT3| 13 Treasons Why
393 394 395 396 397
20K
2M
Zaraki_Kenpachi replied
  • Locked
Poligaf 2017 |OT| It Almost Isn't Fair
9
4K
Bonen no Max'd replied
Top Bottom