538 Model Megathread

[Erich Maria Remarque]

Ezra KleinVerified account
‏@ezraklein
The logic behind the 538 model's relatively pro-Trump results, explained:

http://www.vox.com/2016/11/3/13147678/nate-silver-fivethirtyeight-trump-forecast

You must be logged in to read this quote.

[Speed of Sound]

Update: New Hampsire is out of the freiwal.

I defined the Freiwal as the states that always seem to have a better chance of being carried by HRC than the election according to 538. New Hampshire was among them, but now thanks to Trump leading in a couple polls, it is not

Those polls right or not, God bless NV.

[Blackacre]

Update: New Hampsire is out of the freiwal.

I defined the Freiwal as the states that always seem to have a better chance of being carried by HRC than the election according to 538. New Hampshire was among them, but now thanks to Trump leading in a couple polls, it is not

Those polls right or not, God bless NV.

Yeah, NV's early voting means Clinton is much, much more likely to win there than 538 gives her credit for, and if she wins in NV she can lose NH and still clear 270

[emailking]

Now I think about it, when Silver went on a mini-rant during a recent podcast, emphasising how big he thought Trump's chance of winning was, he sounded genuinely scared. I wonder if he sees it as part of his mission to get Democrats to take the threat seriously and go out and vote. If so, I can't help feeling there are better ways of doing it.

He's just being honest about it. He has a model that was already designed for this cycle. He's inputting the data, and running it, and interpreting the results. It's not "hedging his bets" as you put it. Trump's chances of winning have gone up based on the data, but he's still not favored, and that's all there is to it. He sounds scared because he's left-leaning and probably thinks Trump will be a disaster, even if he's trying not to show that bias, and he realizes the data shows the chance of him winning is not insignificant. That's really all there is to it.

[Erich Maria Remarque]

Yeah. 538 and Upshot are the only media I can read in USA, and about USA.

I follow both Upshot and 538 teams on Twitter and they really hate/scared of Trump, their tweets are often emotional (but it is OK, it is their private thoughts) but nevertheless are their analysis always neutral as possible and based on data.

No one can be completely neutral, but the rest of media don't even try.

It is ridiculous and Breibart'ish to think that they tweaked their models towards any of candidates . Doesn't mean, the models can't be wrong.

[elcorazon]

Yeah. 538 and Upshot are the only media I can read in USA, and about USA.

I follow both Upshot and 538 teams on Twitter and they really hate/scared of Trump, their tweets are often emotional (but it is OK, it is their private thoughts) but nevertheless are their analysis always neutral as possible and based on data.

No one can be completely neutral, but the rest of media don't even try.

It is ridiculous and Breibart'ish to think that they tweaked their models towards any of candidates . Doesn't mean, the models can't be wrong.

I agree with this. sometimes the truth hurts. Although the problem with the rest of the media isn't always that they don't try to be neutral, it's that they don't really understand the polls and they are chasing ratings most of the time. Some are biased (Fox News!!) but most are just dumb/seeking ratings.

[Blackacre]

Yeah. 538 and Upshot are the only media I can read in USA, and about USA.

I follow both Upshot and 538 teams on Twitter and they really hate/scared of Trump, their tweets are often emotional (but it is OK, it is their private thoughts) but nevertheless are their analysis always neutral as possible and based on data.

No one can be completely neutral, but the rest of media don't even try.

It is ridiculous and Breibart'ish to think that they tweaked their models towards any of candidates . Doesn't mean, the models can't be wrong.

I wouldn't follow 538's model if I didnt think it was sound and trying to best portray the state of the race. Theres a lot I dont really get/that confuses me, but I trust Nate Silver.

Having said that, I do have several problems with their models. Polls-plus really shouldn't exist. It was handy after the conventions, but now it's basically polls-only with a bit more of a pro-trump lean. I don't bother with it at the presidential level. It's a shame, because I feel like it could have been better if it accounted for other things that are out of the scope of polls-only. Like, how important is the ground game in this state compared to the nation? Is a certain state harder or easier to poll? What about early voting? Nevada is a state that uses early voting a LOT more than the nation as a whole, and given EV data so far, I completely disagree with 538's assessment of Nevada. I'd rank it 85% dem, not 50/50. The now-cast at least has a purpose, which is to extrapolate on trend lines.

I also dislike how they let one poll swing their forecast so much. Look at North Carolina. Take all the polls that have a weight of at least 2.0, and you see HRC leading in all but 2 of them, mostly by 3-4, though there's a 7 in there. Trump's up in 2: a +7 and a +1. And yet he's favoured to win the state. That's very strange, at least for me.

Still, I'm going to keep using them because I don't want to be like Dick Morris in 2012 and be blindsided just because I didnt like the data

[Erich Maria Remarque]

I also dislike how they let one poll swing their forecast so much. Look at North Carolina. Take all the polls that have a weight of at least 2.0, and you see HRC leading in all but 2 of them, mostly by 3-4, though there's a 7 in there. Trump's up in 2: a +7 and a +1. And yet he's favoured to win the state. That's very strange, at least for me.

Trends. According to their model Trump is now 3 behind Clinton compared with -7 for two weeks ago.

If a poll in a battleground state, showed Clinton +2-5 for 2 weeks ago, it likely would show from T+3 to C+1 now. If you think this is a stupid assumption and doesn't make sense, just look how their model "guessed" new more friendly polls from NH, CO etc.

Early voting? Yeah, it is a good indication, but it is not representative. Neither party ID. For example in Florida according to Upshot poll, Trump was winning about 30% of white uneducated registered Democratic man. We might see something similar in other states as well (yeah, I'm talking about NV).

So according to EV, Clinton is likely to win, but it is really difficult to bake it in into model.

[Mr. Morden]

Like, how important is the ground game in this state compared to the nation? Is a certain state harder or easier to poll? What about early voting?

How can you put ground game and early voting into the model?  There's no way to quantify "ground game" in any meaningful way, and early voting?  It hasn't been around for very long.  The model is based on how accurately various parameters have predicted election outcomes over the past X years.  You can't plug in things that have only been important for the past couple of elections.  You'll just get garbage.

But as I've said before, that's kind of missing the point here.  Yes, there are additional factors that you might think are important in making subjective assessments of the probabilities.  And you're free to use them when you make subjective assessments!  The analogy I used before was if you're watching a football game, and the commentator says "28% of the time that a game has this score, and this field position, with this much time left, the underdog wins."  Maybe there are other factors that you think are important that are harder to quantify, like the weather.  If so, you're free to use them in setting your own subjective probabilities.  You're free to mentally shift that 28% higher or lower.  It's just a baseline, telling you how other teams have done in the past under those conditions.

That's what the 538 model is.  It's just telling you "X% of the time that the underdog has been behind by this much in the polls, he's won."  You're welcome to mentally adjust that up or down depending on what other factors you think are relevant.

[Erich Maria Remarque]

You're welcome to mentally adjust that up or down depending on what other factors you think are relevant.

In my opinion it is how one should use this model. As a prior probability.

Just adjust it for EV, hispanics suuurge, angry NH women or Shy Educated Female Trumpistas

[BlueSwan]

The development over the last few days is super scary. Clinton's win percentage:

Oct 29th: 81,1%
Oct 30th: 78,8%
Oct 31st: 75,2%
Nov 1st: 71,2%
Nov 2nd: 67,7%
Today: 64,7%

That is virtually a linear nosedive, losing an average of 3,28 points PER DAY. As a thought experiment, if the trend continued, on election day the model would have Clinton at 48,3%.

[OSR stands with Israel]

The development over the last few days is super scary. Clinton's win percentage:

Oct 29th: 81,1%
Oct 30th: 78,8%
Oct 31st: 75,2%
Nov 1st: 71,2%
Nov 2nd: 67,7%
Today: 64,7%

That is virtually a linear nosedive, losing an average of 3,28 points PER DAY. As a thought experiment, if the trend continued, on election day the model would have Clinton at 48,3%.

Trump wont win PA or Colorado  unless her chance drop to 42 percent

[elcorazon]

The development over the last few days is super scary. Clinton's win percentage:

Oct 29th: 81,1%
Oct 30th: 78,8%
Oct 31st: 75,2%
Nov 1st: 71,2%
Nov 2nd: 67,7%
Today: 64,7%

That is virtually a linear nosedive, losing an average of 3,28 points PER DAY. As a thought experiment, if the trend continued, on election day the model would have Clinton at 48,3%.

Trump wont win PA or Colorado  unless her chance drop to 42 percent

Harry Enten keeps saying she won't lose NV, which, if true, would add to Hillary's overall chances given the models have NV around 50/50 depending.

[Meclazine for Israel]

The development over the last few days is super cool. Trump's win percentage:

Oct 29th: 18,9%
Oct 30th: 21,2%
Oct 31st: 24,8%
Nov 1st: 28,8%
Nov 2nd: 32,3%
Today: 35,3%

That is virtually a linear ascent, climbing an average of 3,28 points PER DAY. As a thought experiment, if the trend continued, on election day the model would have Trump at 51,7%.

[Speed of Sound]

The development over the last few days is super scary. Clinton's win percentage:

Oct 29th: 81,1%
Oct 30th: 78,8%
Oct 31st: 75,2%
Nov 1st: 71,2%
Nov 2nd: 67,7%
Today: 64,7%

That is virtually a linear nosedive, losing an average of 3,28 points PER DAY. As a thought experiment, if the trend continued, on election day the model would have Clinton at 48,3%.

Trump wont win PA or Colorado  unless her chance drop to 42 percent

Harry Enten keeps saying she won't lose NV, which, if true, would add to Hillary's overall chances given the models have NV around 50/50 depending.

And, more importantly, Harry's saying it because Jon Ralston is saying it. Jon Ralston is the one and only guy when it comes to covering NV for this.

[Slander and/or Libel]

Listening to Harry Enten talk, it sounds like he low key doesn't agree with Nate': model in several respects.

[AGA]

If you scroll down and check the snake of the polls-only forecast, Nate Silver has Nevada voting to the left of North Carolina, but with Trump winning Nevada and Clinton winning North Carolina. Maybe this is because both are tied in the model and the snake just sorts alphabetically at that point.

[Crumpets]

Clinton is up for the day in the nowcast for the first time since October 26th.

[Erich Maria Remarque]

Clinton is up for the day in the nowcast for the first time since October 26th.

Yeah. The race seems to stabilize. Clinton might even regain a bit.

But it will be one more NH polls tonight. From "B" UMassLowell. If it will be Trump-frienldy, he could gane again.



Difference between 538/Upshot over time.

[heatcharger]

She regained Florida in the polls-only model, but I have to say, the huge D-house-effects they've slapped on some of these higher quality polls are absolutely laughable:

[Slander and/or Libel]

She regained Florida in the polls-only model, but I have to say, the huge D-house-effects they've slapped on some of these higher quality polls are absolutely laughable:

It's not like it makes the hugest absolute difference, but it's really quite ridiculous. The act of aggregating the polls is enough to smooth out the differences. Instead Nate needs to say what the polls REALLY said, and then say that some polls are more equal than others, etc. Those two things combined brings the average margin in the 10 polls above from Clinton +0.7 to Trump +0.6.

I'm not really convinced at all that these methods have a ton of rigor, and especially not in combination with each other. There is a huge chance that his estimates are double counting things that are already showing up in the polls on their own.

[Erich Maria Remarque]

She regained Florida in the polls-only model, but I have to say, the huge D-house-effects they've slapped on some of these higher quality polls are absolutely laughable:

Not just house effect.
Nate said, that this also includes trend-line, but I guess it includes more, namely:

You must be logged in to read this quote.

I guess, you refer to Sienna/Marist/Selzer.
Selzer don't have any house effect IIRC, so I suppose I was about 2% adjusting for trend line.
Sienna/Marist do have. So they get house effect + trend linte ~2-3%.

The whole point with 538, is that states are correlated to each other and to nationall polling.

[riceowl]

I know the model doesn't account for early voting, but I'm starting to think Iowa could surprise and they have it as <30% for Clinton in all models.

[Gustaf]

Most of the adjustments do come from trend lines. That isn't unreasonable in itself but I think the problem is that these trend lines are largely derived from all the dumb trackers. And that's where they gain undue influence. The last few days the model has been insane, IMO. Most of the win probability adjustments have been going the wrong way.

[Erich Maria Remarque]

Most of the adjustments do come from trend lines. That isn't unreasonable in itself but I think the problem is that these trend lines are largely derived from all the dumb trackers. And that's where they gain undue influence. The last few days the model has been insane, IMO. Most of the win probability adjustments have been going the wrong way.

Lol, no. If there were no Trump's surge (temporary/artificial or not), we would not see polls from a reputable pollsters showing tie races in CO/NH/NV/FL.

Just for 2 weeks ago, Trump was trailing in most state polls by 3-7% points.