538 Model Megathread

[GeorgiaModerate]

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.

There has clearly been a Trump surge over the past two weeks.  It's starting to look like it reached a peak and has perhaps ebbed a bit.  From Nate Silver:

You must be logged in to read this quote.

[KingSweden]

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.

Predictwise has similarly shown a similar chart in relation to 538. This is why I think Silver's model is useful, it just exagerates trends in comparison to the others

[Erich Maria Remarque]

Predictwise has similarly shown a similar chart in relation to 538. This is why I think Silver's model is useful, it just exagerates trends in comparison to the others

Yeah.

I think, that this election is sooo negative-news driven, that their model is too quick. They should add news-bump prediction (as Post-Convention Bump in polls-plus), hehe

[100% pro-life no matter what]

Clinton's 99.9th Percentile:

Clinton's 99th Percentile:

Clinton's 95th Percentile:

Clinton's 90th Percentile:

Clinton's 75th Percentile:

Clinton's 67th Percentile:

The 50th Percentile:

Trump's 67th Percentile (a narrow Trump win):

Trump's 80th Percentile:

Trump's 90th Percentile:

Trump's 95th Percentile:

Trump's 99th Percentile:

Trump's 99.9th Percentile:

[Devout Centrist]

Sad News from Texas

[Wiz in Wis]

Ok... I'm annoyed by this. 1 poll from UT comes in, with Trump up 7, and the national probability of his winning goes UP 1.3%. That's a ridiculous amount of over sensitivity.

[Nym90]

Talk about polarization. Currently the polls-plus models for both President and Senate have every state with a Senate race voting for the same party for both offices. Every Clinton state voting for a Democratic Senator, and every Trump state for a Republican Senator.

[Erich Maria Remarque]

Ok... I'm annoyed by this. 1 poll from UT comes in, with Trump up 7, and the national probability of his winning goes UP 1.3%. That's a ridiculous amount of over sensitivity.

Mr. Morden explained, but had misstake about on average. I think.

I don't think you need that to explain it.  As I said, shifts or around ~0.5% in the win probability are going to happen even when you don't add any new polls at all, simply because of statistical noise.  And that's an average.  Sometimes the "phantom shifts," that cannot be explained by any particular poll, are larger than that.

[afleitch]

I see the junk polls have erased Clinton's bounce.

[Speed of Sound]

I see the junk polls have erased Clinton's bounce.

Poor, poor Nate.

[afleitch]

I see the junk polls have erased Clinton's bounce.

Poor, poor Nate.

We have less polls this year (and 2012 was less than 2008) because 'who needs polls, we have the models', which in turn leads to less polls going into the model and polls designed to impact upon the model. There needs to be something like the British Polling Council; if you're not signed up and you're not transparent, then it's not counted as a poll.

[Erich Maria Remarque]

Lol, it was a lot of deplorables polls coming during an hour




(((Harry Enten))) ‏@ForecasterEnten  41m41 minutes ago
So we got John Yob's firm (the guy who tried to takeover the Virgin Islands GOP), Trafalgar Group, and Zia Poll with new state polls today.

↓

Jon FavreauVerified account
‏@jonfavs
@ForecasterEnten are you guys concerned about how many garbage vs high quality polls are getting thrown in the model this week?

↓
 
 (((Harry Enten))) ‏@ForecasterEnten  3m3 minutes ago
@jonfavs I mean I am.

[Mr. Morden]

Ok... I'm annoyed by this. 1 poll from UT comes in, with Trump up 7, and the national probability of his winning goes UP 1.3%. That's a ridiculous amount of over sensitivity.

Mr. Morden explained, but had misstake about on average. I think.

I don't think you need that to explain it.  As I said, shifts or around ~0.5% in the win probability are going to happen even when you don't add any new polls at all, simply because of statistical noise.  And that's an average.  Sometimes the "phantom shifts," that cannot be explained by any particular poll, are larger than that.

What's wrong with what I said about 0.5% being an average?  That's around the average shift you'd expect when you run a second set of 10,000 simulations with the exact same polls.  Do you disagree?

[Erich Maria Remarque]

What's wrong with what I said about 0.5% being an average?  That's around the average shift you'd expect when you run a second set of 10,000 simulations with the exact same polls.  Do you disagree?

IDk, it sounded to me as you were talking about mean when you say "the average shift". The mean of shift should be zero.
Or were you talking about abs(mean) or variance/std?

[Erc]

What's wrong with what I said about 0.5% being an average?  That's around the average shift you'd expect when you run a second set of 10,000 simulations with the exact same polls.  Do you disagree?

IDk, it sounded to me as you were talking about mean when you say "the average shift". The mean of shift should be zero.
Or were you talking about abs(mean) or variance/std?

I assume he means root mean square, which in this case should work out to sqrt(2) * the usual standard deviation for sample proportion.  That usual standard deviation is sqrt(p * (1-p) / N), which for p=.643 in the polls-only and N = 10000, is 0.5 percentage points.  Multiplying that by the sqrt(2) factor gives 0.7 percentage points.

A bit more math if you like:

The difference between two independent runs of a normally-distributed random variable with some mean \mu and some standard deviation \sigma is itself a normally-distributed random variable with mean 0 and standard deviation \sqrt(2) * \sigma.

What this means in practice is that, at the moment, 68% of the time the (statistical error only) shift between simulation runs should be less than 0.7 percentage points, and 95% of the time it should be less than 1.4 percentage points.

If you are seeing shifts of 1.4 percentage points more than 5% of the time that means:

1) The poll updates are actually having an effect (which is why they are put into the model in the first place, of course).

and/or

2) You're cherry-picking (consciously or not) the large shifts.

[Blackacre]

I see the junk polls have erased Clinton's bounce.

Are they junk polls though?

Also, I noticed something. HRC's at 64%, but every state in her Freiwal is over 70% except for New Hampshire. Thats only because the model sees Nevada as a toss-up; so the overall numbers are deflated just because of that.

[Ronnie]

I see the junk polls have erased Clinton's bounce.

Are they junk polls though?

Yes, yes they are.

[Wiz in Wis]

I see the junk polls have erased Clinton's bounce.

Are they junk polls though?

Also, I noticed something. HRC's at 64%, but every state in her Freiwal is over 70% except for New Hampshire. Thats only because the model sees Nevada as a toss-up; so the overall numbers are deflated just because of that.

Exactly.

[Erich Maria Remarque]

What's wrong with what I said about 0.5% being an average?  That's around the average shift you'd expect when you run a second set of 10,000 simulations with the exact same polls.  Do you disagree?

IDk, it sounded to me as you were talking about mean when you say "the average shift". The mean of shift should be zero.
Or were you talking about abs(mean) or variance/std?

I assume he means root mean square, which in this case should work out to sqrt(2) * the usual standard deviation for sample proportion.

Explicitly, that's sqrt(2 * p * (1-p) / N), with N the number of trials.  With p currently .643 in the polls only and N=10000, that's an rms average shift of 0.67 percentage points per rerunning of the simulations.

Oh, so embarassing! Of course, it was from run to run. My bad.

[Blackacre]

I see the junk polls have erased Clinton's bounce.

Are they junk polls though?

Yes, yes they are.

*sigh* The thing I dislike most about 538 (aside from the Nevada sh**t) is that their pollster ratings seem to be off and they let junk polls junk up their forecast

[Thomas D]

Nate is getting salty.

---------------------------------------------------------------------
Matthew Dowd @matthewjdowd
How do sketchy states polls have more impact on "data" odds sites (like 538) in this election than actual early vote results?  Bizarre.


Nate Silver @NateSilver538
32s
Next time we'll try the tried-and-true method of making sh**t up as we go along.

[100% pro-life no matter what]

Somehow, a Clinton +18 VA poll helped Trump's chances (combined with a T+6 UT poll).  I guess they view Virginia as gone and say that that means Clinton is wasting a ton of her popular vote there? 

[elcorazon]

Somehow, a Clinton +18 VA poll helped Trump's chances (combined with a T+6 UT poll).  I guess they view Virginia as gone and say that that means Clinton is wasting a ton of her popular vote there? 

I get the logic but I don't believe the model works that way

[Mr. Morden]

Somehow, a Clinton +18 VA poll helped Trump's chances (combined with a T+6 UT poll).  I guess they view Virginia as gone and say that that means Clinton is wasting a ton of her popular vote there? 

Again, it is pointless to read anything into shifts in win probability of a few tenths of a percent, since such shifts could simply be due to random chance in the simulations rather than any real change in outcome of the model being introduced by the latest polls.

[Moderate Pennsylvanian]

Somehow, a Clinton +18 VA poll helped Trump's chances (combined with a T+6 UT poll).  I guess they view Virginia as gone and say that that means Clinton is wasting a ton of her popular vote there? 

Again, it is pointless to read anything into shifts in win probability of a few tenths of a percent, since such shifts could simply be due to random chance in the simulations rather than any real change in outcome of the model being introduced by the latest polls.

In this case, I think it was because of the trend line. The +18 Roanoke College poll was Oct. 25-28, but a more recent (Oct. 29-Nov. 1) Roanoke College poll had already been posted that had Clinton +7.