538 Model Megathread

[The Other Castro]

Now Jon Lovett and Dave Wasserman get into it.

Jon LovettVerified account
‏@jonlovett
I criticize forecasts all the time, but hard to come up with a worse way to criticize the Silver model than the argument in this piece.

Dave Wasserman ‏@Redistrict  43s43 seconds ago
Dave Wasserman Retweeted Nate Silver
Remember when same thinking led HuffPo to decide in 7/15 to cover Trump in Entertainment section, not Politics?

Ryan Grim is probably feeling some regret now.

[‼realJohnEwards‼]

Now Jon Lovett and Dave Wasserman get into it.

Jon LovettVerified account
‏@jonlovett
I criticize forecasts all the time, but hard to come up with a worse way to criticize the Silver model than the argument in this piece.

Dave Wasserman ‏@Redistrict  43s43 seconds ago
Dave Wasserman Retweeted Nate Silver
Remember when same thinking led HuffPo to decide in 7/15 to cover Trump in Entertainment section, not Politics?

Ryan Grim is probably feeling some regret now.

LOL @ Wasserman being a hypocrite. I respect the guy, but he has to admit that 538 did the same thing back then.

[The Other Castro]

Ryan Grim ‏@ryangrim  20m20 minutes ago
@NateSilver538 the piece links to your public reasoning. Happy to add any further explanation you want in there

Nate Silver ‏@NateSilver538  4m4 minutes ago
@ryangrim: The article made clear you have **no f**ing idea** what you're talking about. That's why you contract people **ahead of time**.

Ok chill there a bit maybe.

[SirMuxALot]

Summation of that HuffPo article:

"Nate Silver stupidly believes that *all the polls*, repeat ALLLLLLL THE POLLS have some non-zero chance of being off by as much as three points.  What an idiot unskewer."




"Oh by the way, the polls were all off by about three points in Romney's favor in 2012.  But what an idiot, this Silver guy is, huh?"

[Antonio the Sixth]

I hate to say this because I'm not liking 538's results lately, but Nate is obviously right here and Grimm is being a pretentious dick.

Guess what? 538's model has always been adjusting for trendline, including in 2008 and 2012 when everybody was in awe at its greatness. You are entitled to believe that it's a faulty method, but cries of "UNSKEWING!!!!!!!!" make no sense.

As has been said, the real problem with 538 isn't the trendline adjustment, it's that it pays too much attention to crappy pollsters.

[Lief 🗽]

Sad to see Nate Silver appear to suffer a mental breakdown as his model comes closer and closer to being exposed as a massive fraud.

[Mallow]

It's not his fault there have been so many garbage polls released recently -- but his model is supposed to weigh them accordingly based on their rating. That hasn't happened the way it's supposed to because the sheer volume of junk polls is flooding the model in a way that didn't happen in 2012.

At least in pollster.com you can modify which polls are included.

I hate to say this because I'm not liking 538's results lately, but Nate is obviously right here and Grimm is being a pretentious dick.

Guess what? 538's model has always been adjusting for trendline, including in 2008 and 2012 when everybody was in awe at its greatness. You are entitled to believe that it's a faulty method, but cries of "UNSKEWING!!!!!!!!" make no sense.

As has been said, the real problem with 538 isn't the trendline adjustment, it's that it pays too much attention to crappy pollsters.

Yep.

[Erich Maria Remarque]

I hate to say this because I'm not liking 538's results lately, but Nate is obviously right here and Grimm is being a pretentious dick.

Guess what? 538's model has always been adjusting for trendline, including in 2008 and 2012 when everybody was in awe at its greatness. You are entitled to believe that it's a faulty method, but cries of "UNSKEWING!!!!!!!!" make no sense.

In fact, it is the main future of 538.

As has been said, the real problem with 538 isn't the trendline adjustment, it's that it pays too much attention to crappy pollsters.

It is not completely true.

a) there is not THAT many junky polls.
b) you assume that all/mostly of them are Trump-friendly. It is not true (except last two days, when a lot of (R) polling firms were out). But if you look back, I would say it was more (D) pollsters thatn (R) (again, except last two days).

The main difference of chances 538 vs other is coming of trend line adjustemnts. If you look an nationall polls. Upshot has Clinton +2.4% vs 538 Clinton +2.8% (or +3 in PV projections). Do you think that junky polls caused that?

[Speed of Sound]

Is this supposed to be some kind of humble brag? Good grief.

You must be logged in to read this quote.

[Devout Centrist]

Is this supposed to be some kind of humble brag? Good grief.

You must be logged in to read this quote.

Nate's losing it.

[Erich Maria Remarque]

Is this supposed to be some kind of humble brag? Good grief.

You must be logged in to read this quote.

WTF?

I think that Trump will get 1 extra vote in NY though

[Thomas D]

Nate is under a lot of pressure.  He went 51 for 51 four years ago, and now his model is giving him numbers he know don't add up, and he's handcuffed.

[‼realJohnEwards‼]

Is this supposed to be some kind of humble brag? Good grief.

You must be logged in to read this quote.

Nate's losing it.

[Devout Centrist]

Nate is under a lot of pressure.  He went 51 for 51 four years ago, and now his model is giving him numbers he know don't add up, and he's handcuffed.

Someone explain to me why A rate pollsters are 'adjusted' more than Remington or Survey Monkey? Even ARG and Zubat Zogby get adjusted less.

[Antonio the Sixth]

Is this supposed to be some kind of humble brag? Good grief.

You must be logged in to read this quote.

Nate's losing it.

I actually LOLed at this.

[Erich Maria Remarque]

Nate is under a lot of pressure.  He went 51 for 51 four years ago, and now his model is giving him numbers he know don't add up, and he's handcuffed.

Someone explain to me why A rate pollsters are 'adjusted' more than Remington or Survey Monkey? Even ARG and Zubat Zogby get adjusted less.

TREND LINE and house effect. Was "A" pollsters "old" ones?

[Obama-Biden Democrat]

Sad to see Nate Silver appear to suffer a mental breakdown as his model comes closer and closer to being exposed as a massive fraud.

Nate will have to go around begging for a new contract after being fired by ESPN. Not Good!

 

[Blackacre]

Nate is under a lot of pressure.  He went 51 for 51 four years ago, and now his model is giving him numbers he know don't add up, and he's handcuffed.

Yeah, that clean sweep in 2012 gives him very high expectations and he knows he can't do the same this time.

Even if his model was perfect and had no problems whatsoever, (which I don't believe but let's go with it) this time there are just too many states with probabilities in the 50s or 60s or 70s. In 2012 only 4 states had probabilities below 85%: Colorado, Florida, Virginia, and North Carolina. And only Florida was below 74%.

This time? 8 states, plus ME-2 and NE-2, have probabilities below 74%. A whopping 17 states (plus ME-2 and NE-2) are below 85%. Nate knows that it's far more likely than not that one or more of the states in his far-less-confident model won't go the way it "predicts" it will, and that'll ruin his record. The more it happens, the worse it gets.

PLUS there's the fact that his model uses junk polls and the simple fact that if it shows Trump as the favourite in Nevada, that's one state everybody knows won't go the way the model "predicts" it will

[JohnCA246]

My main criticism would be that his model seems to heavily weigh polls that are a week or two old. Personally, I am more concerned with the last 3 or 4 polls than a high quality poll done two weeks ago. His highest weighed Florida poll was conducted between Oct 20-24th. I have no problem with hedging, if anything more of that needs to be done because polling errors or last minute swings happen.

P.S. what are everyone else's thoughts on this, as mine aren't really supported by history, just that it seems we have experienced polling shifts in the last few weeks that old polls wouldn't pick up on.

[rafta_rafta]

To be fair, that Huffo piece was garbage.  Nate silver should be judged on the accuracy of his model post election. That complaining about unskewing was bogus. Without adjustment and weighting, all models will basically look like the rcp average

[Devout Centrist]

To be fair, that Huffo piece was garbage.  Nate silver should be judged on the accuracy of his model post election. That complaining about unskewing was bogus. Without adjustment and weighting, all models will basically look like the rcp average

But his adjustments, unlike his weighting, seems nonsensical.

[QuickHead555]

Nate's Model was really accurate in 2012, but as this is a really unpredictable election, I honestly don't blame him if he isn't 100% accurate this time. I do feel that he's a bit off, Nevada is almost definitely going to Clinton, but I can see many old white men voting for Trump in Florida.

[IceSpear]

Very glad the discredited fraud's model now has Trump winning NV.

[PresidentSamTilden]

You guys really think HuffPo's model is better than 538? Idk man, I'm with Nate on this one. Even though I think his model will a miss a couple of states this year.

[Angel of Death]

538 addresses the elephant in the casino.