2016 Postmortem
Related: About this forumNate silver is rage tweeting right now because people are calling
Him out. 😂 Check it out!
https://mobile.twitter.com/natesilver538
Response to SDANation (Original post)
NightWatcher This message was self-deleted by its author.
RBInMaine
(13,570 posts)Ms. Toad
(34,062 posts)"So unbelievably lazy."
That phrase screams "Trump" to me.
SusanCalvin
(6,592 posts)sharp_stick
(14,400 posts)because so many people don't understand the model. Lots of people just want him to cheerlead and don't stop bugging him no matter how often he explains the model.
Cosmocat
(14,563 posts)I agree with most here that he has his overall model set up to be really twitchy and fairly conservative (not the insane political bullshit way, just very safe in terms of not making a definitive call).
But, I can see what he is saying about some of the state polling - the last run of PA polls have in fact shown a clear tightening, as an example.
MyNameIsKhan
(2,205 posts)Nate is defending his model but probably we need more transparency from these aggregators. HuffPost released a detailed look into their formula, Nate must do that now.
Otherwise all traffic to his site will be reversed to right wing voters.
Chichiri
(4,667 posts)...in that he honestly had level-headed confidence in his model. But that raging tweetstorm was really . . . emotional. Hot-headed. Not like the Nate Silver I've known in the past. And that bit about "not empirically defensible" -- okay, maybe you can say that about HuffPo, but PEC? Give me a break.
Either Nate's having a rough day today (which, three days before the election, certainly a lot of us are), or he at least suspects his model is fucked up.
BlueStreak
(8,377 posts)1) Enormous amount of early voting, much of which implies GOP crossover to Hillary.
2) Sheer volume of polls, many of which are sketchy at best.
3) Polls that might have been well done in the past seem to have declined in validity, yet 538 rewards past performance.
4) Exacerbation of traditional polling errors (not getting mobile numbers) plus a huge factor in polls underestimating first-time Latino vote.
I think he had a good idea in the original model -- with the key principle being to weight new polls based on the quality of that pollster in the past. It is entirely possible that for the above reasons, Silver's old model no longer tells us very much more than the simple RCP average of polls. Silver is committed to (locked to) his old method, and is current feistiness may be a fear that his vaunted reputation may be about to take a major dive.
doc03
(35,325 posts)I have always thought there will be a significant number of Democrats that will go for Trump, I know a bunch myself.
I also think there may be a lot of people that have sat on the sidelines for years and never voted, I know some of them.
I wouldn't bet a dollar on either of them, I think it is a toss up, this is not a normal election by no means.
duffyduff
(3,251 posts)There is far more likely a huge crossover vote for Clinton than there would ever be for Trump.
You are just not understanding what this election is or the significance of it.
You are basing your view on the ludicrous idea that the Republicans are going to vote straight party ticket this year when this will NOT be the case.
It all hinges on women.
Not so much minorities, but women.
Moonwalk
(2,322 posts)...everyone will believe him and not question him. But the fact is, being a public "predictor" means you're inviting people to question you. It's part of the job. He can say "Nah,nah, nah, I was right" afterwards. But for right now, he's got to say "Wait and see...." not lose his cool. If he can't deal with people doubting him, then he's in the wrong business.
manicraven
(901 posts)In fact, his credibility is on the line, especially if he's wrong again. When asked about his complete blunder during the primaries, he said he ignored his own data because he couldn't believe it. So, what is he doing now?
andym
(5,443 posts)Here are his assumptions:
Assumption No. 1: The high number of undecided and third-party voters indicates greater uncertainty.
Assumption No. 2: The FiveThirtyEight model is calibrated based on general elections since 1972.
Assumption No. 3: The FiveThirtyEight model uses a t-distribution with fat tails, which gives a greater likelihood of rare events.
Assumption No. 4: State outcomes are highly correlated with one another, so polling errors in one state are likely to be replicated in other, similar states.
He uses polling data going back to 1972-- not sure that is a great idea, since the voting population has changed so much. But he claims that is useful.
Linking similar states is useful, but tricky on how states are grouped.
The good news from all of this is that Hillary is likely going to win, but it could be close, so GOTV and we CAN make a difference.
duffyduff
(3,251 posts)This election is a total anomaly, with a third-party-type candidate running on a major party ticket that most of the movers and shakers of the party don't support, while on the other hand you have a candidate posed to make monumental history as the first woman president.
People still think all Republicans are going to vote straight party ticket, when in fact this year is not the case at all.
manicraven
(901 posts)Obviously, that could skew things.
The Great Escape
(1,235 posts)I don't remember him being this thin skinned.
Buzz Clik
(38,437 posts)Dem2
(8,168 posts)I think giving him crap in posts here is fine, but I wouldn't treat him poorly if he showed up here for a discussion.
manicraven
(901 posts)Dem2
(8,168 posts)So I agree, his model is weird and seems overly conservative.
andym
(5,443 posts)I suspect that the model is automated, with historical and dynamic correlation calculations baked in, but I'm not sure. Does anyone know?
Buzz Clik
(38,437 posts)andym
(5,443 posts)It said he is making "trend line" adjustments-- which true is a kind of "unskewing," not necessarily the same kind used by Romney's statistician. But if Silver's unskewing system has worked in the past to give him the high accuracy he has enjoyed, than what is wrong with it now.
SDANation
(419 posts)Then gets mad when some people call him out? Sam wang had better accuracy then he did in 2012?
hlthe2b
(102,225 posts)be feeling right now--no matter how assured they might think that are that HRC will prevail... The alternative is just plain unthinkable.
Nate obviously wants to defend his work, but I think, at heart, he's every bit as nervous as all of us. And, yes, I am aware that he was paid by wealthy conservative donors to speak early in the season, but I do not read into that anything more than I do Bill or Hillary's paid speeches.
That said, I surely hope the competing prediction models that show Clinton faar more likely to win and win decisively, are correct.
LenaBaby61
(6,974 posts)And during the 2008/2012 General election cycles, Nate was golden--kept me calm, cool and collected. But this particular election cycle, something about him has been off to me. Yes, we know that there are reasons voting patterns have been all over the place which has cause volatility to voting this year IE: Under counted Hispanic vote, under counted black vote in some areas, the women's vote coming out big, and I don't think Nate is that much into including early voting into his models which is his choice. But to me, and I'm only speaking for me, I don't get the sense that Nate's as "independent" as he used to be when he was @ the NY Times. He's also been extremely defensive as people continue to call him out because he was wrong about how far Trump would go in the GOP primaries, let alone where that clown Trump is now, and that's running for POTUS. IIRC, Nate called Trump "Sideshow Don" during the primaries.
Anyways, I've been over at Sam Wang's most of this election cycle staying calm and mostly stress-free. Sam's been very bullish on Hillary all year and Nate's been bearish on her.
bushisanidiot
(8,064 posts)And vote. We can be lulled into complacency by too many positive predictions.
Adrahil
(13,340 posts)It's an empirical model. You can certainly disagree with elements of the model, but Silver is not sitting at a keyboard deciding how much up and down to make the odds go for every update. And please remember folks that he STILL is predicting an HRC victory. If his model was off this cycle, then he should do what any empiricist does... try to determine why and seek to fix it.
Buzz Clik
(38,437 posts)Give this a look: http://m.huffpost.com/us/entry/us_581e1c33e4b0d9ce6fbc6f7f?
Silver adjusted all HRC numbers down and all Trump numbers up for no rational reason -- not at all supported by polling data. He then puts out an article that says, "Oh no! Hillary's support is collapsing" when nobody else is seeing that. Even his data doesn't support it.
Adrahil
(13,340 posts)That article is in complete and not accurate.
But in a few days, we'll a result. We'll see how he did. Personally, I think he's gonna be off... HRC will outperform his projections. But I do not suspect him of any malice.
Sam Wang considers an HRC WAY more likely, but his projected margin is pretty damned close to Silvers...
Buzz Clik
(38,437 posts)tymorial
(3,433 posts)In my opinion it is a product of catered news feeds and selective information gathering. The internet allows people to choose what they want to read. Internet tracking and profiling deliver content designed to match a users browsing habits. The way we learn about current events and politics today has drastically changed and as a result our reactions have changed. The main reason why the news media had an unbiased mandate was because they needed to serve everyone in order to stay in business. A balanced report was less likely to offend. Today a balanced story is viewed as conservative by liberals and liberal by conservatives.
The one thing both sides have in common, both want to exist in an echo chamber.
Adrahil
(13,340 posts)We4 saw it during the primaries even worse.
Buzz Clik
(38,437 posts)You can model or you can guess. If you put a guess in your model, it degrades the product.
This criticism has it dead on: http://m.huffpost.com/us/entry/us_581e1c33e4b0d9ce6fbc6f7f?
And it agrees with my similar but (but less in depth) analysis. http://www.democraticunderground.com/12512588713
Everything in my experience tells me that when someone reacts like Silver -- so irrationally and so strongly for no apparent reason -- the raging lunatic is hiding something.
Garrett78
(10,721 posts)...while also thinking a Clinton landslide is more likely than some of the other forecasters do: http://fivethirtyeight.com/features/election-update-why-our-model-is-more-bullish-than-others-on-trump/
FWIW
geek tragedy
(68,868 posts)If Clinton wins 322+ EVs by 5+% then all of a sudden the other aggregators have taken his place as the ones who "got it right."
And the early vote in NV and FL are trending in that direction.
Blue Idaho
(5,048 posts)budkin
(6,699 posts)He has to be praying for a Trump win at this point just so he can tell everyone to bow down.
Capt. Obvious
(9,002 posts)Donald winning would make him look bad
budkin
(6,699 posts)AgadorSparticus
(7,963 posts)He is not the only horse in the race. People are just asking the right questions: why so different? His response is arrogant: we rely on emperical data.....as if the others don't? That is why he is getting grief and now he is rage tweeting? Really??!?!! He has lost his mind......
manicraven
(901 posts)So, Nate was wrong in the primaries by a LOT regarding Trump; in fact, off by a mile, and now he appears to be the outlier against other reputable pollsters, so he's stressed and cranky.
SusanCalvin
(6,592 posts)His method works or it doesn't.
kurt_cagle
(534 posts)His comment that the Cubs had less of a chance of winning than Trump does was a major embarrassment, and I think he's afraid that the Trump effect is something he hasn't accounted enough for yet. He also is not a fan of Hillary Clinton. I think both are coloring his reasoning, making him second guess the data. As a statistician, that's danger territory.
Buzz Clik
(38,437 posts)Indeed.
treestar
(82,383 posts)Cause a lot of DU drama! Some people seem to let it get to them.