Is it just me, or FiveThirtyEight changing?

[diptheriadan]

 Discuss

[LLR]

In what ways? I mean, I don't think so, but maybe you're thinking of something I'm not.

[Free Bird]

I think Nate's been panicking since he botched the GOP primary. He doesn't understand what to do during a paradigm shift because his method is fixated on just the numbers.

[Lyin' Steve]

Yes, after 2012 Nate stopped being just the guy at the New York Times who had all the really good models and became "Nate Silver, the world's greatest statistician."  For some reason, instead of doing what I would have done which is pick some entirely new thing to model (fantasy football?) he hired a bunch of snarky Brooklynites in their 20s who were nowhere near as good at statistics as him but could certainly pump out articles, and built a site based around just shoveling content.  At first all the content was stuff like "our expectations for each of the MLB teams this year" which was ok although there was a lot of filler and bullsh**t along with the real numbers, but then it started to devolve more and more into things like "how many carrots does America eat each year?  I write 1500 words about carrots and shove a half-baked model in at the end to answer the question."  They started following the BuzzFeed model of using clickbait titles and focusing on millenials with sex, drugs, feminism, vulgarity and celebrity/pop culture instead of things that were actually interesting to model.

Then the election coverage happened.  But dozens of other sites had polling aggregators as well.  They adjusted the model to weight by "how good we think the poll is", but nobody really cared.  Then they said, we have to do something really novel that the other sites aren't offering.  So they came up with their "endorsement primary" model.  Every damn article they wrote about the election linked to that model, and they claimed that it would be more accurate than a polling aggregator.  Of course, it wasn't at all, but they embarrassed themselves by sticking by it to the end, eventually claiming "lol it was only an experiment anyway."  In the meantime they just filled the void with a daily slew of 3-4 articles with barely any actual statistics but a lot of #analysis about how Trump had no chance and Cruz couldn't compete because his disfavorables were too high and Bernie was the oldest candidate to ever run before "based on our statistics of looking on wikipedia at the ages of previous candidates" and junk like that.  They basically made fools of themselves because they were completely wrong and they were writing based on their own opinions and feelings, which were the opinions and feelings of a bunch of Brooklyn millenials, rather than reality and hard numbers, which is what people had come to the site for.  It was a big turn-around from 2012 when everyone else was making stuff up and giving #analysis about how Romney could still win and Silver just said "nope, the numbers say Romney has no chance, that's reality."

Now we've moved on to the general election and they've repeated their mistake of adding some phony baloney "Polls Plus" model, for the sake of novelty and attention, that they claim is going to be better than a standard poll aggregator and that they link to and talk about in every single article.  They're still writing opinion pieces with the bare minimum of statistics, but now instead of repeatedly claiming that so-and-so has no shot or other such bold claims, most of their articles are basically "Trump still has a chance", "Hillary shouldn't be too confident", things like that.  And as usual they have some articles like "I looked at some polls and saw that Trump was down with young voters.  Here's 1500 words of #analysis and a chart about it."

It's boring, you can't trust it, 90% of what's written on there is only barely based on hard numbers if at all, Nate himself hardly ever writes anything, and it's strayed really far from what Nate had in 2012.  Their election forecast, without the Polls-Plus nonsense, is actually good, if Nate just made that the site and fired all the millenials and wrote an article every week about how the model has changed and what might change in the future, that would be so much better.

[IceSpear]

^Nate was good in 2012, but he was actually a lot better in 2008. The decline has been going on for a while now. But yeah, the climax was the new site.

[ilikeverin]

Yep yep.  Nate was fantastic in 2008.  2012 was a part of the decline, although it was a rebound off of some of what he was doing in 2011.  2016 is just sad.  I still reliably check it, because I find its poll aggregation some of the most user-friendly, but, sigh.  Harry Enten should be thrown into the Hudson with concrete shoes.

[Ebsy]

Enten is interesting in that he is much better on twitter than on 538 itself.

[heatcharger]

Enten is interesting in that he is much better on twitter than on 538 itself.

Enten was remarkably terrible on both Twitter and 538 during the primary season. He's good on the podcast though.

The best thing that could have happened to the project would have been to transform it into some kind of non-profit, or maybe to fund it as a fellowship through a prominent statistics or a political science department at some university.

This is actually a good idea, but it doesn't surprise me Silver took an opportunity to cash in on his fame and success. As for the site itself some of the content is mediocre but it's sort of the first in the frontier of data journalism, so it would be expected that it would have a rough start. It's much better now than it was at launch.

[Antonio the Sixth]

I don't think it's quite as bad as Averroes and Lying Steve make it out to be, but I agree with the sentiment of both posts.

[Orser67]

I actually think it was pretty bad when he first moved to ESPN but has gotten better. Maybe it's because I know which writers tend to suck (Walt Hickey) and I find the presidential coverage more interesting.

[Tartarus Sauce]

I think all their stuff makes for superb reading. You're all just judgemental snobs.

[afleitch]

I go with Sam Wang at the Princeton Election Consortium.

[Alcon]

I follow it pretty closely, because I do appreciate the basic methodological things they're doing, and they're still worth doing.  They do spend quite a lot of time and work on things like adjusting polls for age, house effect, trends, etc., which I appreciate, because it's a lot better than eyeballing the latest batch of polls.  

I do take their more editorial analysis with a grain of salt, and unfortunately it seems like they often write to their hopes and/or audience.  Their analysis of social issues is often especially guilty of this.  It's sometimes basically advocacy journalism with a lot of numbers, which has a place, but it's not why I've liked the site in the past.  I do actually think that their analysis of the General Election gets too much crap -- they overestimated the application of history to this year's primary, but the General Election probably is a lot more predictable.  Still, more of a grain of salt there than in the past, too.

I do think it's funny that Harry Enten once said he doesn't vote, because he's worried it will cause him to get too emotionally attached to given outcomes, and yet he's clearly very, very emotionally attached to an outcome anyway.  Smart guy, funny, still often enjoy his stuff, but that's pretty damn absurd.

[Vega]

I follow it pretty closely, because I do appreciate the basic methodological things they're doing, and they're still worth doing.  They do spend quite a lot of time and work on things like adjusting polls for age, house effect, trends, etc., which I appreciate, because it's a lot better than eyeballing the latest batch of polls. 

I do take their more editorial analysis with a grain of salt, and unfortunately it seems like they often write to their hopes and/or audience.  Their analysis of social issues is often especially guilty of this.  It's sometimes basically advocacy journalism with a lot of numbers, which has a place, but it's not why I've liked the site in the past.  I do actually think that their analysis of the General Election gets to much crap -- they overestimated the application of history to this year's primary, but the General Election probably is a lot more predictable.  Still, more of a grain of salt there than in the past, too.

I do think it's funny that Harry Enten once said he doesn't vote, because he's worried it will cause him to get too emotionally attached to given outcomes, and yet he's clearly very, very emotionally attached to an outcome anyway.  Smart guy, funny, still often enjoy his stuff, but that's pretty damn absurd.

I think that has more to do with getting sick/burnt out  of politics, which, given his profession, isn't an option

[Alcon]

I follow it pretty closely, because I do appreciate the basic methodological things they're doing, and they're still worth doing.  They do spend quite a lot of time and work on things like adjusting polls for age, house effect, trends, etc., which I appreciate, because it's a lot better than eyeballing the latest batch of polls. 

I do take their more editorial analysis with a grain of salt, and unfortunately it seems like they often write to their hopes and/or audience.  Their analysis of social issues is often especially guilty of this.  It's sometimes basically advocacy journalism with a lot of numbers, which has a place, but it's not why I've liked the site in the past.  I do actually think that their analysis of the General Election gets to much crap -- they overestimated the application of history to this year's primary, but the General Election probably is a lot more predictable.  Still, more of a grain of salt there than in the past, too.

I do think it's funny that Harry Enten once said he doesn't vote, because he's worried it will cause him to get too emotionally attached to given outcomes, and yet he's clearly very, very emotionally attached to an outcome anyway.  Smart guy, funny, still often enjoy his stuff, but that's pretty damn absurd.

I think that has more to do with getting sick/burnt out  of politics, which, given his profession, isn't an option

Maybe, but his explicit explanation was the one I gave (superfluous comma was confusing).  He says that he's voted in some local elections, but plans to avoid the presidential for this reason.  It's a little odd, since he's very open about being extremely disturbed by the idea of Trump winning.

[Negusa Nagast 🚀]

I do actually think that their analysis of the General Election gets to much crap -- they overestimated the application of history to this year's primary, but the General Election probably is a lot more predictable.  Still, more of a grain of salt there than in the past, too.

Agreed. Silver and co bought in heavily to the Party Decides theory of primaries, despite the small sample size of competitive primaries in the modern era and said primaries having data points that the party establishment does not always decide.

Their coverage of the election has drastically improved since Silver's mea culpa on Trump's primary win.