Nate Silver is the founding, or one of the original founders, of 538 - which has, I see, been acquired by ABC News. He's a statistician of some repute, and competent; he's pretty good at explaining and summarizing stats to non-math types. He's gotten a lot of heat for forecasting results that were wrong, starting with the Clinton/Trump election in 2016, but if much of that is because people are terrible at understanding probability, and if you tell the average person something has an 80% chance of happening, they take it at a guarantee.
The issue is that 538 was pretty reliable at the start, but then something happened with polling and things started to go against predictions frequently enough that they sort of became just another voice - instead of a source you might use to bet on. The ABC acquisition probably didn't help.
Anyway, I digress. Nate was the statistician face behind 538. I don't know what he's doing now.
Has it corrected? He mentions in TFA that polling started to get really wrong as early voting for more popular, and wasn't properly accounted for. I admit, I was one of the ones that lost faith in 538 because of the repeated dissonance. They knew something was off, but seemed unable to adjust their models to account for it. I still appreciate his analysis, but I'd love it if he's addressed the issues in the modeling.
Remember, Nate Silver predicted that Hillary Clinton would win in 2016, and when Trump won instead, it was chalked up to the fact that it really was a random chance.
Don't panic about this. Keep quiet and keep doing the work to get Trump thrown out. And charge your mental health bills to the Democrats, for putting up an old man up for election in 2020, one who's even older than Trump, in the first place.
Silver's model will only actually matter once voting starts. Until then he may as well be a poll aggregator. Which, if the polls are flawed, then his aggregation and model will be flawed.
I am legitimately scared for my safety with the upcoming election. I'm trans and if Trump tries to take my HRT away I will end my life. It would be the final straw so to speak. I will not be forced to live a lie.
could try and get diy hrt from reputable tested vendors. if dark web marketplaces can use the mail to ship real illegal drugs, surely some estradiol will be fine.
Party strategists always say their party is going to do well. It’s part of their job. I don’t think this is particularly meaningful, unless you think there’s some particular methodology he has access to that’s better than Silver’s.
In one single election, yes. It means nothing, especially when you understand that his job is not to generate an accurate prediction, it’s to energize core supporters into donating to the campaign.
By the way, you can make the same argument in reverse—Trump always overperforms his polling right? If that prediction is accurate then Biden is absolutely going to get trounced. Now I don’t necessarily think this is correct, but it’s a slightly more sophisticated version of the fallacy you are falling prey to here.
This is a perfectly succinct, textbook example of Outcome Bias.
Betting $1 with a 1 in 3 chance to win $2 is objectively a bad idea; the odds are against you. It doesn't stop being a bad idea if you win the $2 after 1 bet.
Got a text the other day to demand my support for Biden by completing a poll via some suspicious shortened link. Might’ve been legit, more likely a phishing attempt. The wording of it just made me think of the “Trade offer” meme.
Didn’t respond. If this is how pollsters operate they’re gonna be out of business within a decade (should be already) or just continue to get skewed results from braindead fools who click on suspicious links and also vote for blatantly unfit, deranged and dangerous candidates.
I've gotten two texts from two different numbers claiming I'm not registered to vote. Which is weird, because I voted in a primary a couple months back. So I checked my state's voter registration and I'm still there, still getting a mail-in ballot like I asked.
I did a bit of forensics on the links but they just redirect to a GCE instance that returns a 500 error, and the domain registration is anonymized so I can't get any info there. But I'm worried a lot of people are clicking a link that might take them off the voting rolls.
Exactly so. They were always more performative than predictive (remember all those things polls got wrong? No? Funny, that), but in 2024 they’re absolutely reaching and pretending like everything's normal. Trust them, bro.
I think there is a lot this fails to capture because certain things are unprecedented. Michigan's GOP is in utter disarray and it isn't the only one. And overturning Roe v Wade has energized the left and disillusioned whatever center remains.
Now these facts are baked into the polling already, so obviously that's a big concern, but I believe this means polling is too far right across the board. I think who makes up likely voters has shifted. RvW drew in younger voters and I think now that they are engaged they will remain so.
Time will tell. I've seen far less Trump support this year than I did in 2020, which yeah is anecdotal, but I think it's an indicator. Of course, even if I'm correct, Michigan isn't going to carry the election alone, and it looks like the rest of the rust belt is further to the right.
It's all statistics. It means that if we ran the 2024 election millions of times in his model, Trump would win more than Biden. But we will only get one shot, so the number is kind of useless.
I was watching the Mets game this weekend on ESPN, and they were ahead of the Cubs by a few runs. ESPN has a tracker that estimates "Win Probability" and their model gave the Mets a 75% chance to win. But have you seen the Mets this year? They've blown a bunch of games late. Every Mets fan watching knew that their bullpen wasn't good enough to merit that rating.
The Mets did end up winning that game. (Thanks, Grimace.) But that doesn't change the fact that no matter what math is behind their win prediction model, it just doesn't feel right to apply statistics like that to one-off events.
Don't worry, little Yankee fan. It's not your fault you ran headfirst into the Grimace Effect. Now come closer -- I need to bottle up some of those tears for comfort when the wheels fall off the wagon after the ASG, as is tradition.
That's all the media can do nowadays. It's a bunch of journalism graduates twiddling their fingers while cranking out endless "Read the Tea Leaves!" type articles. Everything nowadays is "survey says this", "polls say that", "model says this", "odds predict this or that". It's literally everywhere from sports to politics to the stock market, it requires zero thought or in-depth analysis, and it's both a response to and a cause of the decline in mainstream and investigative journalism. It's team-based tribalism through and through.
Trump does still lead in our national average — however narrowly. But the bigger problem for Biden though is that elections in the United States aren’t determined by the popular vote.
That's a problem for all of us. If the president were elected by popular vote, Trump would never have been president.
Neither would George W. Bush. Republicans have won the popular vote only once in the last 32 years, and that was Bush as the incumbent in 2004 - which wouldn't have happened had Gore been the incumbent.
This is a center-left country, with an election system that gives extreme right-wingers oversized influence.
natesilver.net
Hot