I’m skeptical of that analogy. Scores and plays in football are public information. Actual accurate forecasts of key voters are not, and are in fact unknowable. It’s more like a decision of whether to put in a backup…but you can’t be entirely sure of the quality of your starter’s play.
I understand what you’re saying, but my argument is that looking at special elections (which are always unique) and especially 2022 where the Democrats over performed but still lost seats and control of the House is like watching a team (or a player) have an unlikely run that needed a lot of luck. Eventually the more likely outcome will occur.
OK, a more accurate title would be “538 is no longer using the methodology they have used for the last 14 years, and should be thought of as though they were an entirely new site whose predictions are completely untested, rather than given the benefit of the doubt which their previous track record had earned”.
What’s really sad is that I didn’t realize the change had occurred until I went to it to try to get some predictions for the UEFA contest work was running. Which was totally the kind of content the site under Silver would have had.
I think the reason Democrats do better in special elections these days is because they’re generally low turnout, and Trump has driven most of the high-information always-vote cohort into the Democratic camp, so low voter turnout is now good for the Dems. (Unless, of course, it’s brought about by specifically discouraging Democrat-leaning voters from voting)
Yes, Silver certainly argues that statistical evidence supports his position. If you subscribe to his Substack you can read all his geeky stat articles and check the references if you are so inclined. Admittedly, I’m not a statistician and I haven’t given an equal amount of attention to the other side’s arguments, but his arguments seem sound to me.
Even the geekiest statistical arguments have to start with agreeing to some sets of assumptions. And it is very unclear right now which sets of assumptions are valid this time around. There are first ever circumstances this time …
We can say that 538 is not the model that was Fivethirtyeight, Silver’s site is that.
We can say that 538 is currently heavily discounting polls and otherwise handling polls in ways different than other modelers do.
We can say that their approach is spitting out a result very different than the other major sites do.
We can say that this is this model’s maiden voyage. It has no track record.
Including it in a discussion is valid; citing it as what the polls tell us, or even what the consensus of models tell us, would be very inaccurate.
According to wiki, G._Elliott_Morris built a model for the Economist for the 2020 Presidential election. It gave Biden an 87% chance of victory on August 1st, which Silver thought was dubious at the time. Silver later blocked Morris on twitter. So the 2 analysts have some history.
Morris makes some extraordinary calls, ones that I have doubts about, though they are not absurd. These models are straightforward to create if you have the data, though they are time consuming. Overall, I trust Silver’s background more than I do Morris’. Sam Wang is also worth reading.
Morris’ 538 is free. So there’s that. I can’t say I trust their model, though I would dial the OP back a tad.
Prediction? Sure. Forecasts are easy though.
On Nov 17, 2024 there will be a 31% chance of at least some precipitation in New York City, based on the observation that NYC has 9-10 days of at least some precipitation in November. Unreliable cite, but I hope I’ve made my point. For certain long run forecasts, you can use the historical trend. Put it another way, the odds of the Democrats winning in 2056 are roughly 50-50 (maybe a little lower given the chances of weird events happening such as viable third parties over a 32 year period).
The question is what’s the information value of today’s rain on November rain (nothing) or today’s Presidential polling on November’s outcome (nothing, according to Morris).
Yes and no?
The polling tells us quite a bit. We know the broad shape is very likely to be a close popular vote, within a few percent. The odds of a popular vote blowout either way are quite small.
The polling to date can also tell us that very little moves these numbers. Since March Biden’s been stuck at about 40% ish and Trump a couple of points more. The last week bumped Trump up a notch but not by much given that he was shot, had a powerful picture taken, and a convention.
Big events have had little impact.
It’s not just the rain today, it’s the fact it has rained about the same amount every day for multiple months. THAT becomes meaningful.
Proceeding as is can be realistically expected to continue to deliver similar numbers up to Election Day. Without a major something, bigger than the things to date, polls saying similar then is a safe bet.
Of course given the hard to parse out in polls part is always those who are deciding not only who to vote for but if to vote. And being confident about what those people will do months from now would be foolish. Even the day before!
But essentially discounting a stable set of data as non predictive seems foolish as well.
I have seen numerous tweets today saying Nate Silver is now working for a company funded by Peter Thiel. I’m not sure how direct the connection is, but it’s not a great look.
Not sure what the implication is supposed to be? Thiel’s fund has invested a minority stake in Polymarket. Nate Silver is major talent for betting markets and has been hired by Polymarket as an advisor.
Thiel does not control or have any managerial role in Polymarket.
What’s the issue in regards to how Silver’s own site is run?
I don’t pay any special heed to his pontificating but no less because he has a well paid consultant gig at a company that Thiel has invested in. Why should I?
I hate polls so much I’ve never taken one. Not even now in our texting world. I delete all local and federal poll attempts as delete and block. Am I the only one?
If we resurrect Jimmy the Greek and he says “The Detroit Lions will be president!” I’d trust him more than Mr. Algorithm-I know absolute fuck about politics- Silver.
I was sure his BS in 2016 would shut down his website. He sucks.
No. And the fact that the demographics that ended up voting Trump were more likely to do that in the past, or be harder to reach, has been blamed as the source of past systemic error.
At this level I wonder if these models can possibly be compared except in hindsight? After the election see whose prediction was best.
If Silver has a stake in bets on the election, then he’s naturally going to be using his statistical skills to make the best bets possible. And he also has an incentive to make sure that nobody else makes those same bets, to make his take as high as possible, so he has an incentive to lie to everyone else about what his statistics are saying. Is he actually following that incentive? Who knows?
So nothing to do with Thiel having invested in the company that he is a hired advisor of?
More that he may want to mislead other bettors (like someone writing about stocks might want to mislead other investors)?
Is this a question? Are you asking me if you’re sure? If I wanted to know that, which I do not, I would ask you.
Yes it is a question. @Gorsnak stated it’s not a good look. You dropped links. I’m not understanding what is being implied.