Biden is still more likely to win, according to the FiveThirtyEight forecast

[TheReckoning]

https://projects.fivethirtyeight.com/2024-election-forecast/#:~:text=538%20uses%20polling%2C%20economic%20and,to%20explore%20likely%20election%20outcomes.&text=in%20our%20simulations%20of%20the,of%20no%20Electoral%20College%20winner.

Thoughts?

[7,052,770]

It now comes down to whether people who claim that Trump is an existential threat to our American system really believe it, or were just grandstanding when they said it.

There may be enough of that sentiment that any Democratic nominee will ultimately win. I fear that's Biden's only possible path to winning. I don't think he's going to win any swing voters on his own merits. Sad situation all around.

[E-Dawg]

And Nate Silver, the actual man behind FiveThirtyEight in the first place (and who has a relatively good prediction record), has his own model which estimated pre-debate that Trump has a 66% chance of winning. And his model is very similar to the FiveThirtyEight models used for those past elections. And I'm sure that when the post-debate polls are released, Trump will take the lead in the FiveThirtyEight model, and will expand his commanding lead in Nate Silver's model even further.

[iceman]

their election model must be having some cognitive decline as well.

[ibagli]

https://twitter.com/gelliottmorris/status/1806501909139996835

[○∙◄☻¥tπ[╪AV┼cVê└]

I'm pretty sure Trump would have been already ahead before the debate for a polls only model.

[Annatar]

The 538 model weights fundamentals heavily, so it projects Biden will win the NPV by 2.1% despite currently being behind by 1.3% in the 538 polling average. The model thinks there will be a 3.4% shift towards Biden due to good fundamentals such as good economic indicators.

[Sir Mohamed]

There have just been few polls since the debate debacle. Unless there are, it's hard to measure what exact impact Biden's performance has in such a very numbers based model.

We have to wait to the end of next week and see whether there's movement. I suspect only barely, because opinions at this stage are backed in and Trump just did little to make actual gains among undecided voters (of which there are just very few left anyways).

[President Punxsutawney Phil]

Reasonable.

[Steve from Lambeth]

Also, why are you using the "link to highlighted text" feature? It doesn't work on Firefox and, in case you missed it, every non-stub Wikipedia article is broken down into sections that you can easily hyperlink to.

And this applies doubly when you're literally trying to forward us to the subtitle at the header of 538's most important page this year.

[BlueSwan]

538 is useless after Silver left. Really sad, because I used to love that website.

[Wisconsin+17]

https://projects.fivethirtyeight.com/2024-election-forecast/#:~:text=538%20uses%20polling%2C%20economic%20and,to%20explore%20likely%20election%20outcomes.&text=in%20our%20simulations%20of%20the,of%20no%20Electoral%20College%20winner.

Thoughts?

Seems to me like statistical noise. He wins 506 to Trump winning 490. 1000 is +/- 3.16%. Why don't they run a billion simulations not 1000? To distinguish a 0.8% difference would require a sample of 16k, and that would only be marginally statistically significant. He'd be better off running a simulation in the millions in order to get reliable information about the election.

[wbrocks67]

538 is useless after Silver left. Really sad, because I used to love that website.

Silver had a really bad cycle in 2022 and he didn't really own up to the reasons why he botched it so bad. I feel like he was better years ago but has gotten high on his own supply recently

[Donald Trump’s Toupée]

https://twitter.com/gelliottmorris/status/1806501909139996835

Problem is… Trump gets sentenced in 2 weeks, so polls show will start to muddle.

[wbrocks67]

https://twitter.com/gelliottmorris/status/1806501909139996835

Problem is… Trump gets sentenced in 2 weeks, so polls show will start to muddle.

Yeah there's going to be little time for the dust to settle of the debate before another big event. And then we have the RNC not long after, so it's going to be an interesting month.

[DS0816]

https://projects.fivethirtyeight.com/2024-election-forecast/#:~:text=538%20uses%20polling%2C%20economic%20and,to%20explore%20likely%20election%20outcomes.&text=in%20our%20simulations%20of%20the,of%20no%20Electoral%20College%20winner.

Thoughts?

They’re not worth thinking about.

[VALibertarian]

https://twitter.com/gelliottmorris/status/1806501909139996835

Problem is… Trump gets sentenced in 2 weeks, so polls show will start to muddle.

Yeah there's going to be little time for the dust to settle of the debate before another big event. And then we have the RNC not long after, so it's going to be an interesting month.

Yeah I didn't realize but we've pretty consistently had potential poll-moving events fairly frequently for the past what, 2 months? and not slowing down till the 4 weeks between the rnc and dnc?

[axiomsofdominion]

https://twitter.com/gelliottmorris/status/1806501909139996835

Problem is… Trump gets sentenced in 2 weeks, so polls show will start to muddle.

Yeah there's going to be little time for the dust to settle of the debate before another big event. And then we have the RNC not long after, so it's going to be an interesting month.

Yeah I didn't realize but we've pretty consistently had potential poll-moving events fairly frequently for the past what, 2 months? and not slowing down till the 4 weeks between the rnc and dnc?

Trump trial, debate, immunity ruling, Trump sentencing, convention, a few other things, second convention. Just wild.

[Reactionary Libertarian]

The guy who made this model made the Economist model in 2020 that was totally divorced from reality and showed Biden with a 97% chance of winning. G Elliot Morris is a hack.

[axiomsofdominion]

The guy who made this model made the Economist model in 2020 that was totally divorced from reality and showed Biden with a 97% chance of winning. G Elliot Morris is a hack.

This model is much closer to 538 than the Economist.

Regardless, Biden *did* have an insanely high chance of winning in 2020. Although more like 538's 89% than The Economist's 97%.

[Wisconsin+17]

Well, given that Biden had a +17 poll advantage in 2020, 97% seems low.

Biden's +17 in WI would be the largest landslide since LBJ in 1964. And Biden is no LBJ... Even FDR only cracked that in '32 and '36. 

[oldtimer]

This feels like the 3.6 roentgen meme.

If Morris made a model that's impossible for Trump to lead in it, then it's maximum would be a 50-50 chance.

[axiomsofdominion]

What is happening is that the economy measures that matter to normies, are moving toward Biden which is counteracting the debate, especially since most models discount debates. There's also a lot of reversion to the mean in the models because June is a long time from the election.

So identical conditions next month would shift strongly towards Trump.

[Sorenroy]

[https://projects.fivethirtyeight.com/2024-election-forecast/]

Thoughts?

Seems to me like statistical noise. He wins 506 to Trump winning 490. 1000 is +/- 3.16%. Why don't they run a billion simulations not 1000? To distinguish a 0.8% difference would require a sample of 16k, and that would only be marginally statistically significant. He'd be better off running a simulation in the millions in order to get reliable information about the election.

They run something like 20,000 simulations each model update and select 1,000 of them at random to show users. That still creates a massive set of outcomes (438-100 Dem to 410-128 Rep), but not quite that large.

End of the day though, any changes between the debate and now are - like you said - just noise. There's no big red button they can push at 538 when a candidate has a terrible debate performance. It's all based off of statistical data, which we don't have enough of yet to flush out the pre-debate numbers and override the model's fundamental assumptions. I'm not sure what users like TheReckoning are expecting a statistical model to show. The whole point of a project like this is not to swing wildly off of personal assumptions and expectations. Just give it time for new data to come out and then check back to see what it says.

[E-Dawg]

And Nate Silver, the actual man behind FiveThirtyEight in the first place (and who has a relatively good prediction record), has his own model which estimated pre-debate that Trump has a 66% chance of winning. And his model is very similar to the FiveThirtyEight models used for those past elections. And I'm sure that when the post-debate polls are released, Trump will take the lead in the FiveThirtyEight model, and will expand his commanding lead in Nate Silver's model even further.

https://projects.fivethirtyeight.com/2024-election-forecast/?cid=rrpromo

https://www.natesilver.net/p/nate-silver-2024-president-election-polls-model

And Trump has now barely pulled ahead in the FiveThirtyEight model, and in Nate Silver's model he is up from 66% to 72%.