538 senate model now out

[Antonio the Sixth]

Silver once again shows his hand. He says that this year his default will be the deluxe model for seemingly no other reason other than that it's the most favorable to Republicans.

Don't be such a hack. 538 always uses the fundamentals-inclusive model as its default. It did the same thing in 2018 and in fact that time it led to overestimating the Democrats (since most incumbents were Democrats that cycle). There's a legitimate criticism that his model weighs incumbency too heavily, but that's different from accusing Nate ing Silver of having a partisan agenda.

[Antonio the Sixth]

The amount of bullsh*t Nate Silver gets from statistically illiterate morons who can't fathom a model that doesn't confirm their priors is genuinely staggering. I don't even like the guy! He's just a bland, unoriginal pundit who happens to be good at math. And yeah, he makes plenty of modeling choices that I think are silly (I already criticized the experts thing in 2018). But some of you people are incapable to criticize him on anything more than dumb partisan hackery. It's really sad.

[Antonio the Sixth]

The amount of bullsh*t Nate Silver gets from statistically illiterate morons who can't fathom a model that doesn't confirm their priors is genuinely staggering. I don't even like the guy! He's just a bland, unoriginal pundit who happens to be good at math. And yeah, he makes plenty of modeling choices that I think are silly (I already criticized the experts thing in 2018). But some of you people are incapable to criticize him on anything more than dumb partisan hackery. It's really sad.

It’s not partisan hackery to point out legitimate flaws in the model. Silver isn’t some God whos above criticism.  This model:

1. Has Jones as more likely to win than Ossoff, Bollier and Hegar. I don’t even need to explain why this is absurd.

2. Gives McSally a better chance of winning than James. Sorry, but McSally is down by double digits on average, is underperforming Trump, and would probably lose even if he carried the state. She’s beyond DOA. James on the other hand is performing on par with Trump, has narrowly the gap recently, and is in a much more demographically friendly state.

3. Gives Bullock and Gardner equal chances of winning. Yeah, in a D leaning year an incumbent Republican Senator trailing by double digits in a D trending state is just as likely to win as a Democratic Challenger who’s trailing by no more than 3 on average in a state that’s quite friendly to Democrats downballot. Not to mention that Biden will win CO by more than Trump will win MT, numerous polls have shown Trump winners MT by mid to high singel digits, while nearly ever CO poll has shown Biden up double digits. Gardner requires much more ticket splitting to win than Bullock. To give them equally probabilities of winning is just malpractice.

This has nothing to do with confirming anyones “priors”, it’s about Creating a model that doesn’t contradict polling and partisan lean of individual states just because Silver is worried about getting burned again like he did in 2016.

No, this is exactly what I'm talking about. You could say that the model gives too much weight to incumbency and expert ratings, and I'd be inclined to agree with that, but instead you have to make sensationalistic appeals to specific race forecasts (criticizing a statistical model for producing forecasts you disagree with is basically like being against statistical models as a concept) and baselessly assert that Silver is deliberately skewing his model toward the GOP even though it makes absolutely no sense for him to.

If we needed any proof, the fact you seem to believe he got "burned" in 2016 definitively shows that you have no clue what you're talking about.

[Antonio the Sixth]

Hmm, yes, the "being down 18 points in the polls make you more likely to win" take from the 538 classic model here:

<snip>

The other two models also show (less dramatic) swings towards Jones. I'm dumbfounded at what they're seeing that's good news for him here. This is literally the worst poll for Jones so far this cycle.

Looks like it was an error. They updated it and now Jones only has a 24% chance to win in the Deluxe model (down from 28% yesterday). Probably just reversed the numbers or something.

Not an error per se, but unintended behavior.

See, the rational answer to faulty model outputs is to identify specific design flaws and correct them. But some people prefer to just scream "NATE SILVER IS A HACK HE'S JUST COVERING HIS ASS SOMETHIN SOMETHIN 2016!!!1!!1!11" it seems.