538 model & poll tracker thread

[Beefalow and the Consumer]

A KY average has just been generated: Trump +17.4%.

Trump won Kentucky by nearly 30% in 2016, but 538 only had him up by 20% on the eve of the election. I wonder if we'll see the same thing in 2020, where Trump outperformed the forecast in deeply red states.

[Beefalow and the Consumer]

I pulled together this map from 538's numbers. In states where they don't have their own average, I've just used the average of polls since the race since the candidates have been known. In states with no polling, I've just used the 2016 numbers.

This is my map, assuming the following:

Less than 3% lead = >40%
Less than 22% lead = >50%
Greater than 22% lead = >60%
States without a poll average get slotted in their 2016 order

They have Biden +0.3 in Iowa, so Iowa goes >40% Biden.

[Beefalow and the Consumer]

I imagine some of these outlandish scenarios require some sort of... extreme demographic event.

[Beefalow and the Consumer]

Oh yes.
I forgot there was a grade slider (top-left corner area), to weed-out the lower-graded pollsters.
Nice. Thank you for this image.

I used it a week ago and found that out of all the polls released since the beginning of June by A+ pollsters only one or two didn't show a double digit Biden lead.

Trump has not led in any poll graded B or higher since Biden clinched the nomination. Not a one. I count two polls where Trump and Biden are even, 19 where Biden is up 3 points or less, and 55 where Biden led by double digits.

If the roles were reversed, we'd all be resigned to another four years of Trump.

[Beefalow and the Consumer]

Trump is down to a 15% chance of winning.
Biden has a 35% chance of winning in a landslide.
Georgia is on the cusp of flipping.
I am beginning to wonder where the model can bottom out.

At this point I would put Biden as an odds-on favorite to outperform Obama 2008 both in the popular vote margin and the EV count. I doubt it will be long before 538 flips GA and IA.

[Beefalow and the Consumer]

Today or tomorrow, Biden will probably cross from "Favored" to "Clearly Favored"

I think the "Clearly Favored" threshold is 90%. I'm sure range of expected outcomes will narrow as we get closer to election day, but I don't know if it will happen that quickly.

A continued onslaught of Biden +10% polls might do the trick, though.

[Beefalow and the Consumer]

[Beefalow and the Consumer]

Biden's national lead now over ten points in the 538 tracker:

https://twitter.com/baseballot/status/1314556769226682369

I think this is the first time he's had a lead of at least ten points nationally?

I believe it is.

I'm wondering if Clinton's 6.5 point lead was partially due to Access Hollywood two days previous. Her bump from that peaked at 7.1% (RCP), then crashed after Comey announced he was taking a close-up look at Weiner.

That brings up another point: Oct, 2016 was a crazy month of bombshell news stories and constantly swinging undecideds. 2020 is the polar opposite of that. Biden has led by 6 points or more since George Floyd, and I can't imagine anything short of Biden being incapacitated dropping his lead below 6 points between now and the election.

[Beefalow and the Consumer]

I wonder how long Biden's polling average margin needs to stay north of 10 points before the odds of a Biden landslide ticks up to 50%. It's currently at 36%.

[Beefalow and the Consumer]

I made the following post in the weird 538 maps thread about just how ridiculous Trump's paths to victory are in the model, the number of state upsets he needs to win, and just how little room there is for error:

Here's a compilation of every Trump victory map right now:

Red = Biden in all scenarios
Blue = Trump in all scenarios
Gray = States that differ between scenarios

Even in the wildest of outliers, VA and OR are now gone. That sounds like it should be obvious, but there are still outliers on Biden's end that have MS, TN, KS, IN, and LA going to Biden. Yes, the model is still spitting out results with those states going to Biden, but none at all with VA going to Trump. That's remarkable.

The other notable thing about this is that among the Trump states in EVERY Trump victory, including the ridiculous outliers, we have:

Florida (Biden leads by 4.4%)
North Carolina (Biden leads by 2.7%)
Ohio (Biden leads by 0.9%)
Iowa (Biden leads by 1.1%)
Georgia (Biden leads by 1.4%)

Trump victory maps are as much outliers as maps where Biden wins 430+ EVs. That's where the race is, according to 538.

[Beefalow and the Consumer]

86-14 now.

Wisconsin has replaced Pennsylvania as the tipping point state.

Yes, in the sense that PA zoomed left enough that it pushed Wisconsin down, even though Wisconsin has been steady. NE-02 is about to cross into likely territory as well.

I selfishly want Wisconsin to be the projected tipping point state. I am planning on live mapping (two party vote and swing) the state on election night.

You might just get your wish the way that polling is looking like in PA.

I think it needs to be pointed out that the states are not ordered by probability, they're ordered by expected margin of victory. It's a subtle point, but important

[Beefalow and the Consumer]

Current 538 polling averages, Atlas style!

Biden 334 Trump 164 Tossup 40

T = <1% lead
30 = 1-3%
40 = 3-5%
50 = 5-7%
60 = 7-9%
70 = 9-11%
80 = 11-13%
90 = >13% (or not a realistic battleground)

[Beefalow and the Consumer]

I don't know about anyone else, but the number I'm watching is the "Biden wins in a landslide" probability. It's been as high as 37, now back to 36.

If we keep getting polls like this from now until election day, it may approach 50.

[Beefalow and the Consumer]

I don't know about anyone else, but the number I'm watching is the "Biden wins in a landslide" probability. It's been as high as 37, now back to 36.

If we keep getting polls like this from now until election day, it may approach 50.

Anything more than Biden + 7 and a D senate I'm happy with

+7 is probably just enough to close the door on Trump contesting the results on the grounds of mythical "vote by mail fraud." If Biden wins the tipping point by 5% and wins several more states by more that 1%, there are really no legal or propaganda avenues Trump has to fight the result.

My nightmare, apart from Trump winning outright, is one or two decisive states with a GOP legislature and a result close enough to challenge the mail-in ballots. Or even with Democratic governments. The kidnapping plot against Gov. Whitmer is evidence that there are a whole lot of domestic terrorists out there willing to take matters into their own hands and try and subvert the process.

To avoid this and have an election result that is beyond doubt, a 7% NPV margin is in my estimation the bare minimum. We need to run up the score.

[Beefalow and the Consumer]

Still waiting for Biden to cross that 90% line and become "clearly favored."

[Beefalow and the Consumer]

By my calculations, this is where Biden becomes "clearly favored."

[Beefalow and the Consumer]

By my calculations, this is where Biden becomes "clearly favored."

He'll keep rising a bit each day probably. I guess the reason we've barely seen movement in the past few days is we've received barely any polling.

Biden is blowing out Trump in every national poll that gets released, so I'm guessing all it will take is a few good state polls to put him over the line.

[Beefalow and the Consumer]

In the absence of much new polling, Trump's chances will steadily decrease toward Election Day as the incumbent/economy factor gets less weight.

Especially considering how few undecideds there are. That was the big difference between 2012 and 2016. The 2012 forecast gave Obama over a 90% chance of winning, despite his national lead over Romney being smaller than Hillary's over Trump. The range of uncertainty was much smaller.

As we get closer to Election Day, the range of likely outcomes will definitely narrow.

[Beefalow and the Consumer]

Meanwhile, gamblers on PredictIt has gone full MAGA and flipped Florida red.

If you're looking for a buying opportunity, now's the time.

https://www.predictit.org/markets/detail/5544

[Beefalow and the Consumer]

Meanwhile, gamblers on PredictIt has gone full MAGA and flipped Florida red.

If you're looking for a buying opportunity, now's the time.

https://www.predictit.org/markets/detail/5544

This makes sense if you ignore literally every district-level poll we've gotten outside of Miami-Dade. Further proof that betting markets are ridiculous!

I have no ethical qualms about separating delusional MAGAs from their money.

[Beefalow and the Consumer]

https://twitter.com/adamrpearce/status/1321116988534652929

The biggest negative correlation I could find was WA-MS. I'd struggle to think of any pair of states that are more different from one another politically.

[Beefalow and the Consumer]

The probability of a Biden landslide had been as high as 37%, but it's now back to 29%. That isn't surprising, given the confidence interval narrowing as we approach Election Day.

[Beefalow and the Consumer]

I'm still a bit at a loss to explain how Biden has only gone from 87% to win to 88% to win in the last two weeks. 

It really seems like Biden should have improved more in the forecast than this if there hasn't been significant poll tightening, just from reduced uncertainty.  Didn't Nate say a few weeks ago that Biden would be 95% to win by election day if polls didn't tighten?

The polls have been tightening a bit. It's due to some questionable pollsters, but the model is probably conforming to the assumptions baked into the algorithm.

[Beefalow and the Consumer]

Trump has dropped to 10 in the model.

Heavily favored, here we come! (Three more dots!)

[Beefalow and the Consumer]

It's a rounded 90, he's actually at 89.6 or so.  I think he needs to actually reach 90 to be "strongly favored."

By my calculations Biden is one "dot" away from "clearly favored."