Charles Rotter
The midterms are on November 3. Millions of Americans have already voted, every state that reports early ballots is putting out daily numbers, and the polls keep coming. Everybody is looking at the same spreadsheets. Two completely different elections are being watched.
I’m not making any claim about who’s right, what the numbers mean, or who wins. Every reading below belongs to the person who made it. What interests me is how sure everyone is.
Here is the one on the forecast pages. Nate Silver’s Silver Bulletin generic ballot average had Democrats at D+9 on October 26, and D+9.5 after its likely-voter adjustment. His forecast model gives Democrats a 76 percent chance of taking the Senate. Adrian Beaumont, writing at The Conversation, sums it up:
Silver’s model gives Democrats a 75% chance to win the Senate and a 95% chance to win the House, respectively 75% and 94% on October 19.
And here is the one on a different part of X. An account called Latinx Adjacent Doctor PhD, @TonerousHyus, whose bio reads “Former Democrat who wised up,” posted this on October 24:
YouGov D+5
TIPP D+5
Rasmussen D+4
Echelon D+5Everyone comes around to Tones eventually
https://x.com/TonerousHyus/status/2114134701304582515
Same polling industry, same week. One side averages its way to D+9. The other picks its four and gets D+5. Both will tell you the other side is cherry-picking.
Ohio is the cleanest example, because Ohio’s Secretary of State publishes the early vote every day by county and by party, with party inferred from primary voting history. CNN’s Ethan Cohen looked at those numbers on October 11:
At this point in 2024, Republicans had a 30-percentage point lead over Democrats in counties that voted for President Donald Trump by 20 points or more, according to data from the Ohio Secretary of State. This year, that’s shrunk to just five points.
https://www.cnn.com/2026/10/11/politics/early-voting-midterms-ohio-democratic-shift-vis
Two weeks later the doctor went through the same state’s data and found a different movie:
So Ohio has its Saturday under its belt for voting and I had to know if 18–29 voters went out today. In 2024 today had 28% 18-29 voters and double the votes of Wednesday
Sorry (not sorry) to report the youth are gone
https://x.com/TonerousHyus/status/2114102467067589107
| Who | What they measured | At this point in 2024 | 2026 |
|---|---|---|---|
| CNN, Oct 11 | Republican lead over Democrats among early voters, counties Trump won by 20+ points | 30 points | 5 points |
| @TonerousHyus, Oct 24 | 18–29 voters on the first Saturday of early voting | 28% of that day’s voters | “the youth are gone” (no figure stated) |
One reading counts party share, the other counts age share, and both measure against 2024, a presidential year. CNN did attach a caveat in the same article: “it’s far from determinative,” and Republicans “in recent years have much been much more likely to vote early in person or on Election Day.” Caveats don’t travel as well as headline numbers.
Chris, @chriswithans, reads the polls the way the doctor does, and says why:
In the most shocking news imaginable, Democrats are forced to temper expectations just as the early vote shows the polls are ludicrous because those polls are overloaded with Harris voters, while the early vote clearly is not.
https://x.com/chriswithans/status/2114024956875231277
The Conversation piece makes the opposite argument about the same polls, that likely-voter screens help Democrats because “Democrats are more likely to vote at lower turnout elections like the midterms than Republicans, when Trump is not on the ballot.” One side says the samples are stuffed with Democrats who won’t show. The other says the screen for who will show is exactly what makes Democrats look better.
Chris can also read a number with his team jersey off. Twenty-five minutes earlier the same morning:
If accurate (always a big if), just looks like mutually high turnout to me. Dems running between 2024 and 2022. They should run better than 2022 in the early vote just because of the environment; doesn’t mean E-day will be like 2024 in a midterm.
https://x.com/chriswithans/status/2114018832797016299
Strip out the commentary and that isn’t far from the CNN caveat.
The account that comes closest to just reporting the scoreboard is CA ET Nerd, @earlyvotedata, who posts state-by-state early vote updates all day and leans right. On Florida, October 24:
Republicans were up in Florida until they weren’t…until they might be again.
https://x.com/earlyvotedata/status/2114131440439046424
And on the theory that decides which movie you’re in:
It’s the theory that Democrats are frontloading their votes, so patterns from the prior elections will be different than this one.
A theory that we have to see play out, but it seems to be happening in many other states with different levels of Republicans gaining back.
https://x.com/earlyvotedata/status/2113610215803482233
As both camps frame it, that’s the whole fight. One camp says Democrats are banking votes they would have cast anyway. The other says those are new voters. Each camp reads the same daily totals as proof. November 3 settles it, and not before.
The other side’s election-watchers are just as sure. One left-leaning account with ten thousand followers, @bdquinn, on Monday night:
I just have to look at the LatinX Doctor guy and go “You can’t possibly think these are favorable numbers for Republicans, unless a big share of the Black vote is defecting.”
https://x.com/bdquinn/status/2114892404566196602
Same numbers. “You can’t possibly think.” That’s the tell.
Even the true believers notice it sometimes. Christian Heiens, @ChristianHeiens, who is nobody’s idea of a centrist:
I love how depending on where you look people are coming up with outcomes that rage from 2022 in reverse to 1994 in reverse.
If you look at Michigan, the early vote looks worrying for Democrats, but if you look at Virginia, it’s the political equivalent of the Chicxulub asteroid.
https://x.com/ChristianHeiens/status/2114067188323664141
Scott Adams had a name for this years ago. He called it two movies on one screen. From his blog, February 12, 2017, archived here:
I have been saying since Trump’s election that the world has split into two realities – or as I prefer to say, two movies on one screen – and most of us don’t realize it. We’re all looking at the same events and interpreting them wildly differently. That’s how cognitive dissonance and confirmation bias work.
https://www.scottadamssaid.com/good-example-of-our-two-movie-reality/
And the part people skip:
In order for our two-movie situation to merge back into a single movie, one of us needs to see our expectations violated in ways that even cognitive dissonance can’t explain away.
https://www.scottadamssaid.com/good-example-of-our-two-movie-reality/
Elections are one of the few places where that actually happens on a fixed date. Most of our disagreements never get a final reel.
There’s research suggesting being good with numbers doesn’t save you. Dan Kahan and colleagues, in “Motivated numeracy and enlightened self-government”, gave people the same made-up data table and changed only whether the answer flattered or offended their politics. When the topic was a skin cream, the more numerate subjects got it right more often. When it was gun control, the more numerate subjects split further apart along party lines. Skill went into defending the movie.
The better you are with a spreadsheet, the better your movie looks.
Readers here won’t need the bridge spelled out. Climate has been two movies on one screen for thirty years. Same thermometers, same satellites, same tide gauges, and one theatre watches a slow, manageable warming while the other watches an emergency. The difference is that an election has a hard end date. On November 4 one set of these people will be wrong in public, and there is no adjustment that fixes it.
I’m not going to tell you who wins. I don’t know, and nobody quoted here does either, whatever they post.
Which movie are you watching?
How would you know if it was the wrong one?
And what will you say on November 4 if you were wrong?
