Miller, a professor at Northwestern University, bases his forecasts not on polls, the pundits’ views, or the takes from such “fundamental” analysts as “Nostradamus” Allan Lichtman of American University, but betting prices. His sources are the odds posted on the platform he rates as most trustworthy, PredictIt. He deploys a proprietary model that translates the PredicIt odds into EVs; it’s a system that proved highly accurate in calling both the outcomes and margins of victory in the 2020 presidential race, and the two Georgia Senate runoffs that followed, both clinched by Democrats.

Before unveiling virtualtout.io just a few days ago, Miller updated his wagering forecast each day at midnight. Since late September, the race’s trajectory tracking those postings careened wildly. On September 20, shortly after her big win in the debate, Harris boasted a gigantic lead of around 200 EVs. Trump surged back beginning early last month, and by October 26 had built an advantage that almost equalled Harris’ bulge five weeks earlier. With nine days left, It appeared that Trump was coasting to an easy win.

Harris’ fortunes improved dramatically following Trump’s six-hour rally at Madison Square Garden, a chest-thumping extravaganza that the bettors reckoned would antagonize female voters on the fence. On November 2, Harris regained the lead at 301 EVs, for an edge of 57.

But buckle your seatbelts and grab your hats: The rollercoaster took still another lurching turn: At the close of November 4, Trump vaulted back in front, though by a super-slim 12 EVs.

On the morning of November 5, the Data Scientist’s odds show heavy betting that Trump will win

On Election Day-eve, the virtualtout.io site—it’s updated every minute—showed the EV tallies for Trump and Harris toggling back and forth around the 270 needed to win. Put simply, the bettors were calling the race what the experts have been claiming and polls been showing for weeks: A coin flip. But the next morning as Americans headed to the polls, the outlook shifted once again. Trump’s been consistently ahead so far today. But his lead has varied from big to narrow. Miller cautions that the contest remains highly volatile in its waning hours, and could take still another unexpected turn—this one towards Harris. He notes that November 4 was the most active trading day of the campaign on PredictIt; 518 thousand shares changed hands, over ten times the average during the last few months. And Miller expects even higher trading today. That frenzied activity, he judges, highlights that hordes already wagering on PredictIT are changing their minds about which contender to put their money on, and that sundry new players are crowding the site, and placing bets that could cascade into waves, changing the course of the election.

He cites five reasons why the Dems should still harbor strong hopes. First, as of midnight on the 4th, the figures were extremely close, with Trump ahead by just a dozen EVs. in Miller’s words, separated by “just a state or two.” Second, Miller stresses that the betting markets, including PredictIt, are biased toward the Republicans. That’s because the majority of participants are males who often wager on sports. “You have a big gender gap favoring Democrats in the election, and also a gender gap in the prediction markets,” says Miller. “It’s not as great as for the electorate, but it exists. Also, though betters are backing whom they expect to win, there’s a slight tilt towards whom they want to win, and that also helps the Republican odds.” Miller corrects for the Republican bias using the same methodology he deployed in the 2020 races. But the data scientist’s unsure he’s made a big enough recalibration towards the Democratic side. He acknowledges that if his model doesn’t sufficiently discount the slant towards the GOP, the space Trump enjoys over Harris on virtualtout.io early on Election Day may be smaller than the modest, and vacillating, margin displayed on the site, or even nil.

Third, Miller argues that though the “technicals,” represented by the betting odds, and to some extent, the polls, lean towards Trump, the “fudamentals” should boost Harris’ chances. Among the campaign staples she’s advocating and that usually win, he says, is a centrist, hopeful and upbeat message versus Trump’s extreme positions and negative messaging on immigrants and an economy that’s actually displaying good numbers overall. The dark, dour approach is usually a loser, says Miller.

Fourth, most of the movement in the past week has been decisively in Harris’ direction. Miller says that longer-term trend may be more significant than the Trump spike on November 5.

Fifth, Harris’ closing message sounds like a winner to Miller. It spotlights her willingness to work across the political spectrum, taking the best ideas from such progressives as AOC and inviting a Republican to join her cabinet. “Americans want to see both sides working together, and that’s what she’s advocating by reaching out to Republican voters,” he declares.

“The betters think Trump will win,” Miller avows. Although his model’s was right on in 2020, he’s a lot less sure it’s the North Star this time, especially because the GOP tilt on PredictIt is so hard to gauge. “The ticker will be showing all the twists and turns, which way the race is heading all night, well before the networks call the crucial states,” he concludes. “Things could get wild. I don’t see a definitive, set pattern being established until the returns start coming in. The bettors will change their minds if they see Pennsylvania or North Carolina turning blue. With the data we’re seeing, the market could be dominated by day traders.

But in this contest, the one certainty is that nothing’s been predictable, and for Miller, even the folks risking their own money, who usually get it right in his predictions, may be getting it wrong.