Trump’s darkening prospects mark a dramatic reversal from the election’s dynamic less than 13 days ago. During the first three weeks of October, Donald Trump staged a remarkable comeback, rebounding from a huge deficit to a commanding lead. With less than two weeks to go before Election Day, Trump appeared en route to a smashing victory.

Harris countered the Trump surge by pivoting from an attack on Trump’s policies to spotlighting his “unstable” personality and the “obsession with revenge.” That message failed to resonate with voters as Trump moved relentlessly upwards in the electoral vote count, as forecast by the Miller framework. By contrast, Trump was tapping a powerful undercurrent: The deeply unpopular Biden record, especially on the economy.

“The macro numbers on growth, and employment look good, and the Democrats keep touting them,” says Miller. “But people don’t care about GDP or the national jobless rate. They care that they’re paying so much more for groceries than four years ago, that they can’t afford to buy a first home because mortgage rates are so high, or afford a car loan to replace the beat-up model in the driveway, or that they have no savings and need to work two jobs to get by.”

In other words, although Harris stresses that the stats look fine, people don’t feel fine due to the dollar squeeze in their own lives. Plus, Americans are fretting more and more about this nation’s involvement in foreign wars. The Biden administration’s policy of sending arms to Israel for bolstering its forces in the war versus Iran, and to Ukraine for fortifying its campaign to defeat the the Russian invasion is deeply troubling to a large swath of the electorate—especially since it’s unclear how long those conflicts, and hence our involvement, will last. Trump, on the other hand, has been striking a quasi-isolationist stance on the stump that seems to be finding favor on the trail.

Put simply, Americans are pissed at where Biden’s led the U.S., and by extension, at Harris.

Time was so short, Miller concluded, that Harris was unlikely to significantly close the yawning divide by taking new policy positions, shifting her campaign rhetoric, or even upping her ground game. “My view was that only a major shock could change the course of the race,” he says, “meaning an earthquake that hugely benefited Democratic ticket.”

The Madison Square Garden rally proved the turning point

The earthquake happened. It wasn’t ignited by the Harris camp, but an unforced error by the Trump team. The event that transformed the election was billed as the capstone to a movement bound for glory: The Oct. 27 rally at Madison Square Garden, in the heart of Trump’s home city of New York, before rowdy audience of 20,000.

The spectacle boasted around three-dozen speakers, but few top politicians and no widely-respected statesmen. They divided mainly between highly controversial Trump loyalists who embodied—and from the podium, championed—the candidate’s most extreme views, and marquee names from the worlds of entertainment and right-wing media.

The cast encompassed such divisive figures as disgraced former New York Mayor and onetime Trump lawyer Rudy Giuliani, who was disbarred and suffered a $148 million judgment that forced him into bankruptcy for falsely claiming that Georgia officials tried to rig the 2020 election; Stephen Miller, Trump’s former White House advisor who advocates draconian measures to curb immigration; Robert F. Kennedy Jr., the gadfly Trump has promised to install in a top position if he wins, and whose radical environmental views contradict the former president’s; and Elon Musk, the multi-billionaire given to outrageous tweeting in his self-owned platform X.

Also onstage were Tucker Carlson, fired from his perch as a Fox News host for promoting unfounded conspiracy theories, and Hulk Hogan, who in full wrestling regalia mounted a chest-pounding performance that drew roars from the crowd he dubbed “Trump-o-maniacs.”

The speeches unleashed shockwaves just when Trump needed to stay smoothly on course. The most infamous gaffe: insult comedian Tony Hinchcliffe calling Puerto Rico “a floating island of garbage in the middle of the ocean,” followed by his howler that Latinos “love making babies.” Businessman Grant Cardone’s charge that “Harris and her pimp handlers will destroy the country” amplified the negative messaging. During his address, Trump himself kept the roast going, characterizing his opponent as “a low IQ individual.”

According to Miller, the gathering that Trump called a “love fest” seems to have spread not unity and inclusion, but hostility—especially to women. None other than Nikki Haley, Trump’s leading female rival for the nomination, expressed that view in an Oct. 29 interview for Fox News.

“This campaign is not going to win talking about crowd sizes or whether she’s dumb,” intoned Haley. “Your target market is suburban women who are college educated. This bromance thing, this masculinity stuff [on display at MSG] it borders on edgy to the point where it’s going to make women uncomfortable. Fifty-three percent of the electorate are women. They care about the issues and how they’re being talked to. There was no reason to have a comedian at the event,” where the swipe at Puerto Rico proved “harmful” to the Trump-Vance ticket.

In the days that followed, while the Dems paraded the Hinchcliff disaster in the media, Trump advocates kept the bad vibes coming. A super PAC launched by Musk shared a video on X that that used vulgar wordplay to demean the vice president. Then in a Carlson interview that aired on Nov. 1, Trump denounced former Congresswoman and long-time foe Liz Cheney as a warmonger, and suggested she be sent into battle, proving she’s a coward and hypocrite.

“Let’s put her with a rifle standing there with nine barrels shooting at her, OK?” stated the GOP standardbearer. “Let’s see how she feels about it. You know, when the guns are trained on her face.” 

The MSG extravaganza and the follow-on screwups are sinking Trump’s numbers, says Miller

Miller bases his model not on the polls or opinions of the pundits, but on the prediction or betting markets. His main data source are the prices posted on what he deems the most trustworthy, and highly liquid, political wagering platformÚ PredictIt. Miller applies his own methodology to the PredictIt odds, and translates the result into the share of electoral college votes that each candidate commands at any given moment. Miller updates the counts every minute, and you can follow the race, practically in real-time, on his site Virtualtout.io.

This writer first followed Miller’s methods in the 2020 election, and was impressed by the discipline of his approach and his accuracy in gauging results for both the presidential contest, and the two Senate seat runoffs in Georgia that secured control of the upper chamber for the Democrats.

Miller’s numbers show a jaw-dropping swing to Harris that would have seemed unimaginable two weeks ago. On Oct. 26, Trump tallied 367 electoral votes to just 171 for Harris, putting the GOP nominee 196 in front. The next day, Trump headed his blowout at MSG, and no sooner did the giant screens go blank than he started losing ground. On Monday, Harris gained 18 electoral votes, and she kept improving every day through midnight on Thursday. By then, Harris had gained 58. Trump’s lead shrank by over half from 196 to 80.

The drop accelerated from there. On Friday, Trump’s horde fell by an extraordinary 39 electoral votes, lowering his total to 270 (the number needed to win), against 268 for Harris. By 10 a.m. on Saturday, Trump had shed another 5, putting Harris in the lead by by 273 to 265. All told, in the seven days since Trump peaked on Oct. 25, he’s lost 102 electoral votes. According to the Miller model, what looked like an invincible lead collapsed in a week.

Where does the race stand now, according to Miller?

“It’s been going back and forth around the 270 line,” he told Fortune. “Right now, it’s a tossup according to the PredictIt numbers. The big question in my mind is, how much Republican bias is there in the prediction markets?”

Miller notes that he assessed the GOP tilt in the 2020 races for PredictIt, and made the right adjustments for it, as evidenced by the accuracy of his forecasts. For the 2024 election, he’s using the same correction methodology he deployed four years ago. But now, though he hasn’t put precise numbers on the difference, Miller reckons that the betting sites lean more strongly to the GOP than in 2020.

“It’s my experience working studying prediction markets in the past that lead to that conclusion,” observes Miller. “I believe all of the prediction markets are more biased to Republicans than in 2020.”

The sites overestimate the GOP odds, he says, in part because the betters are mainly males who often also wager on sports, and love taking risks. He adds that Musk’s lavish praise for prediction markets may have pushed the scales in further the Trump-Vance direction, noting that a few Trump “whales” could be inflating his odds of victory on some markets that allow individuals to wager unlimited sums.

A main reason he uses PredicitIt, Miller avows, is that the site imposes a relatively low limit on the dollars any single gambler can place on a candidate. Hence, the platform can’t be swayed by rich Trump fans. Miller reckons that PredictIt’s prices are less exaggerated than the odds on competing sites, but his preferred platform still tips more Republican than 2020. As a result, he says, his model’s electoral vote count could be understating Harris’s newfound advantage.

“Her edge is probably higher than the handful of votes shown on the continuous feed for virtualtout.io,” he says. “It’s just impossible to completely correct for the tilt. Even though it looks like a tossup by the Virtualtout numbers, I’m anticipating Harris-Walz victory.”

It was just a week ago that Miller’s forecast picked Trump as the probable winner. In addition to the expected Republican bias in the prediction markets, the direction towards Harris, says Miller, is consistent with a “fundamental” in calling the outcome. “Candidates usually win by moving to the center, and Harris has claimed the center. Trump’s just kept going more and more to the extremes,” he says.

Miller also notes that his real time “ticker tape,” which he calls “tinyticker,” shows the impact of campaign events instantaneously, whereas the polls capture these changes, and often inaccurately, only after four or five days. His system caught this sudden, massive transition just when it began, and has been tracking its trajectory minute by minute since.

Time looked too short for Harris to rebound, yet it happened. Now, the question is whether in a span of three days, the former president can reverse a trend that he himself set in motion, on one of the world’s biggest stages, in a belly flop heard coast to coast.