“It was staring me right in the face, but at first I missed it,” the Northwestern University data scientist told this reporter by phone on Sunday. “I saw this huge jump in Harris’s support on July 31st, but didn’t put it together with Trump’s appearance at the National Association of Black Journalists convention that day. That event, and not the debate that just made things worse for Trump, marked the decisive turning point in the campaign.”

Miller’s election forecast is based not on polls, but on the prices for both candidates posted on the PredictIt betting site. He regards the PredictIt odds as far more reliable than polls, which reflect voter preferences that are four to five days old. And since they typically survey 500-1,500 likely voters, polls reflect a great deal of statistical “noise”—hence the wide variability in the numbers posted by the various modelers.

PredictIt is the most liquid betting market, averaging around 37,000 wagers a day, according to Miller. And given that each player is subject to a $850 limit, no single bettor or group of high rollers can artificially inflate the odds for one candidate or the other.

Trump led before the NABJ debacle

The Miller model posits first that the PredicIt odds closely reflect popular vote percentages. Put simply, a candidate given a 55% chance of winning, or priced at 55 cents on PredictIt, is likely to receive a similar share of all ballots cast. Second, Miller shows that historically, the popular voting shares closely track the portion of the 538 electoral votes each contender receives. That relationship, he found, has been extremely stable over every race since 1960.

Miller’s homepage, The Virtual Tout, displays a graph showing the share of electoral votes trending towards the Democratic side, overlaid by the events that have significantly moved the odds, and hence the swings in the projected electoral count around the 270 needed to prevail.

Between July 21—the day President Biden left the contest and endorsed Harris—and July 27, her electoral count rose substantially. After that, her numbers went flat for four straight days.

“She was still well behind the former president, and it looked like her electoral numbers had plateaued,” Miller says.

But then, Miller contends, a tremor struck that could very well turn into a Harris landslide by November. On July 31, Trump falsely suggested at the NABJ’s annual colloquy that Harris had altered the way she characterized her racial heritage, questioned her bi-racial background, and charged the VP of “happening to turn Black” and that Harris “now wants to be known as Black.”

Though the incendiary comments raised outrage in the press and among pundits, virtually no one has pegged Trump’s NABJ interview as the pivotal juncture in the election. Miller points out that the PredictIt market turned frenzied that day as bettors shifted en masse from Trump to Harris.

“Over 100,000 shares traded that last day of July, three times the usual number,” he says. “Literally overnight, the election shifted from leaning Republican, to trending Democratic, as Harris surged to over 270. Trump’s statements at the NABJ conference proved a complete disaster for his campaign. It had nothing to do with anything Harris did. The huge shift was all Trump’s doing.”

Following the NABJ debacle, Trump partly closed the chasm—then came the debate

Miller’s chart shows that Harris’s electoral count kept climbing in the two weeks that followed, reaching a peak shortly before the start of the Democratic National Convention. But the Windy City extravaganza itself failed to provide an added bump. By early September, her numbers had drifted downwards slightly. And on Sept. 6, news that Trump’s hush money trial would be delayed until after the election lifted his numbers. The day before the debate, he trailed by only a narrow margin.

“At that point, though Harris still led, the race was almost a dead heat,” says Miller. “It’s remarkable that most of the jumps in Trump’s numbers come as the result of good news about his legal issues.”

Then, the face-off in Philadelphia sent Harris’s forecast electoral count up big time. “That increase was the combined result of the debate and Taylor Swift’s endorsement of Harris,” says Miller.

As of Sept. 22, PredictIt prices suggest Harris’s odds of winning stand at 56.3% versus 43.7% for Trump. Those odds, Miller contends, would translate into an overwhelming win for the vice president with 43 days to go.

“Big events can change things, wars that could alter the race are raging, candidates can make big mistakes,” he cautions.

But right now, he says, Harris is way ahead, and the polls haven’t caught up with the huge win that’s probably building—and started building the day Trump made those disastrous comments to Black journalists, and blew the lead that he’s never regained.