Four years ago, Donald Trump became the second presidential candidate after George W. Bush in 2000 to lose the popular vote and take the White House. It’s a given that he’ll come up short in the national count this time, too. The outcome hinges on Trump’s ability to keep that deficit narrow enough to secure a second victory in the electoral college. In 2016, Trump prevailed with a relatively low number of total ballots, because third-party candidates amassed a far higher proportion than in most elections, and Trump got the maximum mileage imaginable from his haul by edging Hillary Clinton by less than a point in no fewer than three electoral vote-rich states, Michigan, Wisconsin, and Pennsylvania.

To win reelection, Trump needs to immensely boost his performance in the popular vote. Paul Krishnamurty, a professional political gambler and elections expert for the U.K. betting site Betfair––the venue has booked $260 million on the election so far and expects that number to double in the days to come––reckons that Trump must raise his 2016 count of 63 million by roughly 11 million, or 17%, to stand a decent chance. That’s the net increase required, meaning as of today, Trump must squeeze enough new votes from his white, blue-collar, non–college-educated base and groups newly leaning his way to make up for his losses among seniors, independents, and disgruntled Republicans, and still add 10-plus million to his 2016 total. Trump can get part of the way by rallying his hard core of white working-class voters and notching gains with Latinos impressed by his handling of the economy and tough stance on Cuba.

“He seems to think he can get millions of working-class folks who weren’t enthused enough to vote for him the last time, to vote for him this time,” says Krishnamurty. “Some of them will, but it won’t be nearly enough.” Adds University of Florida professor Michael McDonald, who heads the United States Elections Project, an excellent source for voter turnout data: “Trump’s main strategy is attempting to make Biden appear unlikable, like Hillary Clinton. But that messaging is not getting through.” Trump can only win by somehow broadening his appeal in the last days of the campaign. He’d need to bring home millions of the graying Americans appalled by his management of the COVID crisis, and dampen the suburbanites’ surging enthusiasm for Biden. It’s unclear that claiming the former VP would undermine Social Security and Medicare, or his law and order message pledging to protect homeowners by quelling violence in the cities, get him anywhere near the giant gains in the popular vote needed for victory.

A major problem for Trump is that a big chunk of the electorate shunned both the Republican and Democratic standard-bearers last time because they found both of them intensely dislikable. Now, these onetime protesters are back supporting a major candidate because they deem one of the contenders likable, and they’re not suddenly warming to Donald Trump. In 2016, Trump won 63 million votes of the 137 million cast. His total was 3 million short of Clinton’s 66 million, and he trailed her share by 2.1%, 46.1% to 48.2%. The election was highly unusual, because 8 million, almost 6% of the total, went to third-party tickets, notably the Libertarian and Green candidates at 4.5 million and 1.5 million votes, respectively. That 6% is three times the 2% that usually goes to third-party tickets and write-ins and is the non–major-party share expected in this election.

“The 2016 Census reported that a high percentage of Americans didn’t want to vote for a major candidate, because they didn’t like them,” says McDonald. “So they took other options.” He notes that the James Comey letter drove home the narrative that Clinton was secretive and elitist. “It was one of these moments when people’s impressions crystallize,” he says. “But Biden’s approval numbers are much better than Clinton’s.” Trump is clearly trying to lure a big number of the 8 million rebel voters for whom the likability quotient is crucial by painting Biden as a bad guy. “The idea is to motivate people who like or don’t dislike Biden to hold their nose and vote for Trump,” says McDonald. That the strategy isn’t working is evidenced by Biden’s sturdy approval ratings, which in recent Gallup surveys exceeded Trump’s rating by 5 points.

It’s probable, though not certain, that turnout this year will far exceed that in 2016. Trump needs a scenario where sundry more ballots are cast, and where he gets a big proportion of the tens of millions of extra votes, to prevail. It looks now, however, like the lion’s share of the surge will go to Biden. The huge increase in early voting points to a big turnout. According to the Elections Project, 86 million Americans, two-thirds of the total number in 2016, have already voted either by mail or at polling sites. The volume in Texas already exceeds the 2016 total, and over 80% of the people who voted in North Carolina, Arizona, and Florida four years ago have cast their ballots. A Gallup survey from October finds that 69% of registered voters are “more enthusiastic than usual” about the race versus 50% in 2016.

Krishnamurty predicts that turnout could hit 155 million, an 18 million increase over the Trump-Clinton contest. He reckons that to win, Trump would need to get at least 47.5% of that vote to Biden’s 50.5%. (Don’t forget the roughly 2% forecast for mavericks and write-ins.) But just look at how many more ballots he’d need to collect. Of the 155 million votes cast, he’d need to win 73.6 million. That’s 10.6 million more than his 63 million in 2016. “Even at 73.6 million, I don’t think he’d win, but he’d have a shot,” says Krishnamurty. “I’ve been saying he wouldn’t win again from the day he was elected because he’s a divisive candidate, and successful Presidents are unifiers. He doubled down on the strategy of appealing to his base. He’s the greatest showman ever, but he went out and picked fights, and that’s what killed him with the seniors.”

It’s still conceivable that Trump can repeat the sorcery of 2016. But adding almost 11 million votes when he’s losing the 65 and over crowd and doing even more to rile suburban women? Trump needed a bridge to a much broader constituency that he never built, and now it’s a bridge too far.