President Trump will officially announce that he’s running for reelection at a massive rally in Orlando on Tuesday. The early line on his chances: He’s not the favorite, but he’s far from down and out.

The leak of unfavorable reelection numbers from Trump’s own internal polling, which show him behind Democratic candidate Joe Biden in many key swing states, has triggered a rush to judgment among many analysts. While carefully acknowledging things could change, they note that Trump’s job approval rating has been negative through most of his presidency and that he trails most prospective Democratic nominees in the key Midwestern swing states he needs to win. The conclusion is easy to draw: Trump’s probably a goner, despite the strong economy.

It’s true that Trump has been historically unpopular during his tenure. While other presidents have won reelection with similarly low job approval ratings, they all have been in positive territory at some point during their presidency. What’s more, the past two incumbents who ran for reelection, George W. Bush and Barack Obama, both received a share of the popular vote that was extremely similar to their job approval rating right before Election Day. Trump has never polled above 44 percent job approval for any extended stretch of his presidency. Unless that improves a lot, he’s not going to win the popular vote.

But as 2016 showed, he doesn’t have to win the popular vote to win an election. All he has to do is win 270 electoral votes, and with a coalition heavily tilted toward blue-collar white voters, he can do that with as little as 46 percent of the vote. He can hit that mark if he slightly improves his job approval and establishes a strong contrast with (read: run a very negative campaign against) his Democratic opponent.

People tend to forget that Trump won despite record-low favorable ratings because he decisively carried voters who did not like either candidate. The 2016 exit polls showed that 18 percent of all voters did not like either Trump or Hillary Clinton, and Trump carried them by a 47-30 margin. These “reluctant” Trump voters still exist, and Trump will do everything in his power to convince them to cast in with the devil they know one more time.

There’s also the issue of state-level polling error to consider. We know from exit polls that Trump tends to either be roughly as popular or somewhat more popular in the five key Midwestern battleground states — Iowa, Michigan, Ohio, Pennsylvania and Wisconsin — than he is nationally. However, pre-election state polls from these five often underreport his strength or that of other Republican candidates. In 2018, the final state polls for gubernatorial races in Iowa and Ohio severely underestimated the support for the ultimate Republican winners. It’s possible that some pollsters have systematic errors built into their sampling or weighting that pushes GOP support below its real level.

For example, the exit poll for the 2018 election in Wisconsin showed Trump with a 48 percent approval rating there, about four points higher than his national average at the time. It was thus no surprise that incumbent GOP Gov. Scott Walker received roughly 48 percent of the vote in a year where straight-ticket partisan voting based on opinions of Trump was the dominant feature. How believable is it, then, that the most recent Wisconsin poll, taken in March when his national job approval rating was about 42.5 percent, showed Trump with only a 41 percent job approval rating?

We should not overestimate Trump’s chances, either. Republicans might consider a Democratic nominee to hold extreme views, but swing voters might think that person is okay. Trump’s job approval rating remains too low for Trump to win on his own merits even if possible poll error is taken into account. The hot economy could cool markedly by next year, or one of our international confrontations could turn into an armed conflict or even a war. Trump’s prospects could fade just as easily as they could improve. But improve they could, especially once there’s another person to compare him with. Gallup poll data since Richard M. Nixon have shown that presidential incumbents seeking reelection tend to improve in the polls once the general election campaign begins in earnest.

Trump defied history in 2016, becoming the first person since the disputed election of 1876 to win the electoral college while losing the popular vote by two points or more. Don’t rule out him defying history again.

Read more: