One of my favorite metrics for evaluating the 2020 race has been considering the preference of voters who view both major-party candidates unfavorably. In 2016, that group was decisive: Constituting a fifth of the electorate, it backed Donald Trump by a 17-point margin. The margins by which Trump won dislike-both voters in Wisconsin, Pennsylvania and Michigan was enough to make the difference in those states, the ones that handed him the presidency.

Over the course of the past year, then, I’ve been looking for data on that group. When I did, the pattern was consistent. Among those who viewed both President Trump and former vice president Joe Biden unfavorably in Quinnipiac University’s polling, for example, Biden led by 33 points last December. In April, it was 32 points. In July, 27. Other pollsters found something similar; Monmouth University had him up more than 30 with that group in August.

Then, Quinnipiac’s data stopped being useful. I’d ask their always-helpful team for the breakdown among voters who viewed both Biden and Trump unfavorably only to learn, time and again, that there weren’t enough people in that category to be statistically significant.

Why? It’s hard to say, but I have a guess: Too many people liked Biden.

For the first quarter of the year, both Biden and Trump had about equal favorable ratings, according to RealClearPolitics’ average of polls. Through the spring, the two generally moved in tandem, hanging out just south of 45 percent. But then Biden started to pull away. By now, he’s consistently viewed favorably by half of poll respondents while Trump remains mired in the low 40s.

Why does this matter? Because there’s a link between a candidate’s advantage in favorability and the results of the presidential contest.

On Friday, Gallup released its final assessment of the state of the 2020 race, giving Biden a narrow advantage on the question of favorability. Among all adults, 49 percent view Biden favorably, compared with 45 percent who say the same of Trump. That’s a four-point advantage for Biden.

In 2000 and 2004, Democratic candidates Al Gore and John Kerry had 1-point advantages over George W. Bush. In 2008, Barack Obama was viewed more favorably than John McCain by a 12-point margin. You can see each of the past seven races on the graph below. We’ve also included the Biden-Trump margin indicated with a horizontal dashed line and FiveThirtyEight’s national polling margin between Biden and Trump (which currently has Biden up 8.9 points) indicated with a vertical dashed line. Where those lines intersect is, theoretically, where this year’s bubble would show up.

Looking at that graph, it doesn’t look like there’s much correlation between favorability advantages and the results of the election. But, again, I said the correlation was among voters. If we use the favorability data for likely voters, the correlation is much clearer.

And, if you imagine a line roughly running through the bubbles on the chart, you can see why Biden enjoys the lead that he does.

“Biden’s 52% favorability rating is much better than Hillary Clinton’s 43% at the same point four years ago, and comparable to Obama’s in 2012,” Gallup’s Lydia Saad and Megan Brenan wrote. “Trump’s 43% favorable rating among likely voters is slightly better than his 39% in 2016, but otherwise the lowest for any Republican presidential candidate in Gallup records since 1992.”

It’s important to point out that national leads aren’t everything. Two of the three red bubbles, indicating races won by Republican candidates, sit to the right of the vertical dividing line: a reminder that the Democratic candidates won the national popular vote in 2000 and 2016 but not the White House.

During that 2016 contest, observers (including myself) noted that it was unusual in that both candidates were viewed unfavorably by the electorate. There was speculation that perhaps this was the new normal, that national campaigns would, from that point on, invariably involve candidates who got to the general election battered, bruised and disliked.

It turns out that 2016 may, in fact, have been an anomaly — that Trump won despite having a relatively large favorability deficit against Clinton because Clinton was viewed broadly unfavorably. Put Trump up against someone viewed more positively, you get a different result.

Like, say, Joe Biden. The other lesson from 2016 was never to count Trump out. But if he manages a win despite facing the largest favorability deficit in the past eight contests, we’ll once again have to rethink our understanding of national politics.