If Nov. 3 is the final buzzer, then national polls regarding voter preferences in the 2020 presidential election race are the closest thing we have to a scoreboard.

The numbers, at the moment, favor Joe Biden over President Donald Trump. According to RealClearPolitics polling, Biden holds leads in Pennsylvania, Michigan, Wisconsin, North Carolina and Arizona — all key battleground states.

National polls and polling averages have Biden ahead by a sizeable margin. The RealClear Politics average has Biden ahead by 7.7 points nationally; NBC News’ polling average has Biden leading by a margin of 51.7% to 43.9%. FiveThirtyEight’s polling tracker, which aggregates national polls and weighs them depending on reliability and methodology, gives Biden an 89% chance of winning November’s election.

But polls are an imperfect measure, with whoever is leading come Nov. 3 not guaranteed to see their advantage turn into actual votes. The 2016 election between Trump and Hillary Clinton showed as much; Clinton was favored by a vast majority of national polls before losing the Electoral College by a margin of 306 to 232 votes. That predictive misstep has informed the buildup to this year’s election. Just what can we learn from the polls this time around — and where might they mislead?

How polls work

In a recent video segment explaining poll accuracy in the context of the 2016 and 2020 elections, FiveThirtyEight database journalist Dhrumil Mehta said, “A poll is not a crystal ball that tells you who’s going to win the election, but rather a survey taken of a sample of voters that lets us guess at how a population as a whole might vote.”

Despite the ubiquity of the phrase “the polls,” American political polling is not a monolith. Rather, it is a patchwork of various organizations using various methods. According to Pew Research Center, surveys are conducted via live phone calls, online panels, and some combinations thereof. The lion’s share — more than 80% — are now conducted via online opt-in polling, which introduces an element of “bogus respondents” — people answering not to reflect their own opinions but to skew the results.

Pew also notes that technology has given rise to largely internet-based polls that wouldn’t have existed even a short time ago, and that can have flaws more established and practiced polls do not. “In 2016,” Courtney Kennedy wrote, “this contributed to a state polling landscape overrun with fast and cheap polls, most of which made a preventable mistake: failing to correct for an overrepresentation of college-educated voters, who leaned heavily toward Hillary Clinton.”

Mehta noted that state-level polls have higher rates of error than national polls — an important distinction, given that we elect presidents not by national popular vote but by the Electoral College. That difference was crucial in the last election. “While the state-level polls were off by about 3.4 percentage points on average in 2012, they were off by about 5.4 percentage points in 2016.”

Today’s polls, by and large, give Biden a better chance even than they gave Clinton in 2016. But the lessons of the last election have been heeded, by parties across the political spectrum. Biden supporters are hesitant to declare optimism, Trump supporters hold out hope, and politicos and pollsters themselves warn against repeating past mistakes —treating projection as assurance.

Understanding uncertainty

“Polls are the worst form of prediction ever, except for all the other kinds,” statistician and FiveThirtyEight founder Nate Silver told Katie Couric in an interview last month, reshaping a Winston Churchill quote. Asked where pollsters got things wrong in 2016, Silver argued that it was the public’s misunderstanding of what polls do — what information they represent —that led to confusion. “Polls were wrong but by kind of a normal amount, and you expect polls to be off to some degree,” he said, “and the fact that people couldn’t see that ahead of time reflects either their preconceptions about the race or their misunderstanding about polls and shouldn’t have been that much of a surprise.”

Nate Silver holds his phone as he sits on the stairs with his laptop computer at a hotel in Chicago on Friday, Nov. 9, 2012. The 34-year-old statistician, unabashed numbers geek, author and creator of the much-read FiveThirtyEight blog at The New York Times, correctly predicted the presidential winner in all 50 states, and almost all the Senate races. (AP Photo/Nam Y. Huh) | AP

Although FiveThirtyEight itself predicted a Clinton win in 2016, the 29% chance it gave Trump was markedly higher than the chances other polls gave him. In a postelection recap, Silver argued that people misunderstood the margin of error inherent to polling — and the possibility of a Trump victory — in some cases motivated by their own political leanings. “There was a modest polling error, well in line with historical polling errors,” Silver wrote, “but even a modest error was enough to provide for plenty of paths to victory for Trump. We think people should have been better prepared for it.”

Despite longer odds than he faced at this juncture in 2016, Trump retains a path to victory, centering on keeping hold of a portion of the Midwestern states he won in 2016 and protecting Arizona and Florida. “In 2016, his chances of winning the election were those of drawing an inside straight in poker,” Whit Ayres, a Republican pollster, told The Associated Press on Friday. “The question this year is whether he can draw an inside straight two hands in a row.” Robert Cahaly, whose Trafalgar Group correctly predicted the outcome in 2016, is more bullish about the incumbent’s chances. “What we’ve noticed is that these polls are predominantly missing the hidden Trump vote,” he told Fox News, adding, “So if you’re not compensating for this ... you’re not going to get honest answers.”

The only certainty, less than two weeks before Election Day, is uncertainty. Despite their candidate’s healthy lead, Democrats feel “dread,” The New York Times reports, and Trump supporters keep their fingers crossed for a second surprise. If the public — and parts of the political world — still don’t grasp every nuance of polling procedures, they have taken one truth to heart: Votes, not projections, carry the day.