When Donald Trump won the presidential election in 2016, refuting the forecasts of virtually all the pollsters, I promised never to trust polls again. But I interviewed several leading pollsters recently to ask why we should trust them this time, and they offered several good reasons not to dismiss their findings.

They say they have learned from their 2016 mistakes and are pretty confident in their forecasts for the 2020 race.

Right now, virtually all polls — including the one by Fox News, the biggest pro-Trump propaganda outlet — show Democrat Joe Biden with a substantial lead. The Oct. 11 ABC News/Washington Post poll showed Biden beating President Trump by 12 percentage points, 55% to 43%, among likely voters nationwide.

Recent CNN/SSRS and Fox News polls also showed Biden leading by at least 10% of the vote nationwide, while an average of polls by the FiveThirtyEight website gives Biden a 10.6% advantage. But at this time before the 2016 elections, most of these same polls led us to believe that Hillary Clinton would be the next president, and she lost.

Pollsters cite five major reasons why they think that this time their public opinion surveys will be more accurate.

First, they say, it’s wrong to say that polls were wrong in 2016: The average of polls at the national level said Clinton would win the popular vote, and she did — by 2.9 million votes, or by 2.1 percent. This was pretty close to the almost 3% that polls had predicted.

What the polls got wrong were the results in three critical states — Michigan, Wisconsin and Pennsylvania — pollsters say. Trump won those states by a razor-thin edge that was within the margin of error. That allowed Trump to win the Electoral College and become president.

Second, pollsters say their 2016 polls had overstated support for Clinton because their polling samplings did not include enough likely voters without college degrees, who turned out in much larger numbers than in previous elections. Most of them voted for Trump, they say.

Scott Keeter, polling director of the Pew Research Center, told me that, “A lot of pollsters have made changes in their methodologies to try to offset some of the problems they had in 2016,” such as not polling enough white voters without college degrees.

Most pollsters corrected that mistake in the 2018 midterm elections, which turned out to be much more accurate, Keeter said. “With those changes, we have greater confidence that what we’re seeing in the polls today is a genuine picture of reality,” he told me.

Third, in 2016, pollsters failed to do last-minute surveys in Michigan, Wisconsin and Pennsylvania, where previously undecided voters helped Trump win on Election Day. This time, there are fewer undecided voters, and pollsters will conduct their surveys until the very last minute, they say.

“In 2016, at this particular time, a few weeks before the election, we still had about 20% of the electorate that was undecided,” Patrick Murray, director of the Monmouth University Polling Institute, told me. “That number is now below 10%.”

Fourth, Biden’s lead at this time of the race is larger than that of any challenger since the 1936 elections, pollsters say. Biden is the first opposition candidate in the past 21 elections with more than 50% of likely voters, they say.

Fifth, millions of Americans have already voted by mail or plan to vote early to avoid long lines on Election Day because of the COVID-19 pandemic. That minimizes the chances of a late swing vote that could dramatically alter the vote in the final hours before the elections, like what happened in 2016.

I’m still not ready to bet big money that Trump will lose. A lot of things can happen in the last two weeks before an election. But if the polls continue to show Biden with his current lead a few days before Nov. 3, he’s very likely to win — perhaps, even by a landslide.

Andres Oppenheimer is a Latin America correspondent for the Miami Herald.