A poll released by the Trafalgar Group on Monday shows President Donald Trump leads presumptive Democratic nominee Joe Biden by one point, 46 percent to 45 percent, in the key battleground state of Wisconsin.

The poll of 1,021 likely 2020 general election voters was conducted between June 23 and June 25. Trump’s one-point lead over Biden is within the poll’s 3 percent margin of error, so the results reflect a statistical tie between the two.

The results are a significant departure from a slew of recent polls showing Biden leading battleground states Trump won in 2016, including Wisconsin. The most recent polls in Wisconsin prior to the release of Monday’s Trafalgar Group poll have shown Biden with a significant lead, as the Real Clear Politics Average of Polls for the state indicates.

A Marquette University Poll conducted between June 14 and June 18 gave Biden an eight-point lead. A CNBC/Change Research Poll conducted between June 12 and June 14 gave Biden a four-point lead. A New York Times/Siena University Poll conducted between June 8 and June 15 gave Biden an 11-point lead. A Fox News poll conducted between May 30 and June 2 gave Biden a nine-point lead.

The methodology of Monday’s Trafalgar Group Poll differs from those four more recent polls that show Biden in the lead in a number of ways.

First, the Trafalgar Group has a larger sample size and lower margin of error than any of the other recent polls.

The Trafalgar Group’s sample size of 1,021 likely general election voters has a margin of error of 3 percent, while the Marquette University Poll had a sample size of 805 and a margin of error of 4.3 percent, the New York Times/Siena University poll had a sample size of 655 and a margin of error of 4.3 percent, the CNBC Change Research Poll had a sample size of 231 and a margin of error so great as to not even be included in the poll’s report, and the Fox News Poll had a sample size of 801 and a margin of error of 3.5 percent.

Second, and of equal importance, among the polls that had a sample size large enough to identify a margin of error, only the Trafalgar Group Poll focused on likely general election voters, the most reliable indicator of how people will actually vote. The Marquette University Poll, the New York Times/Siena University Poll, and the Fox News Poll all sampled registered voters only, which means the results included some respondents who are registered to vote but not likely to do so.

The Trafalgar Group’s record of battleground state polls in 2016 was quite good, as Politico reported one week after the election:

Few, it seems, paid attention to the surveys from the Trafalgar Group – a Georgia-based consulting firm that, on its website, celebrates the time RealClearPolitics picked up one of its Florida primary polls – showing Trump ahead. The group’s Pennsylvania poll was the only one of dozens since late July to show the GOP nominee in the lead there – but it was also the only poll conducted into this past weekend, as voters made their final choices.

The Trafalgar Group was somewhat prolific on Monday, the day before the election, releasing surveys in Florida (Trump ahead by 4 points), Michigan (Trump ahead by 2 points) and Georgia (Trump ahead by 7 points).

Nate Silver’s 538.com currently rates the Trafalgar Group’s polling with a grade of C-.  On the eve of the 2016 presidential election, 538.com gave Hillary Clinton a 71 percent chance of winning, with an 83 percent chance of winning Wisconsin, and projected she would win 302 electoral college votes to Donald Trump’s 235. On election night, however, Donald Trump won 306 electoral college votes, including Wisconsin’s ten electoral college votes, with Hillary Clinton winning only 232. Trump defeated Clinton by a margin of fewer than 23,000 votes out of 2.8 million cast in the state.