“My feeling is Donald Trump. And I suspect that this applies to many concerned Democrats.”

So writes statistician Nate Silver in the New York Timesadding that his feelings – and yours – cannot be trusted.

The Silver Bulletin writer, FiveThirtyEight founder and former baseball analyst became famous for analyzing quantitative data and statistics of politics, and making election predictions based on probabilistic models that broke down weighted averages of polls.

But at least this year he threw in the towel when he made a strong appeal, saying that former President Donald Trump and Vice President Kamala Harris are in a real battle, with the polls putting them neck and neck .

“You have to come to terms with the fact that a 50-50 prediction really means 50-50,” he wrote. “And you have to be open to the possibility that those predictions are wrong, and that could be the case toward both Mr. Trump and Ms. Harris.”

Silver said that with both candidates within a point or two of each other in the seven battleground states likely to determine the election, a toss-up is “the only responsible prediction.”

Silver modeling of the 2024 race currently reflects that, noting that “we honestly don’t know” who will win.

Silver, in his Times oped introduced a seemingly counterintuitive idea: “It is surprisingly likely that the election will not be a photo finish.”

To that end, he noted that either Trump — whose supporters often have low civic engagement and are harder to track in polls — or Harris — who could benefit from pollsters unknowingly weighing in Trump’s favor — could easily win the election, by aiming for the top. edge of margins of error or beyond.

In fact, he said his model shows a 60 percent chance that one of them will win the Electoral College votes of six of the seven battleground states.

Justin Grimmer, a political scientist at Stanford, raised concerns about election forecasters using probabilistic models in an election campaign interview with the Harvard Graduate School of Arts and Sciences’ Colloquy podcast earlier this month.

“The clearest failure of these predictions is 2016, when the predictions coming out of these models were very confident to moderately confident that Clinton would win,” he said, noting that Silver’s model gave Trump a 28.6 percent gave a significantly greater chance. of winning than other predictors. “Certainly, the consensus view across the models was a Clinton victory. And that didn’t happen. And so they got that election fundamentally wrong. When we think about harm, it’s interesting to think about how these models are consumed by the public.”

Grimmer said many news organizations turned to forecasts after 2008 and 2012, when Silver correctly predicted Barack Obama’s election victories in most states, realizing it could fuel the need for content.

“It turns out it’s a nice summary of what could happen,” he said. “And of course it produces many stories. If the chance of a candidate winning an election goes from 55 to 50 percent, that might be a whole cycle of stories about what happened, why that chance changed and what might have caused it.”