In Pennsylvania, the same two firms gave Lt. Gov. John Fetterman, the Democratic Senate candidate, a smaller lead over Mehmet Oz, and eventually showed Mr. Fetterman trailing — a departure from nonpartisan polls like Marist and Fox News. (He won by nearly five points.)
Worried that the G.O.P.-inflected polls were wrong and were liable to persuade Democratic grass-roots activists to give up rather than go out and knock on doors, Mr. Rosenberg used his podcast and his Twitter account to tell Democrats that their chances were better than they realized.
His bullishness earned him ribbing and ridicule. In an August article calling him “the most optimistic Dem online,” Politico noted that at times it seemed Mr. Rosenberg was pushing his relentlessly rosy view at “profound reputational risk.”
By the fall, perceptions of a red wave were starting to affect strategies, pushing money toward mistakenly perceived trouble spots. Insiders in both parties, mindful of past errors, began to doubt their own internal polls.
“We were limited in how aggressive we could get, because there was a feeling we were about to get absolutely smoked,” said Tim Persico, the executive director of the Democratic Congressional Campaign Committee, which oversees House races. “It comes down to what level of risk you are willing to take if your data is standing alone in a world where everyone else is telling you the opposite.”
A case in point was Wisconsin.
The state Democratic chairman, Ben Wikler, was seeing private internal polling that showed his party’s Senate candidate, Lt. Gov. Mandela Barnes, within a point or two of Senator Ron Johnson, a right-wing Republican and purveyor of misinformation.
But red-hued polling began to show Mr. Johnson pulling away. Data for Progress put the incumbent up by five; Patriot Polling, run by the Pennsylvania high school students, had his lead at eight. By early October, RealClearPolitics was projecting that Mr. Johnson could win by as much as seven points.