During a contentious news conference Wednesday following an election in which the Democratic Party won control of the House, President Trump was pressed both on his optimistic spin about the results and on the rhetoric that he and his party deployed in the past few weeks.

At one point, a reporter asked Trump if the election of two Muslim women to the House was a rebuke of the nationalist arguments that were a mainstay of his recent election rallies.

Trump interpreted the question as being about his commitment to diversity, noting that unemployment among black and Hispanic Americans had dropped during his administration. He then cited a poll showing how well-liked he was in those communities.

“A poll came out recently, where my numbers with Hispanics and with African Americans are the highest, the best they’ve ever been,” Trump said. “That took place two or three days ago, the poll. I have the best numbers with African American and Hispanic Americans than I’ve ever had before.”

A bit later, he referenced the poll again, while excoriating a PBS reporter for asking a “racist question” about whether his referring to himself as a nationalist emboldened racists.

“Why do I have my highest poll numbers ever with African Americans?” he countered. “Why do I have among the highest poll numbers with African Americans?”

Trump was referring to a poll from Rasmussen Reports that was released Oct. 29. It had Trump at 50 percent approval overall, including 40 percent approval from black Americans and 47 percent approval from other nonwhite people.

Trump pays close attention to polling, even if he doesn’t talk about polls as much as he did during the primaries, when he was usually well ahead of his competitors. He’d mentioned polling a few minutes before this exchange, in fact, taking an opportunity to excoriate CNN.

“I’ll give you voter suppression,” he said in response to a question about how the electorate might have been tamped down Tuesday. “I’ll give you voter suppression: Take a look at the CNN polls, how inaccurate they were. That’s voter suppression.”

He’s made the case before that polling is a tool for keeping people from turning out, though it’s not entirely clear how that works. Hillary Clinton was seen as the likely victor in 2016, but there’s no evidence that suppressed Republican turnout; if anything, it might have discouraged Democrats from voting. Regardless, Trump was deploying a very deliberate strategy, using CNN’s poll numbers — an objective bit of data — as a way of trying to show both that CNN was working against the Republicans and, at the same time, peddling incorrect information.

It is the nature of polling that the numbers won’t be perfectly accurate, of course. Polling, unlike many information sources, includes a margin of error in every presentation of results. CNN’s final congressional ballot poll, asking Americans if they preferred Republicans or Democrats in their local House races, had the Republicans trailing by 13 percentage points. As of this writing, the margin by which the Democrats lead in the popular vote is about four points — but that’s with tens of thousands of votes remaining to be counted, including in Democrat-friendly California, where ballots will be counted for weeks to come. (It was the slow tally of California votes that slowly drove Clinton’s popular vote higher in the days after the 2016 election.)

CNN also polled in other key races at about the same time that the Rasmussen Reports poll was released. The cable network had Sen. Bill Nelson (D-Fla.) winning among likely voters by two points and Andrew Gillum, the Democratic contender in Florida’s gubernatorial contest, leading by one point. As of this writing, each of those races is within one percentage point to the Republicans' benefit, which is within CNN’s margin of error.

Which brings us back to Rasmussen. Rasmussen’s results are consistently more favorable to the Republicans than the Democrats, as we’ve noted before. That Trump approval number of 50 percent, for example, was nearly six points higher than RealClearPolitics' average of polls on that same day.

But to Trump’s point, Rasmussen also released a generic-ballot poll on Nov. 5. It had the Republicans leading the Democrats by a point, with a 2 percent margin of error. Meaning that unless the outstanding vote is much friendlier to the Republicans than expected, the Rasmussen result was outside the margin of error and likely to get worse.

CNN was off, and that to Trump was proof that it was biased and working against the Republicans. Rasmussen was off and that was no obstacle to Trump celebrating a (questionable) number from its data.

This is at least the fourth time I’ve written an article that mentions Trump cherry-picking from polls.