Late Wednesday night, President Trump tweeted a celebratory image.

Polling from Rasmussen Reports has Trump at 50 percent approval, the tweet notes, something Trump credits to his hard work. “Promises made; promises kept,” the banners behind him read.

This is not the first time that Trump has tweeted a number from an approval poll. In fact, he’s done it about two dozen times, both his overall approval and his approval among African Americans (generally the only demographic group he singles out for sharing). That frequency allows us to create a sort of Trump Approval Polling Trend chart, stringing together the numbers he’s shared.

So, let’s.

One thing you will notice is that the trend is ... mostly flat. Trump’s tweet about his 50 percent approval is at least the seventh time he has celebrated a 50 percent approval rating, which is a bit like Apple repeatedly sending out news releases touting its stock hitting $170. At some point, a canny observer will note that either nothing much is ever changing or that, between those news releases, the stock is seeing some drops.

Before we figure out which is the case for Trump, we should note that at least three of the times he’s tweeted poll numbers, those numbers have been incorrect, estimated or nonexistent.

There was his tweet noting that his approval was “approximately” the same as Barack Obama’s at the same point in his first term; Trump’s approval in the cited poll from Rasmussen was two points lower.

There was the tweet declaring that his approval was “around 50 percent” in Rasmussen “and other” polling. Rasmussen had him at 47 percent.

And then there was the time in August when he claimed to have an approval rating of 52 percent. It appears he may accidentally have highlighted his disapproval from a Wall Street Journal-NBC News poll.

With those corrections, our chart looks like this.

Back to the other question then. Is this Trump’s approval staying flat or Trump ignoring downward fluctuations?

You’ll have noticed that Trump frequently cites Rasmussen polls. In fact, nearly all the polls he’s highlighted on Twitter come from that pollster. It seems as though Trump often skims the cream off the top of Rasmussen’s polls; the ones he picks out to highlight are usually from new peaks in Rasmussen’s polls or when he passes 50 percent.

When Rasmussen had his approval hitting and then passing 51 percent in early April, it spurred tweets from Trump. Shortly afterward, we wrote an article noting that Rasmussen’s results might be taken with a grain of salt.

Why? Mostly because Rasmussen sees results that are almost always much friendlier to Trump than other polls. If we compare Rasmussen’s results to the RealClearPolitics polling average — an average that itself includes Rasmussen’s results! — you can see that the results from Trump’s favorite pollster are almost always higher or much higher than the average of all other polling.

Since Trump’s inauguration, Rasmussen’s results have been higher than the RCP average 99.4 percent of the time.

Just because a pollster’s figures don’t agree with other polling doesn’t mean that the polls are wrong. They could be tracking a trend that others aren’t seeing. Happily, we can evaluate Rasmussen’s accuracy against a verifiable recent benchmark: the midterm elections.

Democrats won 53.4 percent of the national House vote as of writing to the Republicans' 44.9 percent, a gap of 8.5 points. We can compare that with generic ballot polling, a poll question that asks voters which party they prefer in a generic House contest. The final RealClearPolitics average of polls gave the Democrats a 7.3-point advantage, meaning that the average was off by 1.2 points.

Rasmussen’s final poll had the Republicans with an advantage in House contests of one point. Rasmussen missed the result by 9.5 points.

Minutes after this article first published, Trump had a new tweet.

On Thursday morning, Rasmussen updated its daily approval tracking numbers. According to their “just reported” data, Trump is back down to 49 percent.