This tweet from President Trump is incorrect on at least two levels.

The first level is that his “Republicans are 5-0 in Congressional races this year,” which the media refuses to mention, is true only if 1) you don’t consider the Senate to be part of Congress (which it is) and 2) you skip the House race that Democrats won in California. Trump started talking about how the GOP was “4-and-O” (that being the letter O and not a zero) after the GOP held four House seats given up by people joining his administration and upped it to five after a special election in Utah last month. The only argument for not including the race in California would be that it was a Democratic seat held by a Democrat. But if that’s the standard that applies, Trump is 0 and 0, not 5 and 0.

Including all congressional races, the score (if you will) is 5-2. But before those contests, Republicans held six of the seats.

In all special elections, the Republican record is significantly worse. In state and local races, as tracked by DailyKos, Democrats have won 47 of 68 elections, picking up 13 seats and losing only one to the Republicans.

That excludes the legislative races in Virginia, as those weren’t special elections. There, the Democrats picked up more than a dozen seats from the Republicans, in addition to holding the governor’s mansion. Democrats also won the governor’s race in New Jersey.

So that’s the fuller picture of what happened in elections this year. It’s important to mention that not only because it’s important to note where the president is incorrect (willfully or otherwise) but also because Trump’s incorrect read on the year’s races also informs the other point on which his tweet was incorrect.

When Barack Obama won the presidency in 2008, an argument was made that he had tapped into a new coalition that could propel his party to victory over the short term. Young people, black and Hispanic voters, and women supported Obama heavily. If they kept turning out, the Democrats seemed a lock to hold Congress. Before the 2010 elections, Obama encouraged his base, the voters who loved him, to cast ballots for Democrats in House and Senate races. Far less of his base went to the polls in 2010, and the Democrats were blown out. When Obama was back on the ballot in 2012, his base turned out again — surprising pollsters and giving Obama a wider-than-expected margin. In 2014, though, they again stayed home.

Many Obama voters were voting for Obama, not a new electoral coalition and Democratic parties. (In the wake of the 2010 elections, some expressed frustration about what the Democrats had gotten done with single-party control of Washington.) When Obama wasn’t on the ballot — as he wasn’t in 2010, 2014 and 2016 — they stayed home and the party suffered.

Trump has one advantage in that regard: His base is also heavily Republican, the party that tends to turn out in greater numbers for midterm elections because its demographics correlate to the demographics of people who vote more regularly (older, wealthier). But there’s not much reason to think that Trump’s base will be motivated to turn out in November 2018 to pick at random a Republican in a toss-up race per Cook Political Report, Rep. Rod Blum (R-Iowa). Trump’s base loves Trump; it deliberately and loudly does not love Republicans in general. They came out to vote for him and probably would do so again in 2020, but 2018? Iffy bet.

A year is a long time. On Dec. 18, 2015, Hillary Clinton had a 6.6-point lead over Trump — who hadn’t yet won a single political contest in his life. Things change. But Democrats currently have a double-digit advantage in generic polling on the House next year, the biggest margin in NBC-Wall Street Journal polling since 2008. The president’s party almost always loses seats in midterm elections. Setting aside post-9/11 George W. Bush, the last president to gain seats in the first midterm after his election was Franklin D. Roosevelt.

The problem is that Trump may actually believe his hype. He may be falling into a familiar trap: assuming that everything good that happens in campaigns is because of him and everything bad that happens is the fault of someone else. If you set aside Ed Gillespie’s loss in Virginia and Roy Moore’s in Alabama as functions of those candidates, you’re not coincidentally labeling as outliers two of the counterexamples to your theory of political dominance. If Trump actually believes that the Republicans holding five heavily Republican seats prove that he is a political juggernaut, he may actually then believe that the Trump steamroller will roll through 2018 without any problem.

Believing that while also knowing that he endorsed two losing candidates in Alabama — Luther Strange in the primary and Moore in the general — is an impressive bit of cognitive dissonance.

There’s a third possible way in which Trump’s tweet is incorrect. He states that he said that Gillespie and Moore would lose, which they did. It’s true they lost — after Trump vociferously endorsed their candidacies. Now, it’s possible that Trump nonetheless told people privately that they would lose, either because he thought they might or to build in the sort of escape hatch that he used in this tweet.

Trump does that a lot: reframes what happened to present the interpretation of events that casts him in the most favorable light. If current indicators about 2018 hold, his tweet on Nov. 7, 2018 — the day after the election — should be a doozy.