The Republican presidential nominee had recently avoided the subject, given that many preference polls showed him in a close race with Hillary Clinton.

But on Monday, Trump cited polls following the Republican National Convention that show “the biggest bounce that anyone can remember.”

Yet while all preference polls taken months before the November vote should come with a grain of salt, those published around the conventions are often especially volatile. They require a shaker.

And, by the way, those national polls don’t reflect the way America chooses its president. The White House isn’t decided by a popular vote, but in state-by-state votes that make up the Electoral College.

Here’s a deeper look at post-convention polls:

Trump’s Bump

There have been two major polls conducted since the convention, a four-day gathering in Cleveland that ended with Trump’s acceptance speech Thursday.

The first was conducted Friday through Sunday and found Trump and Clinton in a near-tie. That matched other surveys conducted ahead of the convention that showed the candidates in a tight race.

In his remarks Monday in Virginia, Trump was likely referring to another poll conducted over the same three-day period that showed him up a few points on Clinton, but still within the survey’s margin of error.

The Biggest Bounce

Polling bumps are often the result of intense media coverage of conventions, when voters hear directly from the candidates and their closest allies with very little competition.

Even when there are missteps, such as Melania Trump’s issues with plagiarism last week, voters generally respond well to the weeklong speaking lineup that’s designed to paint candidates in a positive light, disparage rivals and energize the party’s most passionate voters.

Mitt Romney saw very little bump after his convention four years ago, but he was the exception. Since 1996, Republican presidential candidates saw polling increases of between 1.5 and 6.5 points on average, according to an analysis by Tom Holbrook, a political scientist at the University of Wisconsin-Milwaukee.

A Bounce for Clinton?

Al Gore’s bump in 2000 was more than 7 points, but such a spike isn’t guaranteed. President Barack Obama’s numbers jumped less than 2 points after the last two conventions, according to Holbrook’s analysis. Just like Romney four years ago, John Kerry’s numbers were virtually flat after his 2004 convention.

An abundance of historical evidence, however, suggests that Clinton is likely to see a bump by this time next week that could negate Trump’s appearance of momentum.

How Much Should You Read Into Polling Right Now?

Polls conducted during and just after conventions are particularly subject to change. There’s no way of predicting whether a bounce will last after the post-convention glow wears off.

A large bump doesn’t necessarily translate into victory, either. Gore ended up losing despite his 7-point bump in 2000. Obama’s bump was smaller than John McCain’s in 2008, yet Obama won.

Horse race polls are snapshots that can, and often do, change before Election Day. That’s especially true this year, when both conventions are being held about a month earlier than usual.

It’s also important to remember that presidential elections are not decided by a national popular vote. They are decided by the Electoral College system, essentially a series of simultaneous state-by-state votes.

That gives most of the influence to a handful of swing states, such as Ohio and Pennsylvania — which, not coincidentally, are the sites of the Republican and Democratic national conventions.