James pulling the upset in Trump-friendly Michigan is a longer shot than Bob Menendez losing in Trump-hostile New Jersey. I wouldn’t wager on it even with someone else’s money. But I’m throwing this goofy poll at you anyway because:

1) It’s Election Day eve and goofy outliers are fun, like a suspiciously large present under the Christmas tree. What’s in that box? Probably it’s just a pair of socks placed in an oversized container as a gag gift. In fact, that’s definitely what it is. But there’s just enough doubt to create a bit of suspense.

2) I like James and liked him even more after this exchange with Rudy Giuliani. The man’s name is “John,” Rudy. And Rudy — or is it “Randy”? — will never forget it again.

3) This is at least the third bananas poll in the past 24 hours from the Democratic firm Change Research. They’re the ones that had Beto! suddenly tied with Ted Cruz at 49 in Texas and Matt Rosendale surging past Jon Tester in Montana — in both cases the only pollster to show the underdogs catching up at the last minute. Now they’ve done it again with James and Debbie Stabenow in Michigan.

If even one of these races comes out the way Change is betting, they’ll be instant legends. If they hit all three it’ll be the greatest day in the history of polling.

But they’re not going to hit all three. Here’s what Stabenow’s numbers look like in other surveys:

It’s not just the margin, it’s that she’s north of 50 percent in most recent polls. The two in which she’s not have her right on the cusp at 49. How does James deny her a majority? The only trace of credibility in Change Research’s numbers is that the last poll of Michigan before this one had Stabenow’s lead down to seven, a bit of evidence that the race is tightening. But as tight as two points? C’mon. FiveThirtyEight gives her a 96.7 percent chance of winning based on all available polling data and other factors. It’ll be a moral victory if the race isn’t called as soon as the polls close.

A point worth noting about all the Change Research outliers, though: They’re not all leaning in the same direction. If they were showing Republicans overperforming uniformly or Democrats overperforming uniformly, you might detect a “house effect” in which they systematically predicted larger turnout from one party or the other. As it is, they have a Democrat beating expectations in Texas and Republicans beating them in Michigan and Montana. It’s unlikely that a national climate that would place John James on the brink of a shocker in Michigan would also have Beto O’Rourke poised to stun Cruz in Texas, but hats off to ’em for having the courage of their convictions in publishing this data.

One more from Change, just to make things extra fun:

We haven’t paid much attention to Mississippi amid the clutter of all the other midterm races but it’s potentially yuuuuuugely important. If no one gets 50 percent there tomorrow night, there’ll be a runoff between the top two on November 27. Cindy Hyde-Smith, the Republican appointed to fill Thad Cochran’s seat, is favored to finish in the top two but thanks to populist Chris McDaniel she’s all but certain to fall short of outright victory via 50 percent tomorrow. That means she and Democrat Mike Espy will face off in three weeks — potentially with control of the U.S. Senate on the line if Democrats have a big night tomorrow and end up with 50 seats. To make things even stickier, there’s a chance that McDaniel will upset Hyde-Smith tomorrow by finishing ahead of her, guaranteeing a runoff between him and Espy instead. That would be dangerous for the GOP, as McDaniel has polled behind Espy in runoff surveys thus far while Hyde-Smith has polled mostly ahead. In theory the fate of the next Supreme Court nominee could come down to Mississippi’s willingness to grit its teeth and elect McDaniel with Democratic voters out in force to try to flip the Senate. And now here’s Change Research suggesting that McDaniel really might be about to bump Hyde-Smith out of the runoff tomorrow. Gulp.