Speaking as a Bernie Sanders voter (and not wanting to pull rank on anyone, but I've now voted for him four times in various elections), I'm obviously delighted that America's only socialist senator managed to pull off a win against all the odds in Michigan. The most important thing is that it keeps him in with some kind of slim chance of winning the Democratic nomination, but there's another aspect of the story that we shouldn't lose sight of. Have a look at the verdict FiveThirtyEight were offering just a few hours ago...
"According to our final polls-plus forecast, Hillary Clinton has a greater than 99% chance of winning the Michigan primary."
Even more embarrassingly, the site's live-blog was still talking about how boring the evening was and how Clinton was cruising to victory when over a third of precincts were already reporting and she was trailing by thousands of votes. Just what does it take for the alarm bells to start ringing?
It might be worth bearing this latest howler in mind the next time that Nate Silver or one of his fellow 'mystics' blunder into another country's referendum campaign, and ludicrously declare with over a year still to go that there is "virtually no chance" that one side or another is going to win. Once they can get their predictions right on home soil with just hours to go, perhaps they can branch out into the much trickier long-term foreign stuff, but they're clearly not there yet. OK, Silver can claim that he wasn't technically proved wrong about Scotland, but given what we know about how the indyref campaign panned out in the closing weeks and months, it's very difficult to see how he can credibly justify such an extreme statement with the benefit of hindsight.
Uncertainty may be scary to a stats guy, but that's the world we live in.