select all

The Fake Twitchy Hell Dials of the New York Times’ Forecast Only Made Last Night Worse

Photo: rhyselsmore/Twitter

Around 9:30 last night, this tweet popped up on my timeline:

For those who didn’t spend last night glued to both their phones or laptops and the TV, the “hell dial” was the New York Times’ election-forecast dial, which was showing real-time projections about who would be the next president. It started the night with the dials pointing toward an 80 percent chance of a Hillary win. But slowly, and then all at once, it radically reversed itself, as the polls every forecast model uses were proven to be definitively wrong and the needle swung toward Trump. My Twitter timeline for a while was about 40 percent people tweeting screenshots of the dial over and over in agony.

But all the while, the dials kept jittering back and forth, showing small swings — Trump would gain a few electoral votes and then lose them, go from a 55 percent chance to win to 56 and then back down to 52. It was agonizing for supporters on both sides, but also kept you glued to the dials — the feeling of immediacy was both intoxicating and anxiety-inducing compared to the staid talking-head roundtables, on network and cable news, that might stop every 30 to 45 minutes to announce a few more states in between Wolf Blitzer wandering around aimlessly or Mark Shields unwrapping hard candy after hard candy. The feeling was probably best expressed in this GIF posted by Twitter user @rhyselsmore.

The problem? All that twitch wasn’t reflecting small adjustments to the forecast model as new votes came in; it was hard-coded into the dials themselves. The dials were designed to always be jittering around. People much smarter than I am quickly caught on when looking into the code for page. Alp Toker was quick to spot it:

While Lauren Ancona when a bit more in-depth about it:

None of this is to say the Times’ forecast model was fundamentally wrong — Nate Cohn and his team have walked through the math they use, and to my non-statistician’s mind it seems sound. And they did indeed start accurately predicting Trump’s victory long before many others.

Look: The dial faking the shake to keep things interesting is far, far from the biggest problem facing the U.S. right now. The fundamental divide between two halves of the country, the failure of the left to have any sort of answer for those racked by the loss of manufacturing jobs, the fact that Clinton will likely win the popular vote but lose the Electoral College, and the promise of at least four years of a presidency built on vilifying the Other are far, far more disturbing.

But this built-in twitchiness would be fine if it was, say, showing the odds of the Cubs winning the World Series or someone winning Best Actress at the Oscars. The TimesGregor Aisch responded, saying the ersatz unsteadiness of the dials was meant to represent the uncertainty of the model. But when it comes to Americans staring at the screen, wondering what exactly the world is going to look like at the end of the night, just keep the damn dial still until there’s actually something concrete to move it.

NYT’s Fake Twitchy Forecast Dials Only Made Last Night Worse