The Washington PostDemocracy Dies in Darkness

Opinion We’ve become obsessed with statistical probabilities


Richard Zoglin is a New York-based writer and critic.

Is there a more unlikely star in the TV metaverse than Steve Kornacki? The NBC stats guru became a cult hero with his hyperkinetic performances on election night: sleeves rolled up, leaning intently into the big board, punching up the county-by-county vote totals like Mr. Wizard on steroids, turning dry election-night vote counting into edge-of-seat drama.

More recently he’s taken his act to NBC Sports’ “Sunday Night Football”: hovering over another big board, parsing the probability of each NFL contender making the playoffs. On the next-to-last weekend of the regular season, for instance, he pegged the chances at 83 percent for the Indianapolis Colts; a paltry 5 percent for the Pittsburgh Steelers.

But did those numbers really offer any useful information, beyond what an ordinary NFL fan could have gleaned from examining the teams’ records and remaining games? In fact, Pittsburgh, despite its meager chances, made it into the playoffs (before losing to the Kansas City Chiefs on Sunday), and highly favored Indianapolis lost out. Go figure. But never mind the results; not just in sports but across the culture, we’ve become obsessed with statistical probabilities — trying to reduce uncertainty about the future with dubiously precise numbers.

We depend on probabilities, of course, to guide much of our lives. With a 70 percent chance of rain forecast, we’ll probably grab an umbrella on the way out. Economic forecasts help us decide what to do with our savings. Cancer patients look to projected survival rates when deciding what treatments to explore.

But the exactitude of these probabilities, typically based on computer modeling or past results projected onto an uncertain future, conveys an authority that they often don’t deserve.

The battle against the coronavirus pandemic, for example, has been a veritable pageant of probabilities. The Pfizer and Moderna vaccines promise a 90 to 95 percent chance of protecting you against the coronavirus (so most people opt for them instead of the about 70 percent effective Johnson & Johnson shot). Unvaccinated people, the Centers for Disease Control and Prevention reports, are six times more likely than the vaccinated to get sick with covid-19, nine times more likely to be hospitalized and 14 times more likely to die from the disease.

But the anti-vaxxers marshal their own stats: One British study, picked up by the right-wing press, found that triple-vaccinated people were 4.5 times more likely to test positive for the omicron variant than the unvaccinated. Try sorting all that out.

And if the projections turn out to be a little off, they can simply be adjusted in real time. Followers of the political website FiveThirtyEight won’t soon forget the incredibly shrinking probabilities of a Hillary Clinton victory on election night 2016. She started the evening with a comfortable 72 percent chance of beating Donald Trump; the number dropped to 55 percent at 10:10 p.m. Eastern time (when Michigan was called for Trump); went underwater, to 44 percent, at 10:23 (when Ohio was lost); and had plummeted to 6 percent after 2 a.m.

At least they got it right in the end.

The stats-crazy world of sports has become particularly adept at these rolling, real-time projections. Check online for the score of any NBA or Major League Baseball game in progress, and you’ll see the win probabilities bouncing up and down with every three-point play or two-run homer. (The Golden State Warriors lead by 5 points in the first quarter of their Christmas Day game against the Phoenix Suns: Their chances of winning are 54.4 percent. The Suns take the lead by 6 in the second: Now they’re 73.1 percent favorites.)

The NFL’s ultrasophisticated Next Gen Stats can even take big plays and calculate, in retrospect, their probability of success — as in a 14.2 percent chance that Christian McCaffrey can elude four defenders and score on a run from the 7-yard line. It makes the feats on the field seem more marvelous than ever: Look at the odds they beat!

But these suspiciously precise projections should be regarded with caution. In politics, for instance, we’ve learned to be skeptical of pre-election polls, given how wrong they’ve been in recent years. But when a lot of them are digested, weighted and converted into a precise probability figure, down to the 10th of a decimal point, they acquire a bogus air of certitude. Those early projections about vaccine effectiveness, moreover, may well have contributed to complacency about the pandemic — before the new variants emerged and essentially threw the numbers out the window.

As for those rolling adjustments when the probabilities butt up against the real world — well, we can all play that game. According to my own projections (statistical methodology available on request), anyone who started reading this article had a 38 percent chance of making it to the last paragraph. By now, the probability is well up in the 90s.

And if you’ve made it this far, it’s 100 percent. Beat that, Kornacki.