Yardbarker
x
Can analytics help you win your NCAA tournament pool? Don't bet on it

Can analytics help you win your NCAA tournament pool? Don't bet on it

Back in the days before the internet became an all-consuming repository of humanity’s foibles, I decided to attempt a new method for filling out my NCAA Tournament bracket. I called my brother and had him put my infant nephew on the phone. I told my brother to hold up a newspaper and ask my nephew to point out, or coo at or drool on, whichever team he thought might win a game. Full disclosure: I did not win my pool that year. In fact, my nephew is now in college, and I still haven’t won an NCAA Tournament pool since I was a kid myself and chose Indiana to win it all in 1987 — mostly because I admired its candy-striped warmup pants and Bobby Knight’s Tasmanian devil combustibility. 

I’ve picked the tournament badly during years when I’ve watched college basketball games five nights a week; I’ve picked the tournament badly during years when I hardly watched college basketball all season. And this is what makes the tournament such an utterly fantastic event: No matter what you think you know, you don’t really know anything at all, and all the KenPoms in the world — brilliant as they may be at compressing numbers into a narrative —can only get you so far if you are not Biff Tannen

Of course, there are those who would like to convince me that this is changing. In the wake of the sports-analytics revolution, there are people (and companies) who traffic on the notion that they can provide a new and more effective method of winning a tournament pool. But there are limitations to these new methods, and what I wanted to know was: How severe are those limitations? Can you really win your pool just by employing some strategy or zeroing in on a single telling statistic? The answer, of course, at least at this moment, is, No f***ing way. 

There is now apparently a cottage industry of mathematics professors who moonlight by toying around with NCAA Tournament analytics, and one of them, DePaul’s Jeff Bergen, calculated that your chances of picking a perfect tournament bracket are less than one in nine quintillion — which is less than your chances of winning the lottery twice in a row by picking the same numbers. Even if you’re aware of, say, Zion Williamson’s ability to incinerate footwear, and even if you’re familiar enough with Gonzaga’s basketball program that you could point to its general location on a map, your chances are roughly one in 128 billion. 

So unless you stumble onto brilliance, the way some dude named Brad Binder did a few years back when he picked every round of 64 game correctly, you’re still essentially engaging in educated guesswork. And this, says University of Illinois computer science professor Sheldon Jacobson — yet another member of the NCAA-tournament academia-industrial-complex — is where analytics can make a difference.

Jacobson runs a website called Bracketodds, which allows you to generate a computer-aided bracket for both the NCAA men’s and women’s tournaments based on the historical performance of the seeds. His formula also predicts upsets twice as often as picking at random would; last year, it prophesied No. 13 seed Buffalo’s upset of No. 4 seed Arizona. But for the most part, even Jacobson admits that what we can know is far less than what we’ll never know; it’s the difference between forecasting the weather and predicting it. “Patterns in the data suggest that certain upsets will occur, but picking the specific upsets is like finding a needle in a haystack,” Jacobson tells me in an email. “However, with AI methods, we are getting better in narrowing the haystack down.” 

For now, Jacobson suggests picking your bracket based as much on those seeding patterns as on the teams themselves, particularly through the Sweet Sixteen — which feels like a more advanced version of the ubiquitous “pick a 12 seed to beat a five seed” mantra that I grew up with. Maybe, as analytics expert Ed Feng suggests, you also just need to join a smaller pool. But you’re still subject to the capriciousness of a wildly unpredictable event. 

Even amid this revolution in data and analytics, last year’s upset-prone tournament led to a 10-point drop in the average accuracy of a bracket.  So that was my second question to Jacobson: If we’re essentially all thumbing through the same haystack in search of a glint of metal, what fun is it to simply generate a bracket via an artificial intelligence engine? “The 'fun' issue is a matter of taste,” Jacobson tells me. “Analytics transforms a sporting event into a math and stats hoopla. For some, that’s more exciting than the games themselves.” Those people are, of course, completely insane. Which is why I’ll be picking my bracket this year based entirely on the ranking of each school’s philosophy department, just to spite them.

More must-reads:

Customize Your Newsletter

+

Get the latest news and rumors, customized to your favorite sports and teams. Emailed daily. Always free!

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.