Odds are your NCAA bracket has already been busted, but that's half the fun

Did your NCAA March Madness bracket get busted this year? If so, take solace: You're not alone. It's nearly impossible to fill out a perfect bracket.

Last year an estimated 36 million people filled out March Madness brackets, according to the American Gaming Association. And every year millions watch those brackets go up in smoke in the first round by some unpredictable upset.

Saint Peter's busted mine this year by knocking off National Champion favorites Kentucky on their way to a historic run to the Elite Eight. And I wasn't the only one. According to SB Nation, 95.6% of all ESPN brackets had Kentucky advancing to the second round of the tournament.

If, like me, none of your teams are left standing for this weekend's Final Four, don't worry. You're not alone.

Is it even possible to pick a perfect bracket?

According to the NCAA, a perfect bracket has never verifiably been picked. The longest streak happened in 2019 when an Ohio man correctly predicted the first 49 games of the tournament before having his bracket busted on the 50th game when Purdue beat Tennessee in the Sweet 16.

"To get a perfect bracket you need to basically predict 63 games. Right?" says Georgia Tech professor Dr. Joel Sokol, who researches sports analytics including predictive modeling and ranking for the NCAA tournament.

"It's not quite as hard as predicting flipping a coin 63 times in a row because some of those [games] are going to be more likely than others," Sokol tells NPR. "But still, the odds are astronomical against it."

How astronomical, you may ask?

If you were to just flip a coin to determine your bracket selections, the odds of getting a perfect bracket are 1 in 9,223,372,036,854,775,808. That's one in 9.2 quintillion for those of us who don't know our big numbers.

To help wrap your head around how ridiculous that is, there are 31.5 million seconds in each year. 9.2 quintillion seconds would be 292 billion years.

So, pretty ridiculous.

But don't worry sports fans. If you know a little something about basketball, your odds can increase to 1 in 120.2 billion.

Can we predict upsets?

Upsets and underdogs are two of the reasons people love sports in the first place. Nothing is ever certain. But that doesn't mean numbers can't help us understand the probability of tournament outcomes.

Since 2003, Sokol, the Georgia Tech professor, has worked with his colleague Dr. Paul Kvam to put together a ranking method called LRMC.

The inspiration for their method came in 2002.

With five seconds left in the Peach Bowl Classic between Georgia Tech and Tennessee, Georgia Tech had just taken a 69-67 lead over the Vols when Tennessee's Jon Higgins made a desperation 3-point buzzer-beater shot. Miraculously it went in and Georgia Tech lost the game 70-69.

"At the end of the year, a lot of pundits were saying 'If Georgia Tech had just one more win, they would have made the NCAA tournament.' And so, I thought back to this half-court shot going in, and it didn't really make sense that this really rare half-court shot could make the difference between Georgia Tech being good enough or not being good enough," Sokol says. "There must be a better way to do it."

So, the two professors created LRMC, which stands for Logistic Regression/Markov Chain. (It's named after the two primary mathematical techniques used in their system.) They use basic scoreboard data such as which two teams played, whose court they played on and the margin of victory to help rank college basketball teams.

And their system has worked pretty well!

The year after Georgia Tech missed the tournament Sokol released a bracket showing the school making it to the Final Four, which turned some heads.

"It was kind of tough to be at Georgia Tech and say, we have this mathematical system that we don't influence at all. But, hey, it happens to be saying Georgia Tech is going to get to the Final Four! And then they made it," Sokol says. "We were obviously really happy as Georgia Tech professors and we were also happy professionally that it kind of vindicated that pick."

Since then they've tracked nearly 80 different methods for ranking teams, with the NCAA eventually asking them and a few others around the country to help provide rankings for the selection committee.

"Once we started doing that, then I felt like I probably shouldn't be in any brackets, even if there is no money involved," Sokol says.

That doesn't mean that their predictions haven't helped others make some money.

"I've gotten a couple of emails from people over the years, things like, 'Hey, your model helped me win this bracket when I know nothing about college basketball. So, what's your favorite charity? I want to donate a portion of my winnings,' that kind of thing. Which has been really cool," Sokol says.

Even math can't help you predict every bracket buster

Nearly 20 years of experience predicting March Madness results didn't help Sokol and his colleagues this year. Not only did they have Kentucky in the Final Four, LRMC predicted No. 1 seed Gonzaga winning the whole thing.

That was before Gonzaga got knocked off last week.

"I think one of the big takeaways that I've gotten personally from so many years of tracking the NCAA tournament is that there is just a lot of randomness in sports, and that's partly why people like it. I think it's not predictable," Sokol says. "You can do all the math you want and over the long run it might be really helpful, but in any given year you just kind of have to roll with it and sometimes things go your way. Sometimes they don't."

He says not to feel too bad when things don't go your way. Even the experts can't get it exactly right. Sokol points out that the selection committee couldn't predict three of the Final Four teams, with Kansas the only one-seed left in the tournament.

And while we may never see a perfect bracket, filling one out can make watching the tournament that much more enjoyable.

"You never know what's going to happen," Sokol says.

Copyright 2022 NPR. To see more, visit https://www.npr.org.