Tauhid Zaman knows picking a perfect NCAA Tournament bracket is statistically impossible — or close to it.
Zaman’s goals are only slightly more realistic. The MIT professor and algorithmic sports betting enthusiast wants to finish first among the 17.3 million basketball fans who are expected to enter ESPN.com’s Men’s Basketball Tournament Challenge this week.
The Grand Prize of ESPN’s bracket challenge — a round trip to the Maui Jim Maui Invitational and sizable Amazon gift card retailing at approximately $16,950 — offers no incentive for Zaman, who holds three degrees from MIT and teaches in the university’s Sloan School of Management. Instead, he wants to prove his integer programing model is capable of taming the madness in March.
And, in theory, Zaman has the computer model to do just that. He shared the mathematical formulas used to build it during a presentation at the Sloan Sports Analytics Conference, held earlier this month in Boston. The annual gathering is a Woodstock for sports’ big thinkers; participants call it “Dorkapalooza.”
Zaman says his model is, by scientist standards, surprisingly simple. It’s effective, too: He’s already used one version of to win major prizes on DraftKings.com, a website that awards money to those who pick the best daily rosters in different professional sports.
And while the math and computer science used to build the model is far from simplistic, the theory behind it is quite straightforward.
Zaman uses a predictive rankings system such as the NCAA Elo Ratings published at FiveThirtyEight.com as a baseline. He then uses a diversity parameter to make sure each future bracket differs from the ones he’s already created. Then he purposely blows it up. (More on that later.)
Sometimes, simple works.
Zaman built a predictive model for the NCAA Tournament, which starts this week, then — because this is how scientists operate — he used it to see how he would’ve fared in the last two years.
The 2017 NCAA Tournament trial showed promise, as Zaman’s highest-scoring bracket finished in the top 50 in a field of 14.6 million entries.
Zaman was still 100 points off the winner, though, which is the difference between correctly picking an Elite Eight game (80 points) and round-of-32 contest (20 points). (Each round of the ESPN challenge has 320 possible points, evenly divided by the number of games in a round, for a max total of 1920).
It wasn’t until he used his model to run through the 2018 bracket, however, that Zaman knew he was on to something.
Last year’s tournament was the most chaotic one in recent memory. Arizona’s opening-round loss to 13th-seeded Buffalo was a mere footnote in a tournament where UMBC made history as the first No. 16 seed to win a first-round game, stunning top-seeded Virginia. Then there was 11th-seeded Loyola of Chicago, which beat ninth-seeded Kansas State in the Elite Eight and advanced to the Final Four.
Zaman’s best entry navigated past the busted brackets to score 1720, which would have taken home the grand prize.
“(Virginia’s loss) broke a lot of people,” he said. “I’m not going to get broken.”
That’s because Zaman’s system accounts for the chaos. His computer adds a certain amount of “noise” or “crap” — his words — to groups of 20 entries. The added “noise” means that Zaman’s was bound to predict, say, UMBC beating Virginia in at least a handful of his 1,000 entries.
The unpredictable nature of neutral-court college basketball, and the pressure of a one-and-done tournament is why the addition of “noise” ultimately created Zaman’s best-in-class bracket.
“It’s the principle – I want to win it to show I can win,” Zaman said. “I don’t do this for the money. I do it just to do it.”
Winning this year’s tournament challenge won’t be easy.
Zaman is crowd-sourcing what he calls the “MIT Madness Army” to help manually enter his 1,000 brackets into the ESPN contest site.
Victory is less than assured for Zaman and his army, as the madness of the tournament and the large player pool size makes the competition difficult for any contestant.
Yet, his models have proven successful in DraftKings and stock evaluations.
Plus, while Goliaths like DeAndre Ayton’s Wildcats can be easily vanquished by Davids in Buffalo, Zaman isn’t just picking a lottery number or spinning a roulette wheel.
“Here I have a slight edge. I kind of know the good teams and the bad teams,” Zaman said. “It’s a slight edge — I’m not Nostradamus. But with a turd projection and my algorithm, I can make gold.”