UI prof a pro at forecasting NCAA's best seeds
By JEFF HUTH
When you grow up in Montreal, as Sheldon Jacobson did, it almost goes without saying that you root for the Canadiens.
It was Jacobson’s good fortune to come of age as a hockey fan in the 1970s, when the iconic NHL franchise won six Stanley Cups during the decade.
“Kind of like living in heaven,” Jacobson said.
With that background, this University of Illinois computer science professor can relate to the fan fervor he’s seen in the United States from college basketball fans, particularly during NCAA tournaments.
Heck, Jacobson even caught hoops fever when exposed to the sport in the 1980s while earning his Ph.D at Cornell University in Ithaca, N.Y.
“It was a lot of fun,” he said. “I just kind of got into it.”
In fact, March Madness inspired Jacobson to write several published papers on bracketology. And that eventually led to the creation of a student project that Jacobson supervises. Melding their interest in the NCAA men’s basketball tournament with expertise in advanced analytics, Jacobson & Co. debuted the website Bracketodds in March 2011.
“Everybody’s so interested in the tournament that we started to look at the numbers in depth,” Jacobson said.
The result is an online tool for the bracketologist in all of us.
Just don’t expect to find a filled-out bracket at Bracketodds. You’re on your own when it comes to picking, say, how far seventh-seeded Illinois will advance or whether No. 14 Valparaiso will take down No. 3 Michigan State in the second round.
“That’s something we leave to the basketball fan,” Jacobson said. “You decide which upsets you want.”
The folks behind Bracketodds focus on the seeds themselves and the probabilities of their advancement, based on results from all tournaments since 1985 — the year the field was expanded to 64 teams.
Although the tournament has since expanded twice more — and the current field is 68 — the number of seeds per region (16) remains the same.
“And everything we do is focused around seeds, not teams,” Jacobson said.
By analyzing what now is 28 years of data on how specific seeds have performed in various rounds of the tournament — as well as how they’ve fared in specific matchups against other seeds — Jacobson and his team have come up with statistical models that can guide a fan’s picks throughout the bracket.
“It’s a tool,” Jacobson said. “We’re simply saying, ‘When you create your brackets, look at the seeds and how far (you project them to) go, and then compare them to what historically we’ve seen.’ ”
Bracketodds offers visitors that opportunity. For example, let’s say your bracket has teams seeded 1, 1, 2, 2 making it to the Final Four. Go to the Final Four probability model on the site and select those numbers. Then hit the “Show Results” button and see the probability.
In this case, the odds are 19.98-1 against that combination getting to the Final Four. As it turns out, however, that combination historically is better than picking all No. 1 seeds to reach a Final Four. In that case, the odds against it are 47.62-1.
“All the probabilities are small,” said Jacobson, “so what’s really important is to compare the relative probability. Even though they are both small, one (could be) five times bigger than the other.”
During the tournament’s early rounds, history points to the likelihood of one of the bottom-tier seeds springing an upset. Not a 16th seed toppling a No. 1. That’s never happened, at least so far.
“That’s going to change at some point,” Jacobson promises.
But if you’re squeamish about picking a No. 13 seed or a No. 14 or even a No. 15 to win its first game, Jacobson advises giving at least one of those scenarios a shot.
“I can tell you with certainty that somewhere in the first round a 13, a 14 or a 15 is going to win a game because it’s so rare that we never see that,” he said. “Now, I can’t tell you which ones, but I can tell you that you want to have one or two of those events happening.
“The laws of probability suggest that one or two of either the 13s, 14s or 15s are going to win. In fact, it would be a shock if none of them won.”
That’s what history indicates. That’s what the probabilities show. That’s why Bracketodds was created.
“Our model lets the fans see how risky their bracket is or how it’s too conservative by putting in numbers,” Jacobson said. “It gives people the chance to at least understand the consequences of their decisions.”
In the process, it also gives the UI professor and his students a unique way to engage in the madness that is March.
“It’s more than an athletic event; it’s a mathletic event,” Jacobson said. “Engineering and math can be part of this event.”