Profs look for order in March Madness

While the toughest competition most employees face in March Madness office pools comes from die-hard college basketball fans, professors and staff at the Yale School of Management must contend with something else entirely — colleagues who are not only fans, but the brains behind a mathematical model for predicting the winners.

As Illinois prepares to take on North Carolina tonight for the Final Four Championship, SOM professors once again have shown the success of their model with a bracket that mathematically predicted the final matchup between the two number-one seed teams. But despite their general triumph, there are those who question whether office pools depend on anything other than pure luck.

After losing for over a decade in office betting pools, Yale SOM Deputy Dean Stan Garstka and professor Ed Kaplan decided to apply some of the techniques they used daily in their teaching to March Madness betting.

“Since we were trained in using mathematical models for making decisions, we just asked ourselves if there was a way to model this sort of thing,” Kaplan said.

Though the professors correctly predicted the teams in the championship game, they did not pick Louisville or Michigan State to make up the Final Four, Kaplan said.

The SOM professors first wrote the article, “March Madness and the Office Pool,” in 2001. Since then, reporters from media outlets across the country have often called them asking for advice on how to pick winning match ups. Kaplan advised the Wall Street Journal this month that the best method for amateur betters is to decide who they think will make it all the way, instead of predicting winners for the early rounds.

“Most people start by working in the other direction: who will win the first round games, and then who will win between those picked,” he told the Journal. “The problem with this approach is that it is too localized — you don’t look at the big picture. For example, is one region slightly weaker than another?”

Initially attempting to create a better indexing system, the SOM professors eventually decided that the better method would be to consider the structure of single-elimination tournaments and show how to efficiently calculate the mean and variance of the number of total points earned in an office pool for a given slate of predicted winners. The professors used various Markov probability models for predicting game winners based on regular season performance, professional sports rankings and Las Vegas betting odds.

Ariel Schneller ’06, who has placed a wager in March Madness pools for six years running, said he thinks the odds of predicting the eventual winners are quite low.

“It is definitely not impossible to come up with the most likely outcome, but to say that you will have a probable one seems a bit untenable,” he said.

Several factors must be taken into account in predicting the winning team, according to Kaplan and Garstka’s findings. For small office pools, it is important to understand who you will be competing against; but the larger and more complicated the structure of the pool, the more effective the mathematical formulas are for predicting winners.

The mathematical model generally has proven successful, Kaplan said.

“We have done extremely well in some pools, but we are usually at least competitive, Kaplan said.

With a correct application of the model, Kaplan and Garstka have had a 58 percent success rate picking winning teams.

Matt Kornguth, owner of RunYourPool.com, a “pool management” Internet service which helps customers to set up betting pools, said he does not think any mathematical formula could accurately predict the winners.

“I would think that they would be able to pick which teams would not win, but would only be able to narrow the group of potential winners,” he wrote in an e-mail last week.

Kornguth, who has been betting in pools for many years, attributed the success of March Madness office pools to several factors: it does not always take real knowledge of the game to win; the 64-team single-elimination format provides enough different winning scenarios that a single person is likely to win the top prize; that they players are amateur, unpaid athletes increases their enthusiasm; and entering pools is a morale booster for workers at large companies and is being encouraged by managers more and more for that purpose.

As much as luck is important, knowledge of teams and matchups certainly helps, Kornguth said.

“In my experience, you will find that a knowledgeable fan who follows the sport very closely will, year after year, do better than the casual or non-fan,” he said.

Kenneth Massey, a Ph.D. candidate for mathematics at Virginia Tech, has developed a statistical showcase for the tournament. Massey said there are so many upsets that luck actually is a huge factor in correctly predicting the winning team.

“You have to get lucky in terms of picking the right upsets,” he said.

Steve Cuddihy, who runs the Internet betting site OfficePool64.com, said that he has been involved in March Madness office pools since 1997 and has seen a different person win every year.

“With the way the NCAA conferences are set-up, you cannot really judge which conference is better than the other,” Cuddihy said. “I think when it comes down to the tournament it is that is old cliche — any given Sunday.”

Though some professors in the Yale School of Management contend they can mathematically predict NCAA tournament winners, others say it is all just luck.
Susan Keppelman
Though some professors in the Yale School of Management contend they can mathematically predict NCAA tournament winners, others say it is all just luck.

Comments