Oregon’s Diana Inch won the Yahoo bracket contest last year. Beset by constant paparazzi attention and requests for money from distant relatives, her behavior has become increasingly erratic, and she recently checked herself into a luxury rehab facility for “exhaustion.”
Most of the talk about March Madness pools focuses on who’ll get seeded where, or trendy upset picks, or, even on the geekier end, how to best pick the most games correctly using statistical models. But really, filling out a bracket isn’t about picking the most games correctly. It’s about giving yourself the best chance to win. And those are two different things.
In 2007, a pair of math and computer science professors at St. Louis University published a paper about March Madness pools, finding that entrants often select certain teams with a disproportionate frequency compared with those teams’ actual chances of winning. Without an immaculate performance in the early rounds, picking these crowd favorites makes it nearly impossible to win the entire pool. You might be right, but everyone else is too.
This Year’s Most Over- and Under-Picked Teams
“Estimated probability” is a team’s chance of reaching a given point in Nate Silver’s forecast, which is itself a combination of several different computer ranking systems, like Jeff Sagarin’s and Ken Pomeroy’s, and also accounts for recent injuries, geographical location of games, pre-season rankings, and tournament seedings. For “pool probability,” or the probability that other participants have chosen the winner of a given matchup, I used public data from ESPN’s Tournament Challenge. Both data sets were generated at approximately 10 p.m. on Monday, March 18.
But picking too many underdogs, particularly in the early rounds, is a bad move too. In the ESPN Tournament Challenge system, on which this analysis is based, it’s rarely worthwhile to pick a high seed to win one or two games just for a measly point or two, especially when its opponent has a much greater chance to go further in the tournament. Thus the delicate position in which most bracket-pickers find themselves — choosing enough underdogs to separate from the crowd, but still picking teams with a high probability of winning.
This is intuitive to most, but is an element frequently left out of March Madness models — and the reason why so few people adhere to the probabilities all the way through. Most will throw in a few upsets to differentiate themselves, and these are the picks that are most informed by intuition. Here, we’ll attempt to put some science behind those separate-yourself-from-the-crowd picks.
Ideal Picks for a 25-Person Pool
Methodology: Picks have been generated according to software developed by professors Bryan Clair and David Letscher of St. Louis University that considers a team’s estimated probability of advancing in the tournament along with the frequency with which other participants have chosen that team to advance.
In a small pool, it’s prudent to make conservative picks, as less is needed to distinguish your bracket from the pack. But this year, the perceived parity among teams means that the conventional conclusions that this model draws — make cautious picks in smaller pools and bold picks in larger pools — are less pertinent. Unlike previous years, where as many as half of participants picked the overall favorite, the most popular choice this year, Louisville, is the championship pick in just 21.7% of entries. This means that there is less of a need to make bold picks in larger pools. With no bandwagon favorites, stay conservative; the picks shown above for a 25-person poll, which skew heavily toward lower seeds, were extremely similar to the output for a 50- or 100-person pool.
If anything, pool participants have overcompensated for the parity among the field by picking too many underdogs. The three most statistically favored teams — Louisville, Indiana, and Florida — are all being picked less frequently than perhaps they should be. This might be the year that playing it safe pays off.
Nonetheless, at the scale of massive competitions like the ESPN or Yahoo pools, the best chance for winning still comes from a few wise upset picks and avoiding overrated teams. Choices like three-seed Florida for the title, Michigan over Kansas, and Miami falling in the Sweet Sixteen are all relatively defensible picks that comparatively few number of people are choosing.
Just don’t tell your friends you read this post.