March 15, 2009 at 9:55 PM ET
Inside Science News Service / Ivanhoe
Click on image for video: Mathematician Mike Breen says "March
Madness" basketball brackets can generate quintillions of outcomes.
As you puzzle over the "March Madness" basketball bracket for your office pool, you can go with pure luck - and hope you hit a 9,000,000,000,000,000,000-to-1 long shot. Or you can play it safe and just go with the higher-rated team for every game. But how scientific is that? A growing number of online tools promise to give you an analytical edge over your officemates.
Bracketology - the study of the NCAA tournament selections for the purposes of filling out a winning office-pool entry - has been around for more than two decades, even before the World Wide Web began. But bracketologists have really stepped up their game in the past few years, said Pete Tiernan, founder of BracketScience.com.
"You've got people who break down basketball games possession by possession, and use logarithmic analysis to do simulations on the game," Tiernan told me today. "What we're seeing now is the 'Bill Jamesing' or the 'Moneyballing' of college basketball. It just so happens that the NCAA tournament really lends itself to that."
The sheer mathematical breadth of the tournament bracket is breathtaking: There are 263 possible ways to fill out the standard office-pool form (the play-in game between the 64th and 65th teams is usually glossed over), or 9 quintillion permutations.
Just how many office-pool sheets does that translate into? "If you took every school from kindergarten through college, and covered every basketball court with brackets, you would have stacks of brackets 591 miles high on every court in the country," Tiernan said.
You could narrow the odds by sticking with the top seeds: After all, the tournament's selection committee has essentially done all the ranking for you, based on performance during the regular season.
The only problem is that everyone else can do that, too: Chances are your office-pool entry won't stand out from the crowd, and you'll probably end up in the middle of the pack. What's more, some pools are weighted to give extra points for predicting upsets and Cinderella teams. The rewards don't necessarily go to those who play it safe.
That's where high-powered analysis enters the picture.
"The way we think about it is, you just have to do a better job than the selection committee did at picking the seeds," said Tom Federico of TeamRankings.com. The Web site's analysis tool, BracketBrains, matches up historical teams with statistical profiles that are similar to teams in this year's tournament - right down to the perceived home-court advantage for each game.
The software can point to the games where upsets are more or less likely, said Mike Greenfield, one of Federico's fellow "nerds" at TeamRankings.com. "In many cases, the NCAA will mess up, and we'll see that the No. 3 team is probably in the top 25, and the No. 14 should be a No. 10 seed ... so instead of being a 90 percent [probability], it's a 60-40 split."
You do have to pay a subscription fee to get the full benefit of BracketBrains or BracketScience, but for a free analysis with fewer bells and whistles, you can check out the Poologic site - which draws upon years and years of bracketological research. Poologic's Webmaster, systems analyst Tom Adams, estimates that his site's users have won at least $250,000 in pools to date.
If you want to take a chance on winning the office pool - or winning the admiration of your co-workers - you might want to follow a contrarian strategy, based on research published last year in the journal Chance. I've discussed this line of research before, but the latest findings indicate that not picking all the favorites could lead to a better "return on investment."
One of the co-authors of the study, University of Minnesota statistician Bradley Carlin, said you don't want to overdo it: A No. 16 seed, for example, has never won the tournament - so picking your Cinderella from the bottom of the pile probably won't get you far.
His advice is to go with the lowest-rated No. 1 seed or the highest-rated No. 2 seed to take it all, and then fill out your bracket in reverse, using the Sagarin ratings as your guide.
"Then cross your fingers and have another beer," Carlin told me.
Bracketography's David Mihm favors a more mainstream approach: "If you can pick three of the Final Four teams ... those games are typically worth more points. I would really focus on the teams you think are going to be there in Detroit this year."
Over at BracketScience.com, Tiernan said the strategy you decide upon could depend on how big your office pool is.
"In smaller pools, with less than 50 [participants], I don't think you want to go with a contrarian strategy," he said. "The sample size is small enough that if you just do well you're going to be in the running. Now, when you get to 100 people, 200 people and up, you almost have to take a contrarian strategy to win. ... Of course, the risk in smaller pools is that you'll be at the bottom of your pool, and you'll look kind of foolish."
But if you go with a long shot that turns out to be a Cinderella team, even if just for a night or two, you might win a different kind of payoff.
"I ask people, what do you want to be?" Tiernan said. "Do you want to do well in your pool, or do you want to be the guy that they look at and say, 'Wow, he picked Kentucky to go two rounds! He's a genius!'"
That strategy may not be scientific, but it sure sounds like fun.
Update for 8:59 p.m. ET: Poologic's Tom Adams sends along some inside information, just in time for filling out your bracket: "This is a good source of info to find strong teams that are underbet for champ. Memphis seems to be one of them, from the early data (8:50 p.m. in the East). The pick distributions are good estimators of office pool pick distributions, except for localized favorites. For instance, Memphis probabably won't be underbet in Memphis."
Update for 2 p.m. ET March 19: LiveScience reports a couple of new twists in bracketology, including the claim that top seeds don't matter once you get past the first couple of rounds of the brackets, and the development of a new computer model that supposedly outdoes other systems for ranking basketball prowess. Just for fun, compare the Georgia Tech researchers' bracket (which has North Carolina going all the way) against your own picks and the actual results.
For another helping of slightly less scientific bracketology, check out Steve Cuddihy's tips at Officepool64.com. To get the full story about the NCAA tournament, click on over to the college-basketball coverage at NBC Sports.com. For more on the mad, mad mathematics of bracketology, check out Inside Science News Service. And for a contrarian view on the whole idea of bracketology, check out this column from Mike Celizic, who is calling upon the math geeks to "give it a rest."
Thanks to John Cherry of Ivanhoe Broadcast News for assistance on the Inside Science News Service video above, featuring mathematician Mike Breen.