How to Use Stats to Win Your March Madness Pool

By
Casey Prather, whose Florida Gators finished 18-0 in SEC play. Photo: Rob Foldy/Getty Images

There's extra incentive to fill out a great NCAA Tournament bracket this year: Warren Buffett is offering a billion dollars for a perfect one, in the 1-in-9.2 quintillion chance that someone correctly picks all 63 games. But even if you have more modest goals (like winning your office pool), one needn't have spent the past four months breaking down video of Coastal Carolina and Eastern Kentucky to pick teams like a pro. Here, five stat-influenced tips from bracket experts.

1. Fill out your bracket based on the number of competitors.
If you're in a relatively small pool — say, fewer than 100 entrants — you can afford to play things pretty safe. If you pick the Final Four and champion correctly, you'll be in great shape. But in a larger pool, in which multiple people might nail the Final Four, your bracket needs to be a bit of an outlier.

Pete Tiernan, who runs the website BracketScience.com and has been studying Tournament data since 1990, says that to win such a bracket, it's best to take some risks by "curve fitting" your bracket to include just the right amount of chaos. The average tournament, for instance, contains nine major upsets (defined as a team winning a game in which its opponent is four or more seeds higher). And so in that case, Tiernan suggests picking exactly that number of big surprises, using the difference in Ken Pomeroy's Pythagorean calculations for expected winning percentage to identify the strongest underdogs.

Sure, this strategy could lead to your bracket getting busted early — but in a large pool, it still ups your chances of taking the top spot.

2. Pay attention to teams that are hot right now.
Tim Chartier, a math professor at Davidson College, has been looking at Tournament results since 2009 — specifically, he looks for ways to make better predictions. Chartier's research has found that the "recency effect" is a good indicator: Teams that are playing well heading into the tournament, even if they struggled early in the season, are smart picks.

Winning the conference tournament isn't a necessity, but mathematical models that "up-weight," or assign more value to, recent games have worked for Chartier and his students in the past. (Chartier says the concept is similar to the one Nate Silver uses to emphasize polls nearer to an election.)

Chartier says that up-weighting recent games as well as away wins can work even better, so take note of teams that have won tough road games.

3. Do not pay attention to the RPI.
Lazy fans might be tempted to use something called RPI, or the Rating Percentage Index, to determine which of two teams is stronger. It looks like it'd be perfect for this sort of thing: RPI's formula ranks every school in Division I, and results are posted right on the NCAA's official website. But while RPI is a decent indicator of which teams will be selected for the tournament, it's a poor indicator of how they'll do once they get there.

If you're looking to put as little effort into your selections as possible and want to pick games based on rankings systems, Pomeroy's rankings and ESPN's BPI rankings are better bets. (According to Tiernan, picking based on either is better than picking based on seeding, though in recent years, it hasn't been by much.)

4. Know how to spot an upset, but don't get too clever.
There's no single way to identify potential Cinderellas, but Tiernan suggests looking for teams that play a high-risk, high-reward style. Metrics to keep an eye on include offensive rebounding percentage and defensive turnover rate. An example in this year's tournament: Southland Conference champion Stephen F. Austin, which is ranked in the top ten nationally in both of those categories.

Tiernan also suggests looking for teams with imbalanced scoring — in other words, teams that get more than 60 percent of its points from either its frontcourt or backcourt, which could present a matchup problem for a superior team. (Delaware, which relies heavily on its guards for scoring, could be a team to watch if it gets a favorable seed and matchup.)

That said, pick your upsets carefully. Though it's not unheard of for 13 seeds (or even 14 or 15 seeds) to advance out of the first round, it's not a risk worth taking, according to Tiernan. "If you look at the record," he says, "for whatever reason, the 5-12 game seems to be the break point at which the gap in terms of competitive talent is close enough that you can make that pick."

Another reason not to pick a 13 seed: Knocking out a No. 4 too early is more likely to burn you deep in the tournament. A 4 seed is more likely than a 5 to knock off a No. 1 in the Sweet 16. Remember: As much fun as it can be to nail an early upset, with graduated scoring, it's important to keep teams alive deep into the tournament.

5. Don't get too attached to any one stat.
There's no magic statistic (or ranking system) that will predict the winners of games. Chartier, for instance, notes that a team's so-called "four factors" — effective field goal percentage, turnover percentage, offensive rebounding percentage, and free throw rate— are indicative of future success, but even those are hardly the only numbers worth looking at.

Tiernan, meanwhile, uses a set of eight credentials to identify potential champions, which, after all, is pretty much a necessity if you're hoping to win your pool. (Criteria include a coach with a track record of success in the Tournament, an average of at least 73 points per game, an average margin of victory of more than 7, and one of the 75 most difficult schedules in the country.)

But even Tiernan admits that "anybody that tells you that they have the right answers is not being intellectually honest." He adds: "Don't assume that numbers are this fail-safe way to pick these things. I mean, sometimes crazy shit just happens."