Hack Your March Madness Pool—With Science!

Kentucky guard Andrew Harrison (5) and Arkansas forward Bobby Portis (10) vie for a rebound, March 15, 2015, in Nashville, Tenn. Mark Humphrey/AP

Every March, millions of Americans who pay little or no attention to college basketball throughout the year test their knowledge of the sport with their friends, co-workers, and families in March Madness pools. I’ve got friends who couldn’t tell you the difference between a three-point shot and a three second violation who start to opine on the ability of SEC teams to win close games, or the injustice of Arizona’s seeding in the brackets.

A massive industry of pundits, analysts, and bracketologists has sprung up to service this audience, but very often we miss an essential point. Your office pool isn’t a competition held in a vacuum. Its a fight not only to be correct in your picks, but to be differentiated from your fellow entries. That’s where we come in.

Somewhat amazingly to me, this is the now the sixth straight year that I’ve written this post, where I outline a strategy to maximize the value of your picks in an NCAA men’s basketball tournament pool. Over the course of those years, I’ve heard from numerous readers who have used the method to help win their office pools. So here we go — WIRED’s sixth annual guide to March Madness. I’ve made a couple of tweaks to the methodology, but broadly, the idea remains the same.

Generally, most people’s picks in tournament pools look pretty similar. After a while, you see a consensus. For instance, the millions of people in ESPN’s online pool almost universally believe the top four seeds will win their first game (they’re almost certainly right, as a top seed has never lost to a 16 seed in the first round).

But you can look at each round, and each game, and see the percentage of players who’ve picked which team to win. Call it the wisdom of the crowd, which is pretty darn good. Even with the unpredictability of the tournament, the crowd’s consensus picks usually finish in the 80th percentile or so.

But if you run with the crowd, it’s hard to beat them. To do that, you need to look for teams that others are over- or under-valuing. Like so:

Here’s what those numbers mean. They’re the difference between the crowd’s pick at ESPN, and the statistical prediction from team at FiveThirtyEight.com, which uses five computer rankings and two human rankings to evaluate each team. (FiveThirtyEight also makes their data available on GitHub, which is hugely helpful).

I’ve compared that empirically-driven projection with the percentage of ESPN users who pick a certain team to advance to a certain round. A positive number means the stats say a team is more likely to win than the crowd thinks; negative means the stats say they’re more likely to lose than the crowd thinks.

Games that have more than a 10 percent difference are highlighted—green showing teams that are good bets compared to the crowd and red showing bad bets.

The biggest story in college basketball this year is the Kentucky Wildcats, who enter the tournament with a perfect 34-0 record. No team has gone undefeated and won a national title since Indiana in 1976, but Kentucky is a heavy favorite to do just that. FiveThirtyEight thinks they have a 4 in 10 chance of winning the tournament—by far the biggest favorite in the years since we’ve been doing these projections.

But, are they a good pick to win your pool? Exactly half of the entrants of the ESPN pool have picked Kentucky as their national champion, which isn’t insane, given that the stats project them as the champ over 40 percent of the time. That makes them a massive favorite, but if everyone is picking them in your pool, it’s hard to stand out.

No other team is given more than a 11 percent chance of winning by FiveThirtyEight, but their second most likely champ, the Villanova Wildcats, are undervalued by the crowd. Under 3 percent of ESPN entries pick ‘Nova, so that could be a way to differentiate yourself. At the top end of the field, Duke is rated much more highly by the crowd than the computers—it seems that Duke’s reputation as a team that people love to hate isn’t borne out by their picks.

Another team that stands out as a good pick is Utah. The Utes finished the year 24-8 in the Pac 12, and FiveThirtyEight thinks they have a 62 percent chance of reaching the Sweet 16 and a 27 percent chance of getting to the Elite 8. But ESPN entries in each of those rounds have them more than 20 percent lower than the stats—quite a big gap for a team from a power basketball conference.

If you’re entering a pool for the NCAA’s Women’s Basketball tournament, there’s an even more overwhelming favorite—the 32-1 UConn Huskies. It’s not a shock, as UConn has dominated women’s college basketball for the past 15 years, winning eight national titles since 2000, and making the Final Four the past seven consecutive years. ESPN pool entrants have picked UConn to win 61.7 percent of the time, and amazingly, they have probably undervalued the team. FiveThirtyEight thinks Connecticut has a stunning 74 percent chance of taking its tenth NCAA title.

What about the inevitable early tournament upsets? Texas, an 11 seed, is a slight favorite according to the stats in its first round game, but the Longhorns aren’t getting many picks from the crowd. Also, UC Irvine, Valparaiso, and Georgia State looks like underrated upset picks. Meanwhile, Stephen F. Austin is a popular upset pick against Utah; as I noted, Utah is likely a much stronger team than the crowd is showing. One big upset pick for the women’s tourney? James Madison has a 40 percent chance to knock off the Ohio State Buckeyes, but only 13 percent of ESPN entrants have picked JMU.

Of course, this is a high-risk, high-reward strategy. Especially this year, where the field seems so top heavy, it’s hard to imagine the sort of run that the UConn men’s team went on last season, when they won the tournament as a #7 seed. Best of luck to you all. Please let me know on Twitter (@markmcc) how you do, and if you’re looking for the full set of data, you can find it on Google Docs.