- Picking the same bracket as everyone else won't win. Here's how to find undervalued teams in your pool.
It's the time of year where people who don't know the Shockers from the Golden Flashes are suddenly trying to become college basketball experts, all in service of winning their March Madness pool. We've got plenty of ways to help you here at SI, from Luke Winn's Bracket Math column as well as his story on the teams most likely to pull off a big upset, to our panel of experts and their brackets.
But to submit a winning bracket, it's not enough to just be right—you also have to be different. After all, if everyone in your pool submitted an identical bracket, you'd all tie for the win. The best path to taking home the prize is to find ways to depart from the consensus picks, and make smart choices on lesser-known teams. The wisdom of the crowd is pretty darn good (if you look at consensus picks in the big bracket challenges, it would usually finish in the top 15 percent), but to win your pool, you want to find teams that are being undervalued or overvalued by the masses.
For the past eight years (mostly during my time at WIRED), I've broken down the field of 64 to try and find that kind of edge. I've heard from dozens of people who have used this approach to win their pools, by finding a way to zig while everyone else Zags. Although this year, Zagging is a good bet. More on that below.
I've gathered the predictions of two of the most respected stat-based prediction systems: Ken Pomeroy's and those from FiveThirtyEight. Then, I've compared that to the choices that players have made in two of the largest online brackets: ESPN's and Yahoo's. When you compare the stats to the crowd, you get the results below.
Here's what those numbers mean. A positive number means that the stats think that a team is more likely to win than the crowd does; a negative number means that the stats think that they're more likely to lose than the crowd's choices. Games that have more than a 10 percent difference between the two predictions are highlighted--green for good bets compared to the crowd, and red for bad ones.
This year's tournament is notable for the lack of a clear favorite—unlike most past years, there's not a dominant team that's a clear first choice. Interestingly, that's led to the crowd actually overestimating the chances of many top seeds to reach the Final Four. For instance, over 50 percent of fans have chosen Kansas to reach the Final Four, while the statheads only give them a 30 percent chance. That's a huge difference. The other number one seeds are also overrated by the crowd with one big exception: Gonzaga.
In the years I've been doing this analysis, I've never seen a No. 1 seed that's been undervalued by the crowd when it comes to its chances of making the Final Four, but that's the case this year for Gonzaga. The Bulldogs have a 43 percent chance of making it to Phoenix according to the stats, but only 35 percent of pool entrants have picked them. It gets even crazier when you look at the national title game. Stats peg them with a 26 percent chance of playing for the title, while only 15 percent of brackets have them going there. It's hard for a top seed to be a dark horse, but Gonzaga somehow is.
Other teams to consider include Gonzaga's conference mates, St. Mary's. The Gaels get a lot of love from the statistical projections, but not so much from users. Miami also suffers from a relative lack of respect.
When it comes to picking upsets, a couple of teams stand out. Oklahoma State and Wichita State, both No. 10 seeds, look like good picks to pull off the upset—in fact, Wichita State is actually favored in its matchup with Dayton in the first round, both by the stats and the crowd. And Princeton, while a popular upset pick in lots of brackets is still undervalued by the crowd.
Now keep in mind, this is a high-risk and high-reward strategy. But since most pools only offer prizes for the very top entries, there's no reason to pursue a strategy that can, at best, place you in the middle of the pack. Good luck, and please let me know on Twitter (@markmcc) how you do. If you're looking for the complete set of data, you can find it on Google Sheets.