A Stats Guru Explains How to Win Your NCAA Tournament Pool

Having trouble filling out your NCAA tournament bracket? Ed Feng, founder of The Power Rank, can help. The sports analytics website, which he created in 2008, specializes in ranking teams—in multiple sports—and predicting the outcome of future games. For example, according to his interactive tourney bracket, 14-seed Harvard has a 25.7 percent chance of upsetting 3-seed New Mexico, and a 2.1 percent chance of advancing to the Sweet 16. This week, Feng, whose work has been featured in Sports Illustrated, explained a little bit about what he does and gave us a few tips on how to win the bragging rights associated with winning our NCAA tournament pools.

So how did you kind of get into sports statistics?

I’ve always been a sports fan. That was the start. I was getting my Ph.D. at Stanford and learned a lot of basic statistics and physics and was kind of on the academic track until that kind of went awry. I was looking for an alternative kind of career path. I interviewed with Google, and I kind of did my homework on the company and the technology, and came across PageRank and the original paper [about it]. PageRank is the technology for ranking web sites … it did this based on the link structure of the web. So I was reading the paper, and it turns out that it’s the same math that I was doing in my research. So I was like, “Oh, this is really cool.” And then it kind of hit me that, well, let’s do something that’s actually important and start ranking sports teams based on it. So I applied the core idea of PageRank, then I applied a bunch of ideas from my research. Instead of just looking at wins and losses to rank teams, I figured out a way to take margin of victory into account.

How would you sort of explain your rankings to someone that doesn’t know too much about them?

What my core algorithm does is take sports statistics, like margin of victory, and it adjusts it for who you play. For example, in college basketball, there’s a huge difference between playing in the Big Ten, which is widely considered the best conference in the nation this year, and playing in the Southland Conference. And so, my algorithm kind of sorts through all the sets of match-ups that happened. [For example] Michigan played a bunch of really strong teams, we’ll account for that. And at the end of the day, Michigan kind of has an average of margin of victory. The algorithm adjusts that, because [Michigan] played in such a good conference and then, that is what I call the rating. The rating is essentially a number like 14.6. And it gives a prediction for how a team would do against an average Division 1 college basketball team. And then every one is ranked according to that rating.

Do you give advice to friends who are filling out their brackets?

I do. It was kind of one of my [goals] these last couple months to figure out how to best fill out a bracket. It turns out there’s some really interesting math. I didn’t get into it nearly as much as I wanted to. It all kind of involves kind of finding value. It actually does no one any good to pick [overall top seed] Louisville in your bracket. Especially if you’re in a huge pool. Because even though they’re definitely one of the top teams in the tourney, 28 percent of the brackets on Yahoo! pick Louisville. So you’re playing this game, where, even if they win, then you and a bunch of other people have the 32 points for that pick. And then the results of the pool end up depending on earlier rounds, and that can be much more random. You know, my grandma that picks a couple of crazy 13 seeds and ends up taking home the pool.

You sort of just explained this, but what do you think the biggest mistake is that people make while filling out a bracket?

I think the big thing that I’m trying to teach people is that, [first] you don’t want to enter a big pool. The odds of you winning a big pool are just so—even if you have a huge edge—are just so ridiculously small, it’s not even funny. So, that’s the first thing. Even if you get into these medium-sized pools, like 20 people, you’re not trying to predict a champion—you’re trying to win the pool. Those are two completely different things. But if you’re trying to win a pool, you essentially have to make some risky picks. So the thing I’ve been really banging the drum on is picking Florida as the champ. They’re a really undervalued team. Most people kind of come in and see that they’re 13th in the AP poll and that they’ve lost seven games, and don’t think much of them. But they’re actually the best team in my rankings. They have lost a bunch of close games, but when they win, they win by a ton of points. Again, my system kind of adjusts for margin of victory. So, Florida’s a valuable pick if you want to win medium-sized pools.

Any sleepers you’re looking at in this year’s tournament?

One team that I’m really looking at is Creighton. I have no idea what the committee seeded the seventh. They play in a really tough conference in the Missouri Valley. They actually, by my numbers, have, if not the best, the second-best offense in terms of points per possession. I watch a bunch of Missouri Valley [basketball], and [the Bluejays] seem to be really clicking right now. Of course that can all change when you go up against Duke in the second round. But that’s kind of a lower seed that I can see getting to the Sweet 16.

We always see a double-digit seed in the Sweet 16. And I’m looking at this pod with Davidson, Marquette, Bucknell, and Butler. I’m pretty sure one of those is going to come out of there. And the reason is—I don’t necessarily like Davidson or Bucknell—but Marquette and Butler are pretty crazy overrated by their seeds. Butler isn’t near the team that they were a couple years back. They’re not as good defensively. Marquette is actually a team that I really love. They used to be underrated every year because they’d have a bunch of losses in conference when the Big East was really good. But they were still a really good team. Now it’s the exact opposite. Now the Big East is a little bit weaker, Marquette wins part of the regular season title, and is now a 3 seed.

So lastly, what do you have planned for the tournament?

This is actually my job. It’s fun. I will be somewhere watching the games. I used to live in Palo Alto, and I had a group of friends there that we would every year kind of descend on a bar and not leave until the first day of games on Thursday was over. So that was a lot of fun. [This year] I will be at some bar in Ann Arbor, watching a bunch of games.