The Meteorologist Who Changed College Basketball

There wasn’t a lot of interest in college basketball in rural Montana in the late 1990s.

But a meteorologist in the state moonlighting as a high school basketball referee was tracking analytics across the sport — and his work was beginning to garner some buzz online.

“I had a rating system… and I added a blog to it. People started noticing that and posting it to the message boards for their teams,” the then-forecaster, Ken Pomeroy, said.

The makeshift site eventually caught the attention of some of the sport’s top coaches, garnering national acclaim in 2010 when 33-year-old Brad Stevens led Butler on a legendary run to the national championship game. He cited KenPom as a critical factor in his program’s success.

“Brad Stevens said, ‘After we beat somebody, the first thing I do is go to KenPom and look at the next opponent and what they do well and what they don’t,'” Pomeroy said. “It took off from there.”

Stevens is now the president of the Boston Celtics, while KenPom is one of the best-known analytics sites in all of sports, shaping everything from college basketball coaching decisions to fan debate.

Best of 7 spoke to Pomeroy about the early days of his site, the difference between data that’s relevant and data that’s interesting, and why it’s important to have healthy skepticism of numbers.

This interview has been edited and condensed for length and clarity.

Ken, what led you to first explore college basketball analytics?

I was reading about baseball online — just a lot of random people writing about baseball who had access to all sorts of new and interesting data. I wasn't a huge baseball fan, but you could see an audience developing there, and I was like, “Who's doing this for college basketball?”

The stat I really got into was points per possession. Back before the world was so connected, I thought I had this unique idea that I had invented, which I had not, but it revealed which teams weren't putting up a lot of points but still had really efficient offenses. I saw that information and learned a little bit about programming, and decided to put it online for what initially was a very small audience.

How did you come to the decision to give up the National Weather Service job and pursue KenPom full time?

It wasn’t easy. I really liked my job. Weather fundamentally affects everybody, so making good predictions about it was really rewarding.

But the decision was kind of made for me. As things became more popular on the basketball side, it became less of a hobby that I could just spend an hour on updating the site every morning and became something I had to invest more time in, like a second job. At some point, it was just like, “Hey, I probably can make some money off of this. I don't know how much, but let's test it out.” I tested it for one year, and it went really well. At that point, I was like, I’ll probably just do the basketball thing.

You collect so much data. How do you distinguish between what’s relevant and what’s simply interesting?

It’s sort of become an unfortunate issue in college basketball. There are fun facts, statistical coincidences that are interesting but maybe not useful for analysis. For example, a team’s won 10 straight games on Tuesday.

When I was starting, I could put out a stat and people would be like, “That doesn’t mean anything.” Now, anybody can put out any stat with any small sample, and people are less skeptical than they used to be. I’m all for interesting fun facts, as long as they’re couched as such.

I have free throw defense on my site, which I think is obviously not useful, which most people can pick up. But as the unquenchable thirst for content increases, everybody wants new fancy stats, and some of the stats out there can be fun, but they’re often presented as meaningful when they’re not a lot of the time.

It sounds like you think there actually needs to be initial skepticism around metrics to make better use of them.

On the user side, there always should be some skepticism with data. The burden should really be on the person creating the measures to provide evidence that it’s meaningful. A lot of times, that’s not the fun part of analytics. The fun part is creating the data. The less fun part is trying to prove to yourself the data is useful.

As respected as your work is, you get a fair amount of criticism too. When did you develop thick skin?

I don’t know that I have thick skin. I’m pretty sensitive to criticism in general. The best example is the NCAA tournament. There are always upsets, and that’s kind of why we watch. We understand there’s variability in human performance, and there are always going to be upsets. But invariably, people who are waiting to criticize my work are like, “KenPom said so and so was going to win this game. And he was totally wrong. He doesn’t know how to predict anything in this tournament.”

But the point of the ratings is not to predict tournament outcomes with certainty. That’s impossible to do. The point of the ratings is just to give you an idea of who’s better and expected to win on a certain night with a certain probability. People criticize that and don’t understand.

Some of that, I guess, is on me — I could make it clearer about what I’m trying to do. But some of that is on the audience. When they just criticize without trying to understand things, that’s less hurtful but definitely frustrating.

You’re pretty synonymous with the sport now, but how gritty has the journey been to get to where you are?

I started with no connections. I was a terrible player. I was also un-coachable. It’s a really bad combination. From that standpoint, it’s interesting.

Now, the site is mostly automated. But it was 20 years of grinding and programing. Initially, I was updating the ratings every morning and had to run some code I created. When I started doing player stats on a nightly basis, I would stay up until all the games were done every night and run the code to update the ratings. Sometimes, that would be 2 or 3 in the morning. It’s hard to believe I ever did that.

I was kind of stupid, too. I never really hired anyone full time. Once you go through this process and do things by yourself, you feel like you have some sort of special touch that you don’t want to trust anyone else with.

Is there one piece of retrospective advice you’d give yourself?

I started doing the site in its modern form in 2004, but I was calculating points per possession and stuff like that two years before. I felt like there was no way I could make a website and update it frequently and get this information out there — that it all seemed beyond what I could do. But eventually, I just did it. It took time, and it took learning from other people, and a few breaks, but the point is, just have a little more confidence.

You have an idea. Don't be discouraged if it doesn't work out initially because I think it really delayed my development and my work. If I had just had more confidence that I could do this, I would have started earlier.

What I learned…

I wasn’t necessarily expecting Pomeroy, a pioneering analytics expert, to encourage prudence around the application of data, but he offered some helpful insight on why new statistics really need to be scrutinized.

As analytics continue to emerge, the ability to filter what’s revolutionary and what’s simply noise will be even more important.

It was also refreshing to hear Pomeroy say he didn’t entirely know that he had a viable concept when he launched his site.

We tend to celebrate confidence and conviction, yet hesitation and doubt often accompany the beginning of many worthwhile endeavors.

Pomeroy’s perspective reminded me of what boxer and best-selling author Ed Latimore told me.

“We go do these things, and we take these risks, and we look stupid, and we get our egos dragged through the mud. But we learn, learn, learn,” Latimore said.

“Before you know it, people are like, ‘How did you do that?’”