Thursday, August 29, 2013

Q&A with Ken Pomeroy

The summer is coming to an end soon, and while that means football is gearing up, it also means college basketball is about 2 months away.  Since there won't be much to discuss until early October, I decided to put together a piece to get you through the annual doldrums.

I, along with thousands of others, am a subscriber to Ken Pomeroy's website, www.kenpom.com.  Ken is a basketball stats guru whose website provides a lot of stats that the main sports websites (ESPN, CBS, etc.) don't have.  While he'll show the common stats like shooting percentage, he places a lot of emphasis on tempo-free stats (rather than points per game, think points per possession).  He also provides game-by-game predictions for each team and the likelihood of that score outcome (i.e. Marquette plays Butler tomorrow, and 65% of the time MU will win 64-62).

Ken's website has become very popular in the last few years, as coaches and the NCAA tournament selection committee have put more emphasis on stats in evaluating teams.  I caught up with Ken to learn a little more about how he runs the website, his thoughts on the selection committee's recent performance, and whether or not golf will be added to the website.  Enjoy.


1977:  When you watch basketball, how do you watch a game?  Are you churning through stats in your head?  Are you paying attention to a team's intangibles like toughness?  Or do you just try watch a game for the entertainment value?

Ken: If I'm in the stands, I'm pretty close to being just your average obsessed fan. I'm usually aware of the tendencies of both teams. I might focus on some unusual aspect of one of the teams or players while I'm watching. I'll jump out of my seat when someone does something crazy. If I'm with the media, I'll take some notes and monitor stats and substitution patterns more closely and I try to avoid jumping out of my seat.


1977:  Regarding your rankings of each team and pre-game score predictions you have on your site, are there any flaws you would admit to that you're trying to work on?

Sure, there are flaws. The problem is if I spent a whole month doing nothing but working on it, maybe I would improve the predictions by 1% or something like that. Not because my system is perfect, but because there's a limit to how well an automated method can predict scores. Additionally, there are a few dozen systems out there now, and a lot of good ideas have already been taken. There's more of an intellectual payoff for me to spend that time doing some unique research.


1977:   I liked when you used to highlight big wins and bad losses on a team's schedule with varying shades of red and green.  Why did you get rid of that?

Ken:  I just thought it was a cleaner look on the schedule to not have a lot different shades of green and red. I suppose I could add it as a user option. I do feel strongly about quantifying the quality of wins and losses. Too many can't grasp that beating the #10 on the road is more impressive than beating the #1 team at home.


1977:  Are you more intrigued by a historically good team, or a historically bad team, and why?

Ken:  I'm interested in anything unusual, but the Grambling situation last year was just sad. On some level I guess it's interesting statistically, but having a completely uncompetitive team makes a joke of the game.


1977:  Is it possible to get "too caught up" in the numbers?  Or do you feel it's ok to ignore the qualitative factors when analyzing how good a team is?

Ken:  I don't see the concepts of qualitative and quantitative as mutually exclusive.  Someone can say a team is good at offensive rebounding in a qualitative sense and we can look at some numbers to get an idea if that's true. Ideally, I'd like to have a balance between qualitative and quantitative assessments, but that's not always feasible what with 351 Division I teams. Regardless of what kinds of information are available, it's important to understand the limitations of both the numbers and your eyes.

1977:  As you're aware, the Big East has reformed itself into a "basketball-only" conference.  How have you viewed that as far as what it has done for college basketball?  Do you look at it as a good thing that the "Catholic 7" has made a statement about the importance of D-1 basketball, or is it just another sad by-product of being pushed around by college football?

Ken:  In general, I think basketball is less affected by realignment than football. Even in a 16-team conference, you can play every other team at least once in a season.

I don't know that the new Big East really makes a statement. Until football is banned by Congress, that will always drive conference membership considerations in the richest conferences, and there will be a trickle-down effect that other conferences will experience.


1977:  How do you think the NCAA Selection Committee has done in the past few years?  Have they been fairly accurate given the constraints they have to work around?  Are there certain types of teams you think they value too highly or not high enough?  Are they putting in enough mid-majors, or is the BCS bias too prevalent?

Ken:  I'm fine with how the committee has done given the system that's in place. The difference between the 25th best team and the 50th best team is very, very small. You could pretty much take all teams seeded 5 through 12 and seed them randomly and not make the tournament any different. That's how much similarity there is at those seed lines. So when somebody is outraged that a 12 seed should have been 6 seed, I just shrug. It makes almost zero difference.


1977:   Do you still work for the National Weather Service, or are the subscription fees to kenpom.com paying the bills?

Ken:  I've been doing the site full-time since the beginning of last season.


1977:   Regarding the upkeep of the website, does anyone work for you, or is it a one-man show?

Ken:  It's almost all a one-man show. Occasionally people help me out with data collection/analysis projects.


1977:  A lot of people are intrigued by the "luck" factor in your formula (including myself).  I think what makes it intriguing is that you have a subjective input in a formula that is otherwise very objective.  Other than the fact that Dean Oliver came up with it, are there any insights you can share as to how it's calculated, and how much weight (if any) it has in your overall team rankings?

Ken:  Well, luck isn't included in the ratings at all. The formula is kind of complicated but it basically just measures a team's performance in close games. If you must know, it's the difference between a team's actual winning percentage and Dean Oliver's correlated gaussian winning percentage described here: http://www.rawbw.com/~deano/methdesc.html#corrgauss Positive numbers indicate a good record in close games.


1977:  You were working on golf rankings at one point, whatever happened to that?  Did it fall through, or are you planning on bringing it back to the website?

Ken:  It could happen someday. Unlike with college hoops where it is easy to get data in a standard format, each golf tour basically does its own thing which makes data gathering more time consuming.


1977:  Have you done any work in terms of figuring out the likelihood that Tiger will get to 19 majors?  If so, what have you come up with?

Ken:  I wrote about it three years ago (http://kenpom.com/blog/index.php/weblog/entry/the_case_for_tiger_woods) during his low point and although I didn't give a specific prediction, the implication was he still had a really good chance. Of course, he hasn't added to his total since, so it's a closer call now, but I'd say it's around a 50/50 shot to do it. I think he's clearly the best golfer in the world and don't buy that he has some sort of mental block in the majors. His lack of driving accuracy hurts him more in the majors than it does in a typical event because the rough is grown higher.


1977:  How would you describe your golf game?

Ken:  Normally, I would say it's not disgraceful, but it's been trending in the wrong direction recently.  Anyone that plays the game well has my fullest respect.


1977:  Are there any additions you plan on making to the website in the near future?

Ken:  I tend to not get into too many specifics on questions like this, because most of the things I add just happen out of the blue, and the things I promise to do end up not happening. (I'm not crazy, there's a great TED talk on that phenomenon here: http://www.ted.com/talks/derek_sivers_keep_your_goals_to_yourself.html)  I am hoping to build more info into the box scores and player-level stats this season, though.


1977:  Any predictions on a way-too-early Final Four?

Ken:  No, it's way too early.

1 comment:

  1. Nice interview, some interesting stuff!

    ReplyDelete