Yeah, they do advanced statistical analysis in basketball as well.  You might call it basketball sabermetrics, but Eli Witus from Count the Basket prefers “APBRmetrics.”  Basketball has its own numbers that in which it is awash, but are those the right numbers by which to assess a player or a team?  I “sat down” with Eli to ask a few questions about how it’s done in basketball and what Sabermetric folk might learn from looking at the numbers in another sport.
Tell us a little bit about Eli the basketball (and sports in general) fan. Where did you grow up and who were your favorite players growing up?
I was born and raised in Ann Arbor, Michigan, so I’ve always been a big fan of the Detroit teams and the University of Michigan teams. Basketball’s my favorite sport, but I’m also a big football and baseball fan. I started getting really into basketball when the Bad Boys won back-to-back titles and then during the Fab Five years at Michigan. My favorite player in any sport will always be Barry Sanders. When I was younger some of my other favorites were Cecil Fielder and Dennis Rodman.
I don’t recall exactly but somehow I found the APBRmetrics Yahoo discussion group in 2002 or 2003 and went from there. From there I started reading the work of Dean Oliver, John Hollinger, Kevin Pelton, Ed Kupfer, and others. My interest grew and eventually I began posting some of my own stuff.
Baseball is, in the end, mostly a series of individual confrontations between pitcher and batter. Basketball is a 5-on-5 team effort. How has that shaped the types of analyses that you and other folks have done?
That is the main challenge in basketball analysis. Context matters a lot, but we can’t say exactly how or in which areas it matters most. In evaluating player performance, one of the techniques adopted from hockey has been to use the team’s plus/minus (point differential) while the player is on the court as a way of rating the individual player. Team interactions also mean that in basketball roles are very important – not all players have to be scorers, and in fact having too many scorers on the court at the same time can hurt the team. Dean Oliver has looked at the relationship between player usage (how often they possess the ball – which obviously depends on the coach’s system and which other players they are on the court with) and player efficiency (the rate of converting those possessions into points). I’ve done some work looking at the relationship of player rebounding and team rebounding. Another approach that I hope to look at more is to examine how various player statistics change when players change teams, as this can help give a sense of which areas of the game are most context-dependent.
What are some of the discoveries that APBRmetrics have found about which the public might not be aware? Are we watching basketball all wrong?
I’m not big on the “debunking conventional wisdom” aspect of statistical analysis. Probably because I think most conventional wisdom is right, and even if some of it is wrong, I don’t think advanced statistical analysis of basketball has progressed to the point where it can claim that anything is definitively false. That said, in my view the biggest advancement in basketball analysis was looking at the game on a possession by possession basis, something that Dean Oliver deserves a ton of credit for (though he wasn’t the first to come up with it). It’s still commonplace to hear or read a basketball analyst say that a team has a great offense because they score a lot of points per game, or a bad defense because they give up a lot of points per game. But sometimes these results are more due to a team playing at a fast pace (more possessions per game) rather than them having a great offense or bad defense. The same thing can happen in the opposite direction, with a team not scoring many points per game primarily due to the fact that they play at a slow pace. By now these ideas are bedrock in the statistical analysis of basketball, but they still are catching on in the popular media and I’m not sure how much casual fans of the sport are aware of them.
Is there a player in the NBA whom you think doesn’t get his due with the traditional basketball stats?
Well, the glaring weakness in traditional basketball stats is the paucity of stats that measure player defense. Unfortunately the advanced stats that people have come up with don’t do much better in this area. Good perimeter defenders like Bruce Bowen or Shane Battier contribute a lot to their team that isn’t captured in their basic boxscore stats. But I should point out that these kind of players aren’t a “discovery” of advanced statistical analysis – scouts, coaches and general managers know they have value even if there’s no points per game type stat to back it up.
How much crossover do you think there is between the APBRmetric readers and the SABRmetric readers and researchers?
I don’t think there’s that much, actually. I think people tend to stick to their own sport. For me personally I’m a baseball fan and a fan of statistics (obviously), but I’ve always preferred to watch and follow baseball without getting into the sabermetric stuff. I do read a number of sabermetric sites and books now, but that’s mainly to get ideas that I think could transfer over to analyzing basketball.
Got a recommended reading list for someone wanting to know more (other than your blog, of course)?
Sure. Dean Oliver’s book Basketball on Paper is the place to start. The best place to find research and analysis is the APBRmetrics forum.  82games has a great collection of interesting stats and analysis.  If you have ESPN Insider John Hollinger has a lot of good articles there (he used to publish an annual guidebook but no longer does).  Beyond that you might want to explore the links page on my site, which is an attempt to collect and organize all the basketball statistical
sites on the web (unfortunately I haven’t been able to update it as frequently as I would like).
Better posts, more frequently. But seriously, I have a ton of ideas for research that I want to conduct and post about. There are new statistical techniques I want to try, new areas of the game I want to explore, new data sources I want to investigate, and on and on. I don’t really have a plan other than trying to work on things that interest me and that I think would interest others.
Thanks to Eli for taking the time to answer my questions.

1. My favorite basketball metric is the adjusted field goal percentage that takes 3-pointers into account and weights them heavier.
I definitely agree with Eli’s assertion of “sticking to your own sport.” I haven’t missed a Sixers game since February of 1998 but I’m not tremendously interested in doing APBRMetrics; I think they’re great to see but since I do so much with baseball I like that others are doing it.

2. dan says:

I’ve always thought that if we were to analyze football then it could approach the level of baseball analysis that is done today. First thing that comes to mind is a WPA stat for football which I think could be created if the play by play data were available (I don’t know if it is or isn’t, or when it began being tracked).
I was tossing around a WOWY approach to evaluating linemen, but I haven’t thought it through really and it’s entirely possible that it won’t be accurate or meaningful.
I’ll probably think of something to add to this later on in the night.

3. dan says:

Oh and I think someone out there can probably think of a more logical system for evaluating quarterbacks than this monstrosity…

4. dan says:

okay the link didn’t work… here it is:
http://www.nfl.com/help/quarterbackratingformula

5. Dan, I spoke to a statistics class at PSU before graduating this past semester and one of the students involved stays in touch with me; he is extremely interested in developing advanced football metrics and the play by play sounded like it interested him most.
The WOWY approach sounds cool, too.
Who would you debit on certain plays, though, in the WPA? On a reception by the WR, do you split between he and the QB and debit the CB?

6. Pizza Cutter says:

Eric, look up “the football project.” They’re basically trying to start a Retrosheet for football. Not sure how far along they are.

7. dan says:

I’m not really sure who would get credit for what. I havnen’t spent much time thinking about this stuff for football, it just seems like it should be possible.