4-Down Episode 14: Introducing Scott Sereday
The 48MoH crew is happy to announce the newest member of our team, Scott Sereday.
Scott may be a new name to some of you, but many will recognize him from his contributions to the APBRmetrics community and his own site, Basketball-Analysis.Â Heâ€™s currently enrolled at Columbia pursuing an M.A. in statistics.
(Here’s an easter egg to round out the edges. You can be the judge whether that is virtue or vice on display, but it’s not bad for someone who stands 5’11”.)
In addition to our 4-Down podcast with Scott (located at the bottom of this post), we conducted a short interview to give you a better sense of his work and what you should expect from him in the coming months.
48MoH: Greetings, Scott. Tell us about your statistical work.
SS: I currently run a site called basketball-analysis.com. My work includes expected championships added, clutch adjusted plus-minus, estimations of assisted % and efficiencies using box score statistics. In addition to this, Iâ€™ve done lots of analysis of the play-by-play data available at basketballvalue.com and the utilization of statistical methods such as regression analysis.
48MoH: Iâ€™m already lost. Letâ€™s try this a different way. What, in your estimation, is the value of mining data for new ways to see the game?
SS: Well, as humans we typically do a good job of innately telling if something is helpful or harmful towards achieving an ultimate goal. But determining magnitude is a very crude process. I think we can all agree that Tim Duncan creates more wins than Kwame Brown. But without quantitative analysis, how can we tell how many wins this difference amounts to? Additionally, subtle statistical patterns are typically not noticeable to the naked eye. I believe it was Bill James who once stated that one would be unlikely to visually distinguish between a .300 hitter and a .270 hitter.
48MoH: Quantitative analysts such as yourself often talk about the box score as if itâ€™s an embarrassing relic of the past. Why is the box score such a limited (misleading?) rubric for player evaluation?
SS: Many have said before that basketball is behind other sports such as baseball in terms of tracking statistics and the availability of more detailed statistics to the public. Itâ€™s clear to see that a large portion of defensive plays are excluded from the box score. The play-by-play sheds some light on other statistics, such as detailed shooting and situational statistics, but even that misses some telling offensive statistics and struggles to improve the picture of individual defenders.
48MoH: To ask the previous question a little differently, why is play-by-play data useful in helping us understand what happens on the court?
SS: Play-by-play details help determine nontraditional – but still crucially important – statistics such as how many assists are on close field goals and how dependent a player is on other players to help create his shots. For example, Erick Dampier appeared very efficient over the last few years. But the large portion of his scores are assisted close shots, which rely more on the passer than on the scorer, and need to be made at a higher rate than other attempts because of the inherent risk of the entry pass. Dampier shot over 63% from the field over the last 4 years. Sounds great, but consider that over 60% of his made field goals are assisted close attempts. I estimate league average FG% for this type of shot to be around 70%, which put his numbers in a more appropriate perspective.
48MoH: There is another side to this, that of the statistical absolutist. This sort of person — one whose dogged adherence to hard data suggests that findings never err — often turns people off to the work of other quants. You still see value to what we might call a coachâ€™s or scoutâ€™s gut-knowledge, right?
SS: I think the problem is that both sides often try to oversimplify things so they can imagine that they have a complete understanding of everything involved. People who rely heavily on observational intuition may try to say that statistics donâ€™t mean anything or that any statistical analysis they donâ€™t understand is nonsense. But they often take this approach in order to make their opinions seem more valid to themselves. Similarly, people who rely too heavily on statistical analysis often make the mistake of assuming that everything that goes into that data can be fully digested by a simple study. If I beat you in a hand of poker, it doesnâ€™t mean that I am a better poker player than you, it just means I am slightly more likely to be a better poker player than you. Both sides need to remember that my victory isnâ€™t meaningless, but we have to be careful not to draw too many hard-and-fast conclusions from it. The numbers donâ€™t err, they just might not mean what we think they mean.
48MoH: One thing I like about your work is that you recognize your community is not above correction. For example, youâ€™ve recently pointed out that it isnâ€™t always the case, as models such as PER indicate, that fouls represent a negative value. In some circumstances, fouls are good. What do you think are the major shortcomings of the current quantitative analyst crowd?
SS: Man, this is a tough one. I guess it depends on what aspect you want to look into. In terms of running a team, itâ€™s kind of tough to tell because NBA teams are secretive about their data. Iâ€™m sure there is also much data that is not easily available to everyone.
In another aspect, I think itâ€™s interesting that you donâ€™t see any quant guys as talking heads on TV or radio. This might be because their message may not be as readily received by the common fan or because they arenâ€™t as good at making knee-jerk reactions.
48MoH: What first attracted you to the APBR community — perhaps a particular thinker or article? Not that Iâ€™m a quant, but I remember, years ago, getting my mind around the very simple distinction between rebound average and rebound rate. That alone was enough to bend my ear toward statistical refinement.
SS: Iâ€™ve always been someone who wanted proof before accepting something as fact. Just because everyone else held a certain viewpoint was never enough to me. So when I was young, I tried developing my own formulas even when I had no idea anyone else had already been doing the same. Some inspirational names include Dean Oliver, Ed Kupfer and Dan Rosenbaum.
48MoH: What can our readers expect from you this season? Anything fun waiting in the wings?
SS: In addition to applying some of the work I do on basketball-analysis.com to player statistics for 48MoH, I will develop new measurements, such as adjusting offensive ratings to use more detailed statistics to create a more appropriate representation of players like Dampier. (See my articles on estimating offensive efficiencies by shot type: created and assisted and shot location). I will manipulate the play-by-play data to come up with specific clutch statistics or fast break statistics that I believe are useful and interesting. I also hope to come up with improved defensive metrics.
I even have some thoughts on ways to improve the NBA. Some of these studies will take longer to complete, but look for me to write at least weekly regarding lighter subject matter like evaluations of Spurs personnel and other observations pertaining to the basketball world. Also, look for a statistical season preview in the coming weeks.
48MoH: Well, that sounds fun for me and exhausting for you. Thanks for your time, Scott. Weâ€™re all excited to read your insights and, hopefully, watch the Spurs with more intelligence because of it.
And here, dear readers, is your big bag of podcast: