Greetings!

It has been a long time. This has occurred for a variety of reasons, but as an apology, I have written a really nerdy NBA article for all of my loyal readers. The crux of this article will focus on analyzing a paper presented at the Sloan Sports Analytics Conference that I have linked to here.

The Study:

Joseph Kuehn, the author of said aforementioned article, attempts to classify what complimentary pieces are crucial to an NBA roster, and assesses their respective values. His article uses data from the 2014-15 NBA season and is broken down into eleven distinct sections. First, Kuehn seeks to understand what is the most likely outcome of any particular possession. The reasoning why will be important later, however, he finds that the answer is a shot within three feet of the basket. This allows the analyst to frame the rest of his discussion.

Secondly, Kuehn uses a mathematical equation to find values in specific possession results. This leads to a few important findings. First, a player that excels at offensive rebounding, or one that averages 10% more offensive rebounds than a replacement level player, only gains his team 4% more than would be expected otherwise. This is because said player takes away rebounds from his teammates. Moreover, this relationship is even worse with defensive rebounds, as the latter number is slightly under 3%.

Afterwards, Kuehn uses his equation to find how valuable certain players are at benefiting their teammates. Here Kirk Hinrich, Joe Ingles, and Wesley Johnson are the best teammates; whereas, on the other hand, Russell Westbrook, DeMarcus Cousins, and Derrick Rose are the worst teammates. Furthermore, in regards to overall teammate statistical output, Jeremy Lin, Jordan Clarkson, Gordon Hayward, Jordan Hill, and Michael Carter-Williams are the best teammates; however, Rudy Gay, Kyrie Irving, Lou Williams, Anthony Davis, and Kevin Love are the worst.

Finally, Kuehn looks at a regression model to determine if teams and management utilize complementary skills to build rosters. Here, examining the 2014-15 free agency as evidence, the author finds that complementary talents are undervalued.

My Analysis:

I feel like it is important to commend the great statistical work Joseph Kuehn engages in for this article. Especially the latter part of the paper, where he examines how salary is spent, as it is very valuable for understanding how teams pursue free agents. From a standpoint of pure rational choice, which is what Kuehn’s statistics examine, teams overvalue individual box score stats while undervaluing a player’s fit on various teams.

Unfortunately, the first three quarters of the article are littered with statistical oversimplifications. I’ll present my critique in two specific ways.

First, the analysis does a poor job of how a team like the Cleveland Cavaliers could be so successful with such “poor” roster construction. Let’s examine the Cavs’ statistics with and without Kevin Love and Tristan Thompson.

Figure 2 Stats via NBAwowy!

And then the points per possession for each bracket:

Figure 2 Stats via NBAwowy!

While I am running no regression, and thus aspects of a teams play such as floor spacing aren’t being considered, Kuehn argues that players like Love and Thompson do not make their teams noticeably better. Consequently, the stats above should not be viewed as gospel in their own right; but rather, a refutation to Kuehn’s thesis. What Figure 2 shows is that the extra 5.3% in total rebound percentage, for the Cavs, leads to an extra .15 points per possession or 14.3 points per game.

Thus, while Kuehn’s analysis works well with other statistical analyses of the value of rebounding – such as this work by Nylon Calculus – it does not account for the fact that rebounding is an efficiency multiplier. Therefore, while Kuehn is correct in stating that rebounding is a fairly simple stat that is easily replicable, he does not analyze how certain players’ rebounding talents combined can vastly improve a team that takes many three-point shots, such as the Cavaliers.

Secondly, Kuehn accusation that players like Russell Westbrook, DeMarcus Cousins, and Derrick Rose are poor teammates is flawed from conception. This occurs primarily because the paper discounts the mathematic value of box score statistics. Consequently, when one examines NBA.com’s Player Impact Estimate from 2014-15, Westbrook and Cousins are both in the top five overall, Kyrie Irving, Rudy Gay, and Kevin Love are in the top 33%, and Michael Carter-Williams, Jordan Clarkson. and Gordon Hayward lag behind the players labeled as “bad teammates.” This is important, because by focusing on the overall result of a play and where it occurs from, Kuehn ignores more time-tested methods that apply his analytical goal to box score stats.

Conclusion:

Essentially, while Joseph Kuehn’s model has many benefits, it does not necessarily contain the explanatory value he posits in his conclusion. The author examines the benefits of rebounding on a singular plane and totally discounts rebounding stats. This creates a group of statistics that ignores the very value Kuehn is seeking to study: what is a player’s value on his team. Just because, as an individual, Tristan Thompson, Kyrie Irving, and/or Kevin Love may be damaging in a vacuum, does not mean they are all at once. The same goes for other players like Serge Ibaka and Russell Westbrook. Consequently, while Kuehn’s ultimate hypothesis about roster construction should be listened to, his methodological understanding of complementary skills needs more work.

LEAVE A REPLY

Please enter your comment!
Please enter your name here