Tuesday, December 11, 2012

A Closer Look at the "Kobe Assist"

Ten or fifteen years ago, basketball began to undergo a statistical revolution in much the same way baseball had ten or fifteen years prior with Bill James. Over the years, different schools of thought about player value have developed, but there are a few central tenets that most quote-on-quote "advanced statisticians" seem to agree on. One of the biggest such tenets states that points per game are a truly terrible metric for value; good teams do tend to have players that score a lot of points, but so do bad teams, and efficiency is far more important. A related tenet is that generating the most possible points in every possession is more important than generating the most points overall, meaning that if Carmelo scores 35 points but goes 10 for 30 in doing so, his team is highly likely to lose despite his high point totals. In short, the vast majority of players can take lots of shots, and making them is certainly important, but not missing shots is what truly adds value to a team. Missed shots, in all except certain rare end-of-game free throw situations, are bad.

Recently, a Kirk Goldsberry article entitled "The Kobe Assist" generated a lot of discussion amongst the statistical world for attempting to disprove the theory that all such missed shots have equal, negative, value for a team. I encourage reading the entire thing, but for those of you who don't want to/can't access Grantland at work/want to get through this article before your shit is over, Goldsberry's basic thesis is that each player's missed shots are not created equal. In other words, some players are far more likely to get their missed shots rebounded by their own team, leading to (typically) an easy putback. The stats will record this as a miss (a negative event) by the shooter, and an ORB and made shot (positive events) by the rebounder, but since the first shot led to the second, easy shot, some of this positive value should transfer back to the shooter. Unsurprisingly, Kobe Bryant (famous for many reasons, notably large point totals and correspondingly large numbers of missed shots) is the king of this stat, with over 200 such instances over the last two seasons, so Goldsberry names the stat after him.

So, the obvious next question is, does this throw a major wrench into the entire notion of advanced statistics in basketball?

Well, not exactly. The blog "Wages of Wins" responded quickly and decisively to the "Kobe Assist" article with an article of their own that contains some very pointed criticism of Goldsberry's thesis. Again, for those of you currently taking the browns to the super bowl, the author's two primary criticisms are a) Goldsberry does not have a large sample size for his argument, and thus cannot draw meaningful conclusions from his data, and b) Goldsberry gets caught up in the "complexity of basketball" argument and thus leaps to a conclusion which is only moderately supported and he assumes cannot be tested with the appropriate level of rigor. Wages of Wins instead argues that the statistics (at least their advanced ones) already explain 95% of the outcome of any game, and they account quite well for both missed shots and offensive rebounds. While Goldsberry is to be commended for an interesting idea, they contend his argument may be represented in the data, but is a function of randomness over a small sample size rather than a causative trend indicative of basketball success.

Both articles contain strong statistical analysis and make convincing arguments for their own point of view. So who is right? Answering that question completely would require significant time, data, and knowledge of advanced statistics, of which I have probably .4 of the three categories covered. But since the focus here is not on rigorous statistical analysis, my initial impression is that the answer lies somewhere in between the two stances. On the one hand, to argue that a missed shot has more positive than negative value is almost certainly false. If the shot were to go in, the player adds 2 or 3 points of positive value to his team. If the shot is missed, even assuming generously that 40% of the time it generates an ORB and 80% of those ORBs turn directly into two points, the value is still only .64 points, far less than the value of a made shot. Based on these simple metrics, Goldsberry's argument never gets off the ground; any missed shot is clearly worse than a made shot, or even a pass - even the worst offenses still manage about .9 points per possession, greater than the missed shot number above.

However, taking a slightly different perspective on the issue, one can begin to construct some of the underlying principles for Goldsberry's theory. Going back to the first paragraph, I pointed out that efficient scorers (LeBron, Kevin Durant, Tyson Chandler, JaVale McGee) have more value than inefficient scorers (Dwayne Wade, Carmelo, Russell Westbrook, Kobe, Nick Young, etc. etc). But if you look at some notorious "ball hogs" in NBA history, their teams still tend to perform well. In some cases, this can be attributed to other factors (Westbrook plays with the hyper-efficient Durant, Kobe won 3 titles with Shaq), but Carmelo has never missed the playoffs despite the second best player on his teams each year being, by my count, Voshon Lenard, the Camby/Kenyon Martin/Nene pu-pu platter, Andre Miller, old Allen Iverson, older Allen Iverson, Chauncey Billups twice, robot Amare Stoudemire, and Tyson Chandler. Advanced metrics don't do Carmelo any favors and they certainly wouldn't figure that all of these two man combos are playoff contenders. But yet on almost all of these teams, despite Carmelo missing shots, the offense has been efficient enough to make the playoffs even with a generally below-average defense. I can't speculate as to the exact reason for this, but I would guess that when Carmelo's vintage fallaway 20-footers kept going up, his teammates realized going for the rebounds was their best chance to be successful, and Carmelo generated a lot of "Kobe assists" during his career.

Thus, as with most debates in sports, it seems neither side has a 100% convincing argument. While advanced metrics tend not to show it, there does seem to be some value in a player that can attempt a whole ton of shots with reasonable, not excellent, efficiency, if only that their teammates are much more inclined to go after the boards in these situations. If you take a more holistic view of Goldsberry's thesis, there does seem to be value (not as much as Stephen A. Smith would have you believe, but some value) in gunners and players who somehow generate these "Kobe Assists." It will be interesting to see if the favorites of this metric (Kobe and Carmelo, along with Derrick Rose and Dion Waiters) continue to put up these numbers over the next 10 years as the data continues to be tracked. If so, then Goldsberry's argument about this sort of value for specific players will carry a lot more weight. If not, well, we are left with the conclusion that all missed shots are still equal, and still bad.

In any case, I'd still rather have LeBron.

Please contact the author of this post with questions or comments at pabritton42@gmail.com.

1 comment:

  1. I don't think Goldsberry was arguing a Kobe Assist is better than a MADE basket, but that certain kinds of misses are more valuable than other kinds of misses. So he's would say that missed shots are indeed bad, but that not all missed shots are equally bad.

    To me, that makes intuitive sense, with the caveat that which shots are bad depends upon the interaction between the shot's rebound AND the positioning of the offensive players. I don't think he gives the second factor enough credit.

    He acknowledges that who your teammates are has an effect on this (ie, Kobe played with Shaq and Pau--maybe that's why), but ultimately seems to think the player has more to do with the "Kobe Assist" than the team does. I think he left it way short there, especially since we know that certain teams have coaches who prefer them to get back on D than crash the offensive glass (I believe Doc Rivers falls into this camp in an extreme way, though I might be misremembering).