Bruin Sports Analytics
Defensive Deterrence II: Player Analysis
By: Ian Geertsen
So, how does our sample of players shake out according to this metric? Let’s take a look:
Coming in at number one on our list, we have…Jonas Valančiūnas. Ummm, okay? Occupying the next five spots, however, are Rudy Gobert, Jakob Poeltl, Brook Lopez, Myles Turner, and Clint Capela…that sounds more like it. But going back to Valančiūnas, why does this metric value him so highly? For starters, Valančiūnas registered the second best opponent shots from the paint (non RA) value out of the entire sample, coming short only to Draymond Green—these values are based on player-to-team on-off data and do not factor in team responsibility adjustments. He also finished fourth in opponent mid-range shots allowed and fifth in opponent free throws attempted, although he finished just ahead of average in opponent shots from less than five feet and from the restricted area (remember that this is compared to the sample of NBA players I selected, not the league as a whole). This shows that, despite the fact that Valančiūnas didn’t stand out in what I deemed were the two most important categories, he was able to gain value in the metric by shining in other areas. While each of these categories was assigned a unique weighting in the overall formula based on their relative importance, it can be helpful and interesting to see how the players in this sample are valued by each individual part of the metric.
The shots <5ft category registers a variance of 0.0126, squarely in the middle of the five categories; the shots from the restricted area category has a variance of 0.0069, good for the second lowest value. These variance values tell us how volatile each category is, and demonstrates the magnitude with which changes in performance are projected onto changes in value for the metric as a whole. Rudy Gobert, Aron Baynes, and Myles Turner all make their way into the top five of the above categories, although interestingly enough Baynes only ranks 23rd in the overall metric despite this impressive valuation. Similarly, these two categories both see James Wiseman, Derrick Favors, and Nerlins Noel as bottom five talents among this sample, with Jarrett Allen and Kristaps Porziņģis not far behind.
The shots from the paint category’s variance is 0.0246, the highest variance of any of the five categories. This shows that this category is the most volatile, and that changes in this category are more likely to have a larger effect on the metric’s overall valuation as a result. The opposite can be said for our next category, as the mid-range shots category actually registers the lowest variance with a value of just 0.0034. We can really see Valančiūnas’ breakthrough in these two categories—very different from his 20th place ranking in both the <5ft and shots in the restricted area categories. Taking a detour from his performance in the previous categories, the shots in the paint (non RA) category values Aron Baynes as the second worst player in the sample; his data shows essentially the opposite pattern that Valančiūnas’ exhibits.
With a variance of 0.0170, the free throws attempted category ends up with the second highest variance among our five categories. The top five players according to this category’s valuation—Lopez, Serge Ibaka, Capela, Gobert, and Poeltl—are all strong rim protectors and impactful defensive centers, which I hope speaks to the validity of this category. I questioned the credibility of this category the most out of the five, so this is good news.
I included two players in my analysis who were traded halfway through the NBA season, and analyzed their performances on both teams to possibly see how context can affect a player’s defensive deterrence valuation. After spending eight seasons with the Magic, Vučević was dealt to the Chicago Bulls before the 2020-2021 trade deadline, playing 44 games with Orlando and another 26 with Chicago. During his time with Orlando, Vučević registered a defensive deterrence score of 5.41, and he was able to top this score with a value of 5.17 during his games in Chicago. On the flip side, Kelly Olynyk saw his defensive deterrence value shift from 5.85 to 6.01 upon his move from Miami to Houston—Olynyk played 43 games with the Heat and 27 with the Rockets. Obviously there are a limitless number of variables to consider surrounding these numbers, including injuries, motivation, your role on the new team, your relationship with your teammates, the quality of your pregame PB&J sandwiches, the list goes on and on. In the midst of all these confounds, however, we can see that both Vučević and Olynyk registered better defensive deterrence values when playing on better teams. Olynyk saw his deterrence score worsen after he went from the Heat, who would go on to win 40 games, to the 17-55 Rockets—good for dead last in the NBA. Similarly, Vučević saw his valuation improve after leaving the bottoming-out Magic (21-51) for the slightly greener pastures of the 31-41 Chicago Bulls. It also helps that one player went from the better team to the worse team and the other from the worse to the better, as this helps us account for ordering effects (yes I know it’s a small sample, but it’s what we’re working with).
Does this mean that these player’s performance was better on the better team, or are these players being assisted by the play of more capable teammates? While this is really hard to say on the individual level, we can get some insights to this question by looking at the team level. Among the players in my sample, defensive deterrence and games won by each players’ team has a correlation coefficient of just 0.075, suggesting that how good your team is (how many games they win) does not seem to matter much when it comes to your individual performance in the defensive deterrence metric.
When looking more specifically at team defenses we again see very little correlation, this time between player’s defensive deterrence values and their team’s adjusted defensive ratings. With a correlation coefficient of just 0.031, the data would suggest that the overall defensive quality of your team is not highly associated with how well you perform in the defensive deterrence metric.
Going back to Olynyk and Vučević, we can also observe how their respective trades affected their performance in each individual category of the metric. Looking at this data, we can see that both Vučević and Olynyk performed better in the shots from less than five feet, shots from the restricted area, and free throws attempted categories when on the better team than when on the worse team. They also both performed worse in the shots in the paint category when on the better team than when the worse team; the only instance where this pattern does not align is in the shots from mid-range category, where Vučević was valued more favorably ad Olynyk less favorably when on their better teams. While the sample is too small and the changes too negligible to suggest anything significant, I do find these results intriguing.
It is also interesting to note that, for both players, the largest change in any individual category as a result of switching teams came in the free throws attempted category. I am unsure if this speaks to the volatility of the category, the heightened impact that one’s team plays on opponent free throws specifically, or if it’s just due to random chance. All in all there isn’t enough evidence to say that one’s team plays a significant role in affecting one’s defensive deterrence, although if additional data were to look similar to the case studies of Vučević and Olynyk, that conclusion could change.
The Role of Roles
As we’ve mentioned before, the defensive deterrence metric is designed to quantify the value of a specific type of defense, not of defensive impact in general. In order to test the reliability of this metric, and to test the repulsing ability of other kinds of defenders, I added a few players who don’t fit the bill of rim-protecting bigs to the sample. If these players perform extremely well, it will likely speak to some error I made when developing the formula, as wings in general should not have as large of an impact on deterrence as bigs. As long as this is not the case, though, we could say that some non-bigs provide deterrence value when on the court despite the fact that their role likely does not call for this most of the time. Here’s the data of some of our non-traditional defenders:
This short list is comprised of some of the league’s top defenders, as well as some of the leagues not top defenders, to put it nicely. Among these seven players Anthony Davis immediately stands out, ranking 51st in the overall sample and being the only player of these seven to have a defensive deterrence value in the fives. While it is true that Davis did play some small ball center for the Lakers during the 2020-2021 season, basketball reference estimated that he logged 91% of his minutes with a traditional center on the floor. Still, Davis’ role in the Lakers’ defense was centered much more around rim protection than some other players on this list. The same could be said for Marvin Bagley, who is more of a traditional four compared to some of the wings on this list. Behind Bagley we have the clump of Siakam, Randle, and Williamson who all were valued very similarly, and behind them we have OG Anunoby and Ben Simmons. Based on the ordering of this list, it appears that a player’s role in the defense has a very large impact on how they were evaluated by this metric. But what if we could analyze these players without including the context of their roles? By looking at defensive deterrence without the team responsibility adjustment, we can get a glimpse of what that might look like:
While all of these players moved up compared to the overall sample—as we would expect, as they all were devalued by the team responsibility adjustments—Simmons and OG remain at the back of the pack. These two are unquestionably good defensive players, although they spent the most time defending on the perimeter out of this group of seven. This shows that perimeter players continue to score poorly in the metric even after removing the responsibility adjustment from the equation, which to me speaks to the validity of the metric (if we removed the responsibility adjustment and these wings started to look like the best deterrers in the game, this would’ve exposed an issue in the metric). The fact that Williamson’s unimpressive defensive season lands him at the top of this list, however, shows that my formula is far from perfect. We can also further break this list down to see how these players performed in each individual category:
As you can see, these non-traditional players’ average ranks were pretty consistent across categories with the exception of the opponent shots in the paint (non RA) category, where these players were ranked significantly better than the others. This once again speaks to this category being the most volatile and least accurate. This also helps explain how Zion ended up at the top of the previous list, as he finished fourth among the whole sample in unadjusted opponent shots in the paint.
All in all, while the sample we used wasn’t very large and we therefore shouldn’t take too much away from this data, this analysis showed that the role you have does likely affect the impact you have on your team’s defensive deterrence. This analysis also made apparent the significant role that the team responsibility adjustment has on the overall metric, although admittedly it should cause the most change among this group of players as they theoretically should contribute the least to protecting the rim and deterring offensive shooters. When looking just at the team responsibility adjustment, we can see that Simmons, Anunoby, Williamson, and Siakam all finished within the bottom five of team responsibility, as they likely should.
See Part III for the conclusion of this article.
Sources: NBA.com, basketballreference.com, bball-index.com, espn.com.