Which is more valuable: an offensive rebound or defensive rebound? A missed three or missed two? By looking at the impact of various box score statistics on luck-adjusted net rating, we can get an idea.
Three years ago, Benjamin Morris did a study for FiveThirtyEight on which box score stats were most predictive of a player’s impact on winning.
He found that steals were far and away the most predictive of positive team impact, followed by blocks, turnovers, assists then rebounds. But what is the predictive value of a personal foul? A missed shot? An offensive or defensive rebound?
To come to his conclusion, Morris regressed a sample of individual player box score stats (from every player season between 1986 and 2011 where that player missed and played at least 20 games) onto SRS impact. SRS, or Simple Rating System, measures a team’s margin of victory adjusting for strength of schedule. Since we don’t have access to that sample, or a database of players that includes their net SRS impact, we will use our own base of 400 current players with at least 150 possessions played this season including their luck-adjusted on/off net rating, courtesy of Jacob E. Goldstein.
So basically, our study will use a smaller, but more recent sample and per-100-possessions stats that are adjusted for team/opponent shooting luck. Our results will be a little different, but the general idea the same.
Just like Morris explained in the original study, the regression coefficients of each box score stat – points, steals, blocks, offensive and defensive rebounds, free throws (missed, made and attempted), threes (missed, made and attempted), twos (missed, made and attempted), turnovers and fouls – will tell us how many net points per 100 possessions that stat is likely to produce.
Here are the results:
- Clearly, made shots are very strong indicators of positive team impact. No surprise there.
- A steal or block over a foul – let’s call these “Kawhis” – is the most valuable defensive play included in this study, followed by steals.
- Fouls on their own can sometimes be indicative of defensive activity, which might partially explain why personal fouls don’t have a more significant negative impact on luck-adjusted net rating. But if we count a player’s Kawhis – their defensive plays over fouls – we get an accurate measure of defensive activity and efficiency. As the regression coefficient of 4.5 indicates, Kawhis are valuable not only because they end opponent possessions, but they also often lead to fastbreak opportunities and shots with high expected value.
- (For reference, this season’s leaders in Kawhis: Anthony Davis, Lebron James, Jimmy Butler and two Spurs – Kyle Anderson and Danny Green. All time leaders are Hakeem Olajuwon, David Robinson, Ben Wallace, Andrei Kirilenko and Tim Duncan).
- It’s also worth noting that regressing steals, blocks, Kawhis and fouls onto luck-adjusted offensive and defensive impact separately indicated that these stats have a more significant impact on offense than defense. In other words, they are more likely to lead to efficient offense than they are to limit your opponent’s scoring.
- Offensive rebounds are somehow nearly as bad as turnovers in terms of their predicted impact on net rating.
- This was the most surprising result to me personally. I’ve always thought of offensive rebounds as not only having an impact outside of the box score by deflating opposing defenses, but also as an excellent way to generate efficient shot attempts, as putbacks are the second-most efficient playtype behind transition opportunities.
- This is especially true for individual players with high offensive-rebounding rates, as scheming to crash the offensive glass as a team involves bringing guards into the paint when shots go up, thus conceding transition opportunities to your opponent. Having a high-volume offensive rebounder theoretically allows teams to benefit from this efficient playtype without having to concede as much in transition.
- Diving into the data a little deeper confirms that hypothesis in some ways. I sorted the original sample by offensive rebounds per 100 possessions, and I found the top 100 offensive rebounders have a better median net rating (-0.9) than numbers 101-300. Of the four quadrants, numbers 201-300 had the worst median net rating at -1.8, followed by 101-200 at -1.15. The bottom 100 offensive rebounders were at -0.3.
- The fact that high-volume offensive rebounders generally had more positive team impacts than average-volume offensive rebounders leads me to believe there is still a benefit to having a high-volume offensive rebounder, especially in comparison to scheming for offensive rebounds because teams who scheme to crash the offensive glass (and thus concede more in transition) will generally have offensive-rebounding numbers more evenly-distributed among their players, explaining why average-volume offensive rebounders have the worst median net rating.
- In other words, my theory is that there is greater value in an offensive rebound achieved by having one good offensive rebounder than there is in an offensive rebound achieved by having multiple players crash the offensive glass because there is a lower opportunity cost on defense when you send only one player. But that is mostly conjecture and it might be worth studying separately another day.
- Three-point attempts, even if they don’t go in, help teams more than their two-point counterparts.
- Specifically, two-point attempts being more deleterious than turnovers seems counterintuitive. But high turnover rates are generally indicative of volume playmakers like Russell Westbrook, James Harden or Lebron James, all of whom have unequivocally positive team impacts. High two-point attempt rates are harmful unless they are counterbalanced by efficiently converting them, which is demonstrated by made twos grading as the second-most valuable box score stat in this study.
- Made threes are also more valuable than made twos, so the raw efficiency threshold for a volume three-point shooter is lower than that for a volume two-point shooter, which is borne out both by these results and the fact that, well, shooting 45% on twos is the same as shooting 30% on threes in terms of points generated. You don’t have to make as many threes to be efficient as you do on twos.
- Missed shots have a more positive impact than raw shot attempts.
- Another fascinating result of this regression. Perhaps the soundest explanation is that high-usage players, superstars, generally are less efficient than role players because they are tasked with creating and converting more difficult shots. Thus, they have higher miss rates that correlate with their positive overall team impact.
- The difference in predicted impact between a missed shot and a shot attempt is not all that significant, but these results may indicate the value of having a player that has below-average scoring efficiency but can create and score on more difficult attempts (thus having higher miss rates) versus having players with average scoring efficiency at lower volume.
If there are any sweeping conclusions to be drawn from this admittedly limited study, there are three. First, a missed shot is not just a missed shot. Willingness to shoot threes can change how opponents defend an offensive player, which can have a positive impact on an offense. See: Smart, Marcus.
Second, the value of an offensive rebound can vary widely depending on how that offensive rebound is achieved. Scheming to crash the offensive glass comes at cost. Having a great offensive rebounder does not.
Third, evaluating defense with box statistics can be significantly improved by looking at it from an efficiency perspective and understanding that while steals and blocks do not tell the whole story, they are extremely valuable when counted relative to fouls. The San Antonio Spurs are masters of this, as they consistently have multiple players leading the league in this stat. Clearly, this is no small part of their sustained success, especially on the offensive end.