It’s an important question to ask: do empty statistics exist in basketball? Short answer: no… per se. However, you’re reading this with the intention of me expanding on the topic, so I will continue. This concept of production not leading to value on the court or team wins is a fascinating one. Whether you want to frame the discussion harshly by saying a player has “empty stats” or the more polite way of a player has “good stats but on a bad team,” determining if box score statistics paint an accurate picture of a player’s value on the court is not only a discussion worth having, but also one that is ever present in online NBA communities, sports talk radio, and everything in between.
Every team has that guy who fans point to and say something along the lines of “his numbers are empty and don’t contribute to winning.” DeMar DeRozan gets the brunt of this by a segment of Raptors fandom. Despite averaging 23.0 points, 3.9 rebounds, 5.2 assists, and being named an All-Star last season, Toronto was +9.9 with DeRozan off the floor compared to +7.2 when he was on the floor, an on/off differential –2.7, per Basketball-Reference. For the Knicks, the empty stats target is zeroed in on Enes Kanter, a player whose on/off differential was –2.3 as a bad Knicks was less bad with him off the floor.
Unfortunately for Kanter, he was the inspiration for this article. When you look at Kanter’s box score stats from last season, they look quite impressive: 14.1 points on 63.0 true shooting percentage, 11.0 rebounds, and a career-high 1.5 assists. His per-100 possession numbers are even more impressive, as he averaged 27.1 points and 21.1 rebounds. How could anyone say with a straight face that a player who averages an efficient double-double is not providing value, not contributing to winning, and is putting up empty statistics?
Well… Enes Kanter, despite averaging an efficient double-double, does not provide value, does not contribute to winning basketball, and puts up empty statistics… sort of.
“Empty stats” is a misleading term. The issue isn’t that a player’s box score statistics don’t have value — they do and contribute to winning basketball games — but rather a player is performing so poorly in other facets of the game that it is taking away the positives of putting up good basic box score statistics. This is where DRE enters the programs.
DRE is an acronym for “Daily RAPM Estimate.” The metric was developed by Kevin Ferrigan back in 2015 and then updated in 2017. His goal was to track game-to-game performance fluctuations by determining what weights to place on specific box score metrics. Ferrigan’s methodology is somewhat of a raw, stripped-down version of Basketball-Reference’s box plus-minus. He ran a linear regression of per-100 possession basic box score stats against 14 seasons worth of multi-year RAPM.
Ferrigan reworked the regression’s explanitory variables a few years later and settled on the following: points, two-point attempts, three-point attempts, free throw attempts, offensive rebounds, defensive rebounds, assists, steals, blocks, turnovers and personal fouls (sound familiar?). The new, reworked regression was statistically significant and had an adjusted R-squared value of 0.5327, meaning that those basic per-100 possession box score statistics explain 53.27 percent of the variation of RAPM. It also means that 46.73 percent goes unexplained, but more on that later.
All but offensive rebounds and personal fouls were statistically significant at the 99th confidence intervals — and guess who happens to be atop the list of the one stat that is statistically insignificant. Do I really need to write who it is?
To calculate a player’s DRE, you use per-100 possession values in the following equation:
The sum of this equation is a player’s estimated RAPM based on per-100 possession box score statistics. Below is a table consisting of a number of different metrics for the Knicks this past season, including and sorted by DRE. Players with less than 1,000 possessions were removed. Sorry, Kuz.
You see it too, right? Enes Kanter has the second-best DRE and the worst differential on the Knicks last season. I know you’re asking yourself, “what does this mean, Drew?” and I’ll tell you right now: Kanter’s per-100 possession basic box score stats are overestimating his actual RAPM by a noticeable amount. Kanter goes from the Knick with the second-best estimated RAPM (or DRE) to a player who has a negative impact when he’s on the court. His box score metrics do not accurately predict his RAPM.
I completely understand that an adjusted plus-minus metric like RAPM is not the end-all be-all — I know you’re thinking this right now (probably… maybe). Reducing a player’s contribution to winning into one statistic is quite reckless.
With that said, RAPM is a foundational metric with NBA advanced analytics. Statistics like box plus-minus and player impact plus-minus are based off it. The metric is one of the best predictors in a team’s future performance. Its multi-year version is even more powerful for predictions and stat creation. There are no box score components to its calculation, which allows for analyses to be performed in order to measure the impact of different metrics. RAPM does provide legitimate value that cannot be ignored or overlooked.
Kanter is an excellent example for why individuals should not just look at the basic box scores and come to a conclusion about if a player is “good” or is a positive on the court. His stats are not empty, but rather his poor defense, lack of passing, and how his shot profile clogs the lane do not allow for his efficient, restricted area scoring and rebounding to add value nor be a plus on the floor. While the later to issues are important, Kanter’s issues are primarily defense related:
He’s impacting the game negatively in ways that a box score simply cannot capture, hence why there is still much left unexplained (46.73 percent) in the variation of RAPM.
An estimate like DRE cannot fully register that 87.3 percent of Kanter’s two-point attempts came within 10 feet of the rim and simply cannot capture the nuances of defense. Sure, capturing steals and blocks is important, but defense is so much more (we can debate whether or not individual defensive rebounding should be considered good individual defense another day). How well was the pick-and-roll contained? Were the help rotations correct? Did the switch make sense? Was the defender in a position to also defend the drive as he chased his man off the three-point line? Etc. Plus, defended field goal percentage differential within six feet with some added context for volume is a much better metric to measure rim protection as it also takes into account blocked shots.
Nevertheless, a box score estimate of RAPM and calculating the differentials is a useful way to begin asking questions about if a player’s basic statistics are indeed contributing to winning basketball. You can go in the opposite direction and look at a player like Frank Ntilikina and ask “how does he have a positive RAPM despite his box score numbers being trash?” No matter how you use these DRE differentials, make sure you follow up with some additional research to see what is going on. They are definitely useful.
Final Note: If you’re unfamiliar with RAPM and don’t know why I like using it and RPM in my articles/work, I would suggest reading this article and this article to help get a better understanding of how these adjusted plus-minus stats function and how they are calculated.