Time for another edition of the NBA mailbag.
Throughout the NBA season, I will be answering your questions about the latest, most interesting topics in basketball. You can tweet your questions using the hashtag #peltonmailbag or email them to peltonmailbag@gmail.com.
This week's edition includes how LeBron James' passing Karl Malone for an all-time record is both bad and good, how different analytical measures help profile players' strengths and weaknesses and why the Minnesota Timberwolves can afford to win a game or two.
"I just saw that LeBron James overtook Karl Malone for the most career turnovers. While this is traditionally a negative stat, these two are obviously all-time greats. My guess would be that having the most career turnovers would mean that you have had a long career and a high usage rate. Is there a correlation between having lots of career turnovers and being an all-time great?"
-- Brett
Your guess is on point. Here are the top 10 all time in turnovers, along with their rank in my wins above replacement player (WARP) metric, which also dates back to when player turnovers were first tracked (1977-78).
The correlation isn't one-to-one -- Michael Jordan ranks just 30th in career turnovers -- but it's pretty good. To find a player on the turnover leaderboard who isn't an inner-circle Hall of Famer requires going down to 13th and Reggie Theus, who ranks 159th in WARP and was a two-time All-Star. Theus is the only player in the top 30 for turnovers who is not in the WARP top 100.
All other things equal, turnovers are a bad thing. As you note, though, all other things aren't equal when it comes to total turnovers because they tend to be associated both with high-volume shot creation and minutes played.
On a per-minute basis, there's a modest positive association between turnovers and player rating because of the former factor. But when you add in the career length component, it becomes a strong one.
In a certain indirect way, then, LeBron becoming the turnover GOAT (greatest of accumulating turnovers?) is a point in his favor.
"With all of the All-Star Game and MVP discussion recently, I've seen a number of articles looking at which players are rated highly across different holistic value stats. What I'm curious about is which players are most uniquely rated by the different individual statistics. In particular, knowing which players are uniquely rated highly or uniquely rated poorly in comparison to the other value statistics might give an idea of the profile of the ideal and anti-ideal player for each statistic."
-- Dave Henderson
Great question. It's important to understand why these metrics differ from each other, and edge cases are a good way to illustrate those differences. I've included eight value metrics, looking at the per-minute/possession version of each:
My WARP stat
Basketball-Reference.com's box plus-minus (BPM)
Estimated plus-minus from Dunks & Threes (EPM)
ESPN's real plus-minus (RPM)
Daily plus-minus from Kostya Medvedovsky's DARKO projection system (DPM)
Regularized adjusted plus-minus from NBAshotcharts.com (RAPM)
The first two of these metrics rely exclusively on box score data. The last one is strictly plus-minus data, adjusted for teammates, opponents and game-state factors. The other five combine box score stats and adjusted plus-minus, making them more comparable. That's clear from looking at how well the stats correlate to each other:
WARP and BPM are easily the most similar of these stats, while RAPM is most dissimilar from all the rest -- although slightly closer to RAPTOR, suggesting it relies more heavily on adjusted plus-minus than the other combo stats.
After normalizing all these stats to put them on the same scales, here are the players who come out relatively best/worst by each of them.
Best by WARP: Robert Williams, Tyrese Haliburton, Zach LaVine
Worst by WARP: Draymond Green, Kevon Looney, Jakob Poeltl
Of these metrics, WARP puts the least emphasis on team defensive performance, which makes it more stable when players change teams but means some of the contributions of a strong defender such as Green aren't captured as well. Naturally, the players who score well by WARP are generally from weak defensive teams.
Best by BPM: Nikola Jokic, Domantas Sabonis, Nikola Vucevic
Worst by BPM: Dillon Brooks, Goran Dragic, Devonte' Graham
The updated version of BPM still seems to overvalue big men relative to perimeter players, as the average center rates 0.4 standardized deviations better than league average. This particularly seems to affect skilled offensive big men such as the trio best by BPM.
Best by EPM: Wendell Carter Jr., DeAndre Jordan, Zach LaVine
Worst by EPM: Furkan Korkmaz, Doug McDermott, Markieff Morris
EPM actually has a higher average rating for centers, although it's not quite as apparent at the extremes as with BPM. On the flip side, EPM's model doesn't seem to value low-volume, offense-first players.
Best by RPM: Lonzo Ball, Devonte' Graham, Kyle Lowry
Worst by RPM: Robert Covington, DeMar DeRozan, Luka Doncic
I wrote about two of the three overperformers when RPM first debuted this season. Covington is a particularly interesting case because he scores well by the RAPM component of RPM. The combo models are essentially zero sum in terms of team success, and RPM ultimately gives the credit for the Blazers' success to their guards, leaving little for Covington and fellow starters Derrick Jones Jr. and Enes Kanter.
Best by RAPTOR: Clint Capela, Jusuf Nurkic, Jakob Poeltl
Worst by RAPTOR: LeBron James, Domantas Sabonis, Russell Westbrook
As noted, RAPTOR tends to rely relatively more heavily on adjusted plus-minus, hurting LeBron, Sabonis and Westbrook relative to their box score stats. Although RAPTOR doesn't have centers as a whole rated as highly as BPM and EPM, defensive-minded bigs tend to rate relatively well.
Best by DPM: Giannis Antetokounmpo, James Harden, Kyle Lowry
Worst by DPM: Chris Boucher, Clint Capela, De'Andre Hunter
There's an important distinction between DPM and the other combo stats: It's designed to project player performance rather than measure value, so it includes performance from previous seasons. Hence the players who rate relatively well by DPM are past standouts, while it's slower to capture improvement from the likes of Boucher and Hunter. This is important to keep in mind when using DPM for awards purposes, although it tends to be closer to how I view All-Stars.
Best by LEBRON: Luka Doncic, Rudy Gobert, Hassan Whiteside
Worst by LEBRON: Joe Harris, Jeremy Lamb, Norman Powell
In addition to Gobert and Whiteside, Myles Turner is also near the top of LEBRON's relative ratings, indicating it places high value on shot-blocking. Conversely, shooting specialists look somewhat undervalued.
Best by RAPM: Willie Cauley-Stein, Pat Connaughton, Thaddeus Young
Worst by RAPM: Bradley Beal, Nikola Jokic, Nikola Vucevic
Because it's based entirely on plus-minus data, RAPM doesn't favor any particular type of player. Instead, the overperformers have seen their teams play much better with them on the court than their box score stats would suggest and vice versa.
This exercise is a good reminder that although value metrics often overlap, it's important to look at more than one to get a full picture of value because each has its quirks and weaknesses, and some players rate very differently depending on the rating system. Just don't let that be an excuse to sift through ratings until you find the one that matches your desired conclusion.
couldn't you argue the wolves have a fair amount of incentive to tank the rest of the season, given dropping to the 4-7 range in the lottery would hurt the odds of keeping their pick, and the team desperately needs an impact player?
— s rosenbud (@rosenbud_) February 22, 2021
-a depressed wolves fan searching for some hope
This question stemmed from my piece on the Minnesota Timberwolves' coaching change, which argued that the team doesn't have to worry about hurting its pick by winning games now. As a refresher, the Golden State Warriors get Minnesota's first-round pick to complete last year's D'Angelo Russell trade unless it lands in the top three of the lottery.
When the NBA altered the lottery process in 2018, I took a look at the impact of those changes on how valuable it is to enter the lottery in each position. Although there is a massive difference in value between the typical No. 1 pick (who produces $30 million-plus in value above and beyond his salary) and the typical No. 5 pick (about $17.5 million in net value), the lottery odds flatten that gap so that the expected difference in net value between having the worst record and the fifth-worst record is barely more than $1 million.
The situation is worse for the Timberwolves because they can't take advantage of the lone remaining benefit for the team with the worst record -- being assured no worse than the fifth pick.
Based on those estimated pick values, here's how much value Minnesota can expect based on where the team enters the lottery.
Because the odds of getting a top-three pick are the same for the bottom three teams, it makes no difference whatsoever if Minnesota -- which has the league's worst record -- passes two teams in the standings.
Passing additional teams would hurt the Timberwolves' chances of landing in the top three, but that's offset by the fact that if Minnesota keeps this year's pick, Golden State gets the team's first-round pick without any protection in 2022.
Naturally, that tradeoff is worth it for a top-three pick in this year's top-heavy draft, but it mitigates the motivation for the Timberwolves to try to lose games for lottery positioning.