What stands out in the first release of ESPN's real plus-minus for the 2020-21 season?
More than a month into the NBA season, RPM ratings have settled down to the point that they become more reliable. Still, the top of the leaderboard (including a familiar face at No. 1, LeBron James) doesn't look exactly like the MVP rankings to date. There are also some surprising players high up in the rankings who might not continue to play so well.
I took a closer look at the ratings for 10 players to help understand what they tell us about RPM and the players in question.
About RPM
Before we get to that, a quick refresher on RPM. First developed by Jeremias Engelmann and Steve Ilardi and now managed in-house by ESPN Analytics with Engelmann having joined the Dallas Mavericks' front office, RPM incorporates a player's box-score statistics, advanced statistics generated by Second Spectrum camera tracking and how teams perform with players on and off the court, adjusted for their teammates, opponents and game state. For more details on how it works, check out the piece I wrote last year about adjustments to RPM and how to best use it.
As part of "RPM 2.0," ESPN Analytics is also explicitly creating ratings that are based solely on on-off data (known as regularized adjusted plus-minus, RAPM) and a statistical plus-minus that does not incorporate on-off data (similar to box plus-minus, albeit with the distinction of including tracking data). I'll refer to both at times as we talk about why players rate the way they do.
Because RAPM tends to require very large samples to stabilize into reliable ratings, early in the season RPM will look much more like the statistical prior. As the season goes on and RAPM becomes more meaningful, it will make up a greater part of RPM ratings.
One last note: ESPN Analytics has calculated RPM back to 1996-97, the first season for which complete play-by-play data is available, and you can now find all these historical ratings on the RPM page.
Now, let's get to the interesting ratings.
LeBron James | Los Angeles Lakers | RPM: +7.5 (1st)
Ordinarily, James topping the RPM rankings would hardly qualify as notable. He has led the league in RPM four times. Yet it's interesting that James only recently began getting buzz as an MVP favorite despite the Lakers cruising to the league's best record and point differential.
Part of that is James splitting credit with co-star Anthony Davis, who actually has the better rating in FiveThirtyEight's similar RAPTOR metric, where James is 14th on a per-possession basis.
RPM favors James for several reasons: He scores well in two important tracking categories, passes ahead (which tend to lead to transition scores) and adjusted opponent field-goal percentage. Per Second Spectrum data, opponents are shooting an effective field goal percentage of 43.8% with James as the nearest defender this season, which is second lowest among players who have defended at least 100 such shots.
Additionally, the Lakers have played better in minutes with James as their only All-Star on the court than with Davis -- same as the 2019-20 regular season, though the trend reversed during the 2020 playoffs. Per NBA Advanced Stats, they have a plus-6.2 net rating in 248 minutes with just James in the lineup as compared to getting outscored by 2.3 points per 100 possessions with just Davis (178 minutes).
CJ McCollum | Portland Trail Blazers | RPM: +7.3 (2nd)
McCollum is a surprising name to see near the top of the RPM leaderboard, but it's really more a case of McCollum performing at a level he has never reached before over an extended stretch of his career than anything unique to RPM.
As I noted in writing about McCollum's untimely foot injury, he led the league in my wins above replacement player (WARP) metric at the time he was sidelined.
Stephen Curry | Golden State Warriors | RPM: +6.4 (3rd)
Because three championships and two MVPs is apparently not a sufficient résumé, this season has inexplicably become a referendum on Curry's value. So far, so good in the eyes of RPM.
Curry actually has the highest statistical plus-minus of any player because of his efficient, high-volume offense without dominating the ball.
Kyle Lowry | Toronto Raptors | RPM: +5.3 (4th)
Given Lowry hasn't performed as well in terms of box-score stats as in 2019-20 on a Raptors team that's off to a slow start, his RPM rating was one of the most surprising to me.
In part, this reflects that RPM 2.0 has highly valued Lowry's contributions at both ends of the court. He has finished in the top 10 every season since 2015-16, including leading the league in 2017-18. Like Curry, Lowry rates well in terms of both passes ahead and generating shots early in the shot clock. He also rates well in the tracking categories that are important on defense, including deferring rebounds (generally a measure of boxing out to allow teammates to secure rebounds), forcing turnovers by his matchup and adjusted opponent shooting.
Kyrie Irving | Brooklyn Nets | RPM: +4.9 (6th)
Although Irving has been on fire this season, making 42% of his 3-point attempts and 58% of his 2-pointers (both career highs), somewhat surprisingly his advantage over equally hot teammate Kevin Durant on a per-possession basis can be traced to the defensive end.
The statistical plus-minus model values Irving's above-average rebounds deferred while docking Durant for his own turnover rate (which tends to create fast-break points for the opposition) and his matchup infrequently turning the ball over. (Cleaning The Glass data shows Durant's teams have typically tended to force far fewer turnovers with him on the court.)
Lonzo Ball | New Orleans Pelicans | RPM: +3.5 (18th)
Given trade rumors and calls to remove Ball from the Pelicans' starting five, it's amusing to note that RPM actually rates him the most effective member of said lineup to date. New Orleans has in fact played better with Ball on the court (his plus-6.3 net differential is second-best among Pelicans with more than 150 minutes of action, per NBA Advanced Stats), but his RAPM doesn't really explain his rating.
Instead, it's primarily about Ball's tracking stats, and in particular his league-leading rate of passes ahead. That definitely helps, as the Pelicans average 2.4 more fast-break points per 100 possessions with Ball on the court. However, that isn't much relief to their struggling half-court offense.
Wayne Ellington | Detroit Pistons | RPM: +3.2 (25th)
As with McCollum, the lofty rating for journeyman Ellington stems primarily from his hot start. He has made 54% of his 6.5 3-point attempts per game so far this season. Only McCollum among regular players has made more 3s per 36 minutes, per Basketball-Reference.com. So Ellington is even higher in RAPTOR, which rates him 14th.
Inevitably, Ellington will cool. Using Kostya Medvedovsky's padding method, Owen Phillips' most recent newsletter found that Ellington can be expected to shoot about 40% from 3-point range for the rest of the season based on his percentage to date and attempts. That's more in line with Ellington's career mark of 38%.
Luka Doncic | Dallas Mavericks | RPM: +1.4 (68th)
Doncic's middling rating was another surprise given he's No. 1 overall in my WARP metric so far. Despite performing well in several key metrics, including the league's best adjusted assist rate, Doncic's offensive rating is limited by his dominance of the ball (he also tops the NBA in time of possession). And his defensive rating is poor due to his high turnover rate and low rate of forcing turnovers by his matchup. (Dallas indeed forces turnovers on just 12.5% of opponent possessions with Doncic on the court, per NBA Advanced Stats, as opposed to 17.2% with him on the bench.)
Nonetheless, Doncic has finished with a better rating than this in each of his two NBA seasons and seems certain to improve over time.
DeAndre Jordan | Brooklyn Nets | RPM: -3.0 (371st)
Jordan's 83% shooting so far this season looks like a typo. Per Stathead.com, Mitchell Robinson's 74% shooting last season was best ever by a player with at least 50 shot attempts, so we'll see whether Jordan can inaugurate the 80% club. Yet he still rates worse than replacement level because of, well, everything else.
Jordan's turnover rate (31% of the plays he has finished) also looks like a typo, and opponents are shooting an above-average 62% when he's the closest defender on attempts inside five feet, per Second Spectrum tracking on NBA Advanced Stats. Given Jordan's important role after the Nets traded Jarrett Allen as part of the James Harden deal, his poor play is a major concern.
P.J. Tucker | Houston Rockets | RPM: -5.0 (447th)
The Tuckwagon has sprung a flat thus far this season. Because of his limited involvement on offense, Tucker has never rated particularly well at that end by RPM but has compensated with strong defense. That has not been the case this season, putting Tucker's rating in the bottom five.
The Rockets have allowed an incredible 21 more points per 100 possessions with Tucker on the court, according to NBA Advanced Stats. That's in part due to fluky 3-point shooting. Opponents have hit 39% of their 3s with Tucker on the court and a preposterously unsustainable 26.5% with him on the bench. Still, contenders interested in dealing for Tucker will have to evaluate whether he's truly experiencing a drop-off in his play at age 35.