Clippers and Real Plus-Minus

On Monday, ESPN brought adjusted plus-minus into the mainstream, a refined variant designated Real Plus-Minus (RPM), developed by Jeremias Englemann and Steve Ilardi. Adjusted plus-minus looks at each matchup between units in every game of the season and uses a regression to isolate the players from who they play with and against.

Since the initial iterations, developers are using more powerful regressions and working to account for lots of other variables, like coaching, aging and score effects. RPM is the most modern instance, a further development of Englemann’s xRAPM. Analysis in 2013 by now FiveThirtyEight contributor Neil Paine found xRAPM to be much more accurate than all other box score based advanced stats.

RPM isn’t the singular “golden stat,” the largest issue being sample size, but it does look at player value from a vastly different angle than other catch-all stats that have come before it. Defense is accounted for as half of the game. And all the gritty, clichéd things that basketball pundits love are finally a part of the conversation. And again, it’s more expansive than anything that’s come before in the public domain.

Let’s look at few of the more interesting Clippers via RPM:

J.J. Redick (-0.16)

The case of J.J. Redick and RPM is a fascinating one. He’s clearly a good offensive player. He has a +2.93 Offensive Real Plus-Minus (ORPM) this season and was +2.40 last year. But he might be the player whose defensive perception diverges mostly widely from all available evidence, especially among the community of smart basketball people.

J.J. Redick was not a credible defender at the start of his career. But maybe to counter the public’s overreaction, maybe because somebody saw it in a few games, maybe because of a few Stan Van Gundy quotes, the narrative of Redick as a good “team defender” started in the basketball community. And once something like this starts, especially on the data-lacking defensive end, it’s hard for much to get in the way of confirmation bias.

But RPM does, and it provides reasonable evidence that Redick is not quite the defender we think. He’s has a -3.09 Defensive Real Plus-Minus (DRPM) this season, but that could be dismissed on flukey small sample sizes and injuries. The issue is, lots of previous seasons corroborate this year’s estimate. His DRPM was -3.60 last year (another peculiar season of being on a terrible team coupled with a trade), -2.80 in 2012 and 2011 and -2.40 in 2010. Those years in Orlando with Van Gundy and Dwight Howard are the most concerning.

Sample size and variance from season to season is an issue with RPM, but when it points to something year after year after year, it’s pretty hard to dismiss.

DeAndre Jordan (+4.27)

The first big surprise of the RPM list may be Jordan, who’s estimated to make a larger per possession impact than any other center in the league. Jordan has the second highest ORPM among centers of a modest +1.00, likely due to his efficient finishing and the way his lob threat opens up the floor for shooters. Jordan only rates as the 18th best center defensively–between Robin Lopez and Anderson Varejao–but centers make such a big impact defensively, that along with his good ORPM score, DeAndre rockets up near the top of the overall rankings. Jordan’s defensive impact (+3.27) is suggested to be about the same as the offensive impact of Carmelo Anthony.

Blake Griffin (+3.97)

The most surprising thing about Blake Griffin’s low ranking is his smaller than expected ORPM. It may just be a function of sample size–Griffin’s elevated play has been relatively recent–and data from previous seasons is taken into account, regressing his impact this year onto prior expectations. That’s another thing the RPM framework can struggle with: age-curve defying sudden leaps in player performance. The good news for Griffin is that DRPM sees him as a pretty average defender for a power forward, not the minus public perception seems to label him as. That’s fine for the Clippers as long as he’s playing alongside Jordan.

Matt Barnes (+3.36)

Barnes has long been heralded as an underrated player based on his defense, but RPM delivers another surprise here: most of Barnes’ impact is estimated to come from the offensive end. Small forward as a position is fairly weak offensively, but it’s still a shock that his ORPM is fourth in the league at the position. I’m still not sure what it is, exactly. His shooting? Screening? That’s probably my favorite aspect of RPM: unlike boxscore based advanced stats, it can hint at things we don’t realize about players and spark further investigation. Again, RPM has flaws, but is based in lots of data. Barnes probably isn’t the fourth best offensive small forward in the league, but his ORPM being there can help prompt further investigation as to whether we’ve been missing some aspect of his offensive value.

Chris Paul (+7.58)

Blake Griffin has gotten buzz for the three spot on MVP ballots, but RPM reaffirms that Chris Paul is still the best player on the Clippers. By a wide margin. RPM loves Paul, and he’s been no lower than third in the league over the last three seasons. Paul ranks as the fourth most valuable offensive player in the league, behind LeBron James, Kevin Durant and Stephen Curry. He may not be quite warrant another first team All-Defense selection (that would be…Eric Bledsoe), but CP3’s no slouch on that end either, ranking fourth in DRPM among point guards.

Ryan Hollins (+0.19)

Sure, he’ll jump for any pump fake and doesn’t have stellar positioning, but Hollins is still a solid defensive player. That may be one of the biggest takeaways of adjusted plus-minus research–if you’re seven feet with a pulse, you’re going to have a pretty big impact defensively.

Just because Hollins has a higher DRPM score doesn’t mean he’s a better defensive player than Jordan, though. There are important role caveats to consider. Hollins averages 6.5 fouls per 36 minutes, meaning he’d foul out long before playing the 100 possessions DRPM is suggesting he’d contribute to. Hollins can also probably focus more exclusively on the defensive end than Jordan. And RPM highlights that exclusive focus too, as Hollins has a -3.72 ORPM.

***
Again, RPM isn’t a magic bullet stat and there are lots of qualifiers that come with it. Accounting for role is still an issue. Previous years improve predictive ability, but don’t necessarily reflect what has happened this season. And sample size can still screw things up.

But it’s unique in how it’s calculated, actually showing new information and challenging preconceived notions of player value. Thinking of it as another flawed end-all, be-all stat does a disservice to the new ways in which it can help us understand basketball.