2011 WNBA Regularized Adjusted Plus-Minus: Who Were The Top Offensive & Defensive Performers?

Yesterday I took a look at plus/minus ratings with a focus on four-year Regularized Adjusted Plus Minus as a metric that measures player impact (click here to revisit that).

Today, I continue looking at RAPM, but narrow the scope to a one-year perspective to get a sense of how RAPM matches conventional wisdom about the top performers of the 2011 WNBA season. Before jumping in to these, it's worth noting Dan Rosenbaum's point from 2005: he suggested then that adjust plus/minus required half a NBA season's worth of games (41 games at 48 minutes each) to produce stable results; a WNBA season does not meet that standard at 34 40-minute games. That's why I began with the four-year ratings yesterday.

Nevertheless, Jeremias Engelmann put together the one-year results as well (click here) and they offer some interesting food for thought as we think about how our favorite teams can improve during the long WNBA off-season.

Just as we did yesterday, it sometimes helps to compare a new metric to existing metrics to help establish a common frame of reference for how exactly it's challenging the status quo. So in addition to the top RAPM ratings, I'm providing three other metrics that we might be more familiar with:

• Net plus/minus: This is an obvious one as those "raw" numbers ultimately underpin RAPM. The differences between the one-year data RAPM data and NPM are very interesting although it's worth noting that they're not the same because they're not supposed to be - that's the purpose of the adjustment. Net plus/minus, its strengths and limitations were described in yesterday's post. But for our purposes here, we'll be comparing only the on court data from NPM to the RAPM data for offense and defense and comparing that to league average. So the rule of thumb is that if, for example, there's a player whose offensive "on court" rating is above league average and in above average minutes will probably have a good offensive RAPM. It's similar for defense, with the normal contextual caveats of course. For your reference, the league average for both team and opponent on court data is just under 77 points per 40 minutes, which makes sense given that the average points per game scored in 2011 was 77.3.
  
• PER: This is John Hollinger's Player Efficiency Rating and James Bowman described that previously here. It arguably correlates best - but still not well - with RAPM.
• MEV: This is the David Sparks Model Estimated Value metric*, which I like because of how it fits into a variety of team and individual stats in the Sparks constellation of metrics. It's not widely used ... at all ... but I like what it enables for analysis. Click here for more on that metric.

So with that, here are the top 15 players in 2011 by RAPM are as follows, along with other metrics used in the WNBA:

Top 15 players in the 2011 WNBA regular season by RAPM in comparison to net plus/minus, MEV & PER.

What might immediately stand out here looking at this as a whole is that 8 of this top 15 are also among the top 15 in net plus/minus.

Top defensive players by RAPM in 2011

Looking at those 7 players who "fell off" after the adjustment, so to speak, one thing that stands out is that all of them might have benefited from "the backup effect" - most were high usage players that were significantly better than the player that replaced them. But before we dig too deeply into that matter, let's look at the top players by defense and offense.

Top 15 defensive players by RAPM for the 2011 WNBA season.

So 11 of these players and their defensive abilities were discussed by Richard Cohen in our winding exchange about All-Defensive teams previously (click here to revisit that). As an example of the on court rating rule of thumb, Camille Little has a good defensive RAPM and opponents scored only 67.3 points/40 when she was on the court, which is well below league average in a lot of minutes.

By that on court raing logic, Erin Phillips really jumps out on this list. But rather than go over the players we "left out", I'd actually rather return to the discussion of Sophia Young, a player who has solid defensive statistics and made the top 15 RAPM list overall almost entirely due to her defensive ability.

What Young does well defensively is get steals, with a 3.25% steal percentage - her athleticism certainly makes her disruptive and she has great awareness, instincts, and sense of space defensively. There's no doubt that she's a smart defender. But what still makes her inclusion on a list like this interesting is that she's also an undersized defender, which allows bigger players to both rebound and score over her around the basket.

Again, as Clay Kallam already commented on my All-Defensive team post, defense is "not me vs. you, it’s us vs. them." So while any quantitative defensive rating is going to be incomplete, a lot of these are justifiable by observation; a few might require a bit of convincing. 

In any case, not one of the 7 top net plus/minus players who "fell off" with adjustment and regularizing were among the top 30 defensive players by RAPM, with Becky Hammon being the highest rated (#32, 0.6).

But interestingly, for all the praise McCoughtry gets for her defense, it was her offense this season that carried her to the top of the RAPM ratings.

Top offensive players by RAPM

Defensively, McCoughtry is #22 by RAPM. Offensively, she tops the list.

Top 15 offensive players in the 2011 WNBA season by RAPM.

First, it's just sort of interesting to note that this list apparently had an unspoken "Conference Finals only" rule attached to it, with only Plenette Pierson crashing the party. That's not entirely arbitrary - the majority of this list of top offensive players played for four of the top six offensive teams in the league by points per possession. With the exception of the Catchings, they were also all on teams that played at a fast pace.

Continuing the discussion about Seimone Augustus from yesterday, she might not have a great net plus/minus (-0.1) but her team was above on offense when she was playing and she obviously played a lot of minutes.

But what remains interesting is that there were All-Star players among the top 15 in net plus/minus who weren't in the top 30 by RAPM offensively or defensively: Becky Hammon (#39 offensively), Danielle Adams (#40 offensively), Crystal Langhorne (#48 offensively), Sue Bird (#68 offensively). Trying to figure out what that set of otherwise very good players have in common - and yet don't have in common with some of the others - is difficult.

Yet digging through some stats revealed something potentially interesting about this subset of players that ranked high in net plus/minus but lower than one might expect in RAPM offensive ratings.

All but two players in the top 30 RAPM offensive ratings were in the 63rd percentile or above in something called "floor percentage", which is described here as "the fraction of a team’s or individual possessions on which there is a scoring possession" or here as "the probability that his team scores at least 1 point" when a player uses a possession. When you think about what RAPM is trying to do - measure a player's impact on the score - it makes a lot of sense that floor percentage would matter to the offensive portion of the rating.

But what separates Harding, who is at exactly that 63rd percentile in floor percentage, from Bird, who's in the 60th percentile? Usage rate. And in fact, the effect is even larger for post players - Tina Charles and Sylvia Fowles have two of the highest floor percentages in the league and Langhorne isn't far behind at 6th. All three of them are relatively high usage (24% +) post players; low usage post players not in the top 30 (or so) have two-point percentages under 50% and high turnover ratios almost as a rule. In other words, RAPM offensive ratings place a high value on players that maximize touches by producing points for their team without the expense of wasting possessions in the process.

Conceptually, that makes a lot of sense, but begins to break down outside that top 30 and there's a lot to be said about that - and floor percentage in particular* - as it relates to what players contribute and whether we find RAPM useful.

Notes:

* In putting MEV side-by-side with PER, it's worth noting that the league-high in 2011 was 20.20 (Sylvia Fowles) and the league average was 6.88 (somewhere between Marissa Coleman and Kalana Greene), with a large "middle class" and small elite. So not only does the discrepancy between those numbers not mean much, but every player listed above is "above average" by MEV; no so for PER.

** Five of the top 30 floor percentages in the league belong to the Lynx starters (the same five you see above). That's more than any team in 2011. The Lynx were also the highest synergy team in the league. To me, that's not a coincidence - again, it's not just that they were talented, but they had a mix of players that worked extremely well together to get shots.