Statistical +/- for Rockets

2010 February 1
by durvasa

Using the statistical +/- model described at the excellent blog here and here, and data provided by, I calculated boxscore ratings for all NBA players who have played at least 500 minutes this season. Statistical +/- is not a +/- rating in the conventional sense, in that is not based directly on how many points the team scores over the opponent with the player on the floor. Rather, it uses indicators in the traditional boxscore stats (such as minutes/g, points/min, reb/min, etc.) to estimate a player’s +/- impact per 100 possessions. The particular formula described in the links above was arrived at using 7 years worth of NBA data. It is not a fool-proof way of rating a player’s impact, and I don’t think such a system exists. But it serves as an alternative to PER, or NBA EFF, or whatever other method you like to use to assess a player’s worth based on his individual stats. Like any statistical basketball rating system, it should be treated as a supplement to actually watching the players perform.

Here is a list of the top players so far this season based on this statistical +/- model:

Player              Team     PS     GP     mpg      SPM
james,lebron        cle      SF     48     38.6     13.6
wade,dwyane         mia      SG     47     36.4     9.7
paul,chris          nor      PG     38     38.8     9.2
duncan,tim          san      C      42     32.5     7.8
howard,dwight       orl      C      47     35.1     6.8
ginobili,manu       san      SG     40     26.7     6.7
bryant,kobe         lal      SG     47     38.4     6.7
smith,josh          atl      PF     46     33.7     6.5
durant,kevin        okl      SF     46     39.9     6.2
roy,brandon         por      SG     40     38.2     5.8
love,kevin          min      PF     28     30.2     5.6
billups,chauncey    den      PG     38     32.7     5.6
anthony,carmelo     den      SF     38     37.9     5.5
kidd,jason          dal      PG     46     36.1     5.5
camby,marcus        lac      C      45     31.1     5.5
gasol,marc          mem      C      46     35.6     5.2
wallace,gerald      cha      SF     45     42.1     5
oden,greg           por      C      21     23.9     4.8
gasol,pau           lal      PF     30     36.4     4.8
davis,baron         lac      PG     46     34.2     4.7
rondo,rajon         bos      PG     43     36.9     4.7
granger,danny       ind      SF     30     36.3     4.6
nowitzki,dirk       dal      PF     46     38.1     4.6
bosh,chris          tor      PF     47     35.9     4.6
hilario,nene        den      C      46     34.3     4.2
arenas,gilbert      was      PG     32     36.5     4.2
pierce,paul         bos      SF     39     36       4
wallace,ben         det      C      45     30       3.9
frye,channing       pho      C      48     29.8     3.7
williams,deron      uta      PG     41     37.1     3.7
wallace,rasheed     bos      C      41     24       3.6
murphy,troy         ind      PF     37     31.1     3.6
kirilenko,andrei    uta      SF     40     29       3.6
boozer,carlos       uta      PF     45     35       3.6
lowry,kyle          hou      PG     46     25       3.5
iguodala,andre      phi      SG     46     40.1     3.5
martin,kevin        sac      SG     14     36.2     3.4
odom,lamar          lal      PF     47     31.3     3.3
harden,james        okl      SG     46     22.9     3.3
noah,joakim         chi      C      44     32.9     3
anderson,ryan       orl      PF     39     16.3     2.8
fernandez,rudy      por      SG     30     22.9     2.7

Here is the list for the Rockets:

Player              Team     PS     GP     mpg      SPM
lowry,kyle          hou      PG     46     25.0     3.5
battier,shane       hou      SF     46     33.1     0.8
ariza,trevor        hou      SF     45     38.2     0.6
hayes,chuck         hou      C      46     21.9     0.5
landry,carl         hou      PF     45     27.2    -0.4
brooks,aaron        hou      PG     46     35.4    -0.5
scola,luis          hou      PF     46     30.8    -0.8
budinger,chase      hou      SF     38     18.6    -1.9
andersen,david      hou      PF     46     14.2    -2.3

There are some surprising results for the Rockets players. Lowry leads the team by far. The two rookies (Budinger, Andersen) are at the bottom, which actually shouldn’t be a surprise. Everyone else is between -1 and 1.  While having a rating less than 0 is technically “below average”, keep in mind that “average” in this case is much higher than the statistical median for all the NBA players. That is, there are far more below average players in the league than above average players. An average player corresponds to a player who is basically starter-quality. If you are familiar with John Hollinger’s PER system, this would be a player with a 15 PER.  Compare these results, based on boxscore indicators, to the results Wayne Winston had for the Rockets as of his January 20th blog on the team.  It may not be obvious from the lineup data he shows in that link, but it is possible from the information he provided to arrive at his adusted +/- ratings for each of our 9 main rotation players — both at the offensive end and defensive end. I show this graphically below:

This depicts offensive and defensive adjusted +/- for 9 Rocket players.

Left to right, players go from bad to good offensively. North to south, players go from bad to good defensively (the more negative, the better).  Winston’s results differ substantially from what the statistical +/- would predict for some of the players, in particular rookie David Andersen.  I tend to trust the results from the statistical +/- formula more in this case. Winston has Andersen as rating well above average both on offense and defense. But Andersen, in my humble opinion, is the worst defensive player the Rockets have had since Steve Novak; I can only assume that Winston’s method is allocating defensive credit to Andersen at the expense of other Rocket players (especially Landry and Budinger, who rate as awful on defense according to Winston). Winston’s results on the offensive side make a little more sense, subjectively.

Returning to the statistical +/- results, it is a surprise to see both Battier and Ariza rate so highly. The SPM model likes players who take a lot of threes (which they do), probably because such players tend to spread the floor, leading to more efficient team offense. I had thought that Ariza’s poor scoring efficiency would sink him for this rating system, but it turned out that he does enough other things well according to the model (particularly his high 3-point attempts, his steals, and his versatility index) to end up rating above average. Remember before when I said this is not a fool-proof system? I believe Ariza is a case where his offensive ineptitude is not being well-captured by this model; this is a case where Winston’s assessment rings more true.

To close, below I show a histogram of SPM ratings for all NBA players who have played at least 500 minutes:

SPM Distribution for all NBA players (Rocket players highlighted)

Again, 6 Rockets fall between -1 and 1 in the histogram above. The two rookies have less than a -2, which may after all be appropriate (though I think Budinger has shown the potential to become a starter-quality player some time soon).  According to statistical +/-, Kyle Lowry has separated himself from the pack. He does not score a lot of points, and he does not shoot (or make) a lot of threes. But, by the box score, he does everything else really well. He gets to the line, he gets a ton of offensive rebounds (leads the league amongst PGs), he gets a good number of steals, and he has a high versatility index. And this doesn’t even account for the fact that Lowry leads the league in drawing offensive fouls. Has he been the Rockets best player this season? I am not sure I would say that, but I do think he’s been the most important Rocket in regaining lost momentum from the starters’ frequent lulls to start each half. And I think many people, especially the national media at large, have under-appreciated his impact.

Bookmark and Share
  • Easy
    duvasa, I don't have time to decipher the formula. But since this method is using traditional boxscore numbers, I would venture to guess that it would not be as accurate on the defensive impact as it is on the offensive end, since the boxscore numbers are notorious in lacking in measuring defensive performance.

    Am I understanding it right?
  • durvasa
    I think that's likely true. It formula was arrived at to estimate overall impact in terms of adjusted +/- (which is equal parts offense and defense). But I think its likely biased towards the offensive component simply because that's easier to predict using the boxscore. I don't think its as offensively-biased as some other boxscore stats like PER, but that is more of an intuitive guess on my part.
  • physicsgeekandrocketsfan
    Alituro, I agree, Lowry is something else. You gotta give credit to Morey for seeing that when he was on the Grizzlies, I mean he wasn't even getting that much playing time back then. Tsunami, you're also right about Lebron, this shows that adv. stats can still capture unbelievable star power like Lebron. Which brings me to my question, who is the poor bastard with -7.5?
  • durvasa
    That would be Sasha Pavlovic of MIN, who with a SPM of -7.01 fell in the (-7.5, -7.0) bracket in the histogram.
  • Tsunami
    the distribution with LeBron by himself way out in space is awesome.
  • Alituro
    Not only a great player statistically, as proven above, but much like Battier, there is so much Kyle does that can't be quantified. Last week he was on a 2-on-1 fast break, I think with Chase trailing and a bigger defender in front on the inside. He used his body to screen the defender all the way to the basket, keeping his body between the trailer and defender, facilitating an uncontested layup. It was chalked up as an assist for Kyle, but one that needed a serious asterisk and footnote. I can't count the amount of times I've seen him set screens on the break at the perimeter, again, assists with asterisks. Seen him drive through the lane and spot a defender with his back to the ball and ram it straight at him, knowing that there will never be a charge called if the guy's back is to the ball, and-1 opp on the way. Cross-court passing accuracy? Again, assists with asterisks. One of the top rebounding points in the game, even with bench minutes. It's like he has a basketball encylopedia in his head that he's able to refer to in a split second as a situation arises. If the Rockets are able to pick up another consistent scorer besides Brooks, I'd like to see Kyle start and have Brooks off the bench for some instant scoring when needed. If Conroy can be 3/4 as effective as Kyle, then I'm all for selling high on Brooks to bring in that Scorer.
blog comments powered by Disqus