Friday, January 30, 2009

Sofamore Statistics are Stoopid

Today I'd like you to join me in a brief discussion of the Rookie/Sophomore roster selection for the 2009 All-Star Saturday festivities.  The rosters for the game are as follows:

Player (Team)PositionHeightWeightSchool/Country
Michael Beasley (Heat)F6-9245Kansas State
Rudy Fernandez (Trail Blazers)G-F6-6185Spain
Marc Gasol (Grizzlies)C7-1265Spain
Eric Gordon (Clippers)G6-3222Indiana
Brook Lopez (Nets)C7-0260Stanford
O.J. Mayo (Grizzlies)G6-4210USC
Greg Oden (Trail Blazers)C7-0285Ohio State
Derrick Rose (Bulls)G6-3190Memphis
Russell Westbrook (Thunder)G6-3187UCLA
Head Coach -- TBD
Assistant Coach -- Dwyane Wade
Player (Team)PositionHeightWeightSchool/Country
Aaron Brooks (Rockets)G6-0161Oregon
Wilson Chandler (Knicks)F6-8220DePaul
Kevin Durant (Thunder)G-F6-9215Texas
Jeff Green (Thunder)F6-9235Georgetown
Al Horford (Hawks)F-C6-10245Florida
Luis Scola (Rockets)F-C6-9245Argentina
Al Thornton (Clippers)F6-8220Florida State
Rodney Stuckey (Pistons)G6-5205Eastern Washington
Thaddeus Young (Sixers)F6-8220Georgia Tech
Head Coach -- TBD
Assistant Coach -- Dwight Howard

Today, John Hollinger of ESPN, who specializes in the study of "accurate measurement" statistics (think football's DVOA) and who coined the PER ("player efficiency rating" - a rating which Hollinger believes can accurately rank any player at any position) took the rosters of this year's rookie/soph game to task because he believed they ignored the truly good young players (that is, those with high PERs or other statistics he deems accurate assessors of talent).  To put it plainly, John was upset that the NBA had not been using his statistical methods to assess the play of young basketball players. 

The rosters for the R/S game are chosen by a group of assistant coaches (I'm not sure if it's all of them or a committee) who watch the young players during the season and presumably review their statistics when they make a selection.  If a rookie is in a coach's conference, they will have seen them play twice before the All-Star game; if not, once.  The coaches are the beneficiaries of reports on the prospective rooks/sophs from their scouting staffs during the regular season (particularly important with rookies) and they have, of course, considerable experience and expertise in basketball.  

Mr. Hollinger, on the other hand, has a background entirely based in online sports journalism.  He is without graduate degrees, scouting experience, and as far as I know never played above high school (if he even played there).  ESPN has somehow managed to prevent lechrous NBA teams from hiring Hollinger to give them an edge over their competition with his impressive statistical analysis - he has no professional experience other than his ESPN work.   

Although it may seem strange in light of the foregoing, Mr. Hollinger, who got his start with the web site humbly described as "the basketball page for thinking fans", believes that he knows more about who should and who should not be on the rookie team than the NBA's assistant coaches.  I believe that he is incorrect and seek to discredit him.  His original article, from which I am about to refer, is available here.  

The first selection Hollinger takes issue with is the selection of Eric Gordon over Kevin Love.  He states that Gordon was picked because of his gaudy per-game statistics (13.8, 2.5, and 2.5) and, to quote him:  "Don't get me wrong; Gordon is going to have a fine career, it seems, and in almost any other year he'd be a shoo-in for the team. But he made this squad mainly because the forlorn Clippers have no choice but to play him extensive minutes."

Despite good play by the T-Wolves in January, I don't think Hollinger would describe either Love's or Gordon's teams as "good".  It's true that Gordon gets to start because Cuttino Mobley went down and that Love has the shadow of Al Jefferson to deal with, but there is one plain fact that Hollinger is missing.  Kevin Love is not getting time on a bad team because of the bad team's decision not to play him.  Kevin McHale, no less than one of the top five post players in history, is starting Craig Smith over Mr. Love.  

Hollinger explains what McHale and the NBA assistant coaches are missing:  Love leads the league in offensive rebound rate, as I mentioned the other day, but his prodigious work on the boards has gone largely unnoticed because he plays only 23.2 minutes a game, far less than Gordon's 32.2.  (Note:  Hollinger is incorrect about Love's rebound rate - he only leads rookies - he is third in the NBA.)

Do you see how this works?  Although Kevin Love's own coach, who watches him practice, spends from 8-10 hours a day with him, and knows more than we could ever hope to about his personality, temperment, and ability, does not play him more than Craig Smith, and he is wrong not to, because Kevin Love has amazing ORP48 (Offensive Rebounds Per 48 Minutes) statistics.  But, just for the sake of argument, let's take a look at how the ORP48 statistics relate to player skill.  Below is a chart of the Top 15 ORP48ers*:


*Note that the list display starts at five and includes eighteen entries because the top five all played five games or less and there were three people on the list who should not be applicable due to too few games played who I couldn't figure out how to remove.

That chart is a who's who of marginal big men in the NBA.  Are some of them (Dwight Howard, Marcus Camby) excellent rebounders?  Absolutely.  Are some of them horrendous guys who can hustle their asses off with no regard for fouling because they don't have to play that many minutes?  Yes!  (See:  Joel Pryzbilla, whose stats are almost identical to Love, Aaron Gray, Reggie Evans, Marcin Gortat).  In my mind, Love can fall into the latter category as easily as he can the first, and Hollinger gives no other statistic to support the assertion that Love is more deserving than Gordon.  Even if he is correct that Love is the third-best offensive rebounder in the league, does this fact prove he has more skill than someone who is, say, an excellent scorer?  Or passer?  

The story with these players' rookies season thusfar illustrates their skill level and, in my mind, shows how they should be rated respective to each other.  Love has been playing consistently all year, and has shown some improvement.  In November, he averaged 8.4 points and 6.1 rebounds per game.  In December, he averaged 6.6 ppg and 9.1 rpg.  In January, he averaged 12.5 ppg and 10.1 rpg.  These numbers are, obviously, mediocre.  

Gordon's numbers tell a similar story.  It's just a better one.  In November, he averaged 7.8 ppg, 1.4 rpg, and 1.7 apg.  In December, it was 13.9 ppg, 3.7 rpg, and 1.8 apg.  This January, he has averaged 21.4 ppg, 2.4 rpg, and 4.3 apg.  

If anyone wants to argue that a player who is their team's top scorer every night should be cut from a team because a guy who is good at offensive rebounding in limited minutes and who is warming the pine behind Craig Smith is statistically better than the scorer, I think it's safe to say that person is not qualified to question the assistant coaches.  

Let's continue with Hollinger's "snubs".  Hollinger notes:  Because of that [coaches looking at only simple statistics and ignoring minutes played], we'll have the Human Goaltend Violation, Al Thornton of the Clippers, playing in the game instead of Houston's Carl Landry, and we'll have New York's Wilson Chandler instead of Toronto's Jamario Moon.

(What is a human goaltend violation?  Does Al Thornton goaltend often?  I tried to find statistics reflecting the same and I don't think the NBA keeps them.)  

Some of my readers may recall that I am a big Carl Landry fan and certainly have no axe to grind.  I also like Jamario Moon, one of the best dunkers in the NBA.  

However, in this analysis, Hollinger is simply ignoring the reality of Landry and Moon's situations.  Carl Landry, love him though I do, is judged by Rick Adelman, one of the better coaches in the NBA, to be worse than Luis Scola (12.3/7.9).  I have nothing against Luis Scola, but he's not that good.  There's a simple commutative rule in sports: if you're getting benched behind a guy who's not that're not that good either.  Coaches with thirty years of experience rating players don't usually make mistakes on their own teams (though it's not unheard of) and I don't think anyone would argue that Rick Adelman benching Landry is an exception to that rule.  

So why does Landry deserve to make the team, you ask?  Landry, in particular, is a stunning exclusion. I've written many times already about his impressive per-minute numbers the past two seasons, but suffice to say that he ranks right between Josh Smith and David West in PER, and that his numbers are actually down from what he did as a rookie.

See...he's right in between Josh Smith and David West.  If only Rick Adelman knew. 

I went on ESPN and looked at the season PER ratings.  David West is rated 49th in the NBA (20 spots behind Marreese Speights...who was not even chosen for the rookie team!), and I can't find Carl Landry.  ESPN would have me pay a fee to view beyond fifty, and he's presumably there, ahead of Al Thornton.  But in my mind, Thonrton (who averages about 17 and 5) being behind Landry (who has scored twenty points twice this season (20 and 21)) in PER is more of a condemnation of PER than an accurate assessment of skill.

The Chandler/Moon comparison is also quite damning.  Jamario moon is a 28 year old sophomore whose statistics dropped from last season (8.5 ppg) to this one (7.5 ppg).  He is athletic, a good defender...and he scores 7.5 points per game despite starting all but seven Raptors games this year.  That makes him, as far as I know, the second least-effective starter in the NBA (after Ben Wallace).  Fine me a worse one and I'll give you credit.  

Wilson Chandler, on the other hand, is 22 years old, and went from scoring 7.3 ppg, 3.6 rpg, and .5 apg last year to 13.8 ppg, 5.0 rpg, 2.1 apg, and also throws in a block and a steal per game.  

I know there are factors (D'Antoni) that might skew Chandler's statistics but I don't see how one can argue that someone who actually got worse since last year, is old, and is the second-worst starter in the NBA should replace a young, up-and-coming player who happens to shoot for a low percentage.  

Hollinger's last point is that Aaron Brooks should be replaced by Ramon Sessions.  I have no problem with that, but would like to take this opportunity to cite Hollinger's draft analysis of 2007 which stated, based on the same statistical analysis that now vindicates Sessions, that: 

Nevada guard Ramon Sessions, and Florida guard Taurean Green are fringe first-rounders who are best to be avoided.

Hollinger for his draft analysis had rated Sessions behind Josh McRoberts ("he has the assist to turnover ratio of a point guard"),  Mike Conley ("in the last six drafts, only Chris Paul ranked higher than Mike Conley"), Nick Fazekas ("in a dead heat for No. 4...he's like Nick Collison with a jump shot"), and many others. 

The point I'm trying to make here is this:  John Hollinger and his ignorant yet arrogant analyses are full of shit.

As always, e-mail me at


Prolow said...

Hollinger sucks. A lot. Good post

John said...

Hi James...I am very impressed with the quality and depth of your writing!