The Return of Bynum Gets Even

The last few days I have spent either grading or doing something to avoid grading (I hate grading).  While I have been busy (or not busy with this work), a column I wrote on Andrew Bynum and Kobe Bryant has been getting quite a bit of attention. 

Henry Abbott at TrueHoop placed it at the top of Tuesday’s Bullets.

The LA Times Lakers Blog also commented on the column (The (or at least “an”) answer lies within).

Such prominent attention has led the page view count on this one column to pass the 20,000 mark, making this the most read column in the history of The Wages of Wins Journal. 

Frankly, when I wrote this column on Monday night I didn’t think such a story would get an unusual amount of attention.  This is hardly the first time I had challenged the notion that Kobe is the greatest player in the NBA.  For example, see the following:

Kobe Myths

Kobe Bryant is Unhappy

As these previous columns indicate, part of the story I told on Monday night was simply a re-statement of something I said earlier.  Consequently, the Bynum column really didn’t seem much different than any other column I write each night, except I failed to check my spelling (or at least, I am pretty sure that’s not how you spell “similar”). Still, this story did spark quite a reaction, so I thought I would take some time to clarify what I was saying with a sequel.

What follows is “The Return of Bynum Gets Even.” The sequel, which critics must note goes on far too long, is divided into two parts.  The first part re-iterates the story told in the original.  The second part is a reaction to the reaction.

The Story and the Story Behind the Story

Let me start by telling the story behind the story.  

When I started thinking about the column (sometime Monday morning) I was going to compare Bynum and Chris Kaman. In fact, the original title was “The Best Center in LA.”

But then I started thinking about what Kobe said this summer.  And so I took the story in a different direction.

Despite my different focus, though, you can tell I couldn’t abandon the Kaman angle.  In the middle of the column I bring up Kaman, although frankly he has nothing to do with the basic story I was now telling. If we had a competent editor at the WoW Journal (since the editor is me, we clearly don’t), the Kaman stuff would have been cut from the piece.  But since I did the analysis, I felt compelled to throw it in.

Although I left the Kaman material in, the primary story was not a comparison of Bynum and Kaman, but a comparison of Bynum and Kobe.  And here are the basic points I made about the latter comparison.

1. The Lakers kept Bynum this summer over Kobe’s very strong and vocal objection.

2. After 17 games this season, Bynum had rewarded the faith his team showed in his talent.  Per 48 minutes he has been one of the most productive centers in the game.

3. As noted in the column, this analysis was based on 17 games.  This is not much of a sample.  So although I state “…Bynum could argue he has done more (than Kobe) on a per-minute basis”, one could also argue that Bynum only played well for 17 games. In contrast, Kobe has played well for more than a decade.  In sum, we should not conclude – and I don’t think I did — that Bynum is now one of the best players in the game.  All we can say is that across 17 games, he has done pretty well.  And if Bynum keeps producing, Kobe has found the star teammate he demanded this summer.

4. At the end of the column I made an attempt at humor (and frankly, I thought it was a successful attempt).  Bynum has been more productive than Bryant on a per-minute basis (again, across 17 games).  You can see this in WP48, or if you don’t like Wins Produced (can’t imagine that would be true, but it’s possible) in other metrics like NBA Efficiency or plus-minus (the latter was noted by several people posting comments in this forum).  Again, this is only across 17 games.  So we should not make too big a deal about it. But given these early numbers, it’s somewhat funny to think of Bynum “pulling a Kobe” and demanding the Lakers trade Bryant for Jason Kidd.

Okay, those were the four basic points I was making.  Let me also emphasize the big story I was trying to tell.  Kobe demanded the Lakers trade Bynum this summer. The much-maligned Lakers front office said no.  After 17 games, it looked like the front-office knew a bit more about building a winning team than Kobe. 

Reacting to the Reaction

So that was the story I told.  Tom Ziller, at AOL Fanhouse, I think captured much of the sentiment offered against what I said.  Not to pick on Mr. Ziller, but I thought it might be useful to go through his column and offer a few responses.

Let me start by noting that Kelly Dwyer, the latest blogger at Yahoo! sports, linked to Ziller’s column with the following statement:

FanHouse’s Tom Ziller. Another swift and accurate dismissal of Dave Berri’s findings. Always fun.

Given that introduction, here is the Ziller’s column.  I will post this paragraph-by-paragraph, providing responses as we go along. 

TZ: In today’s edition of “Analysts Speeding Toward the Far Reaches of Absurdity,” popular (in corners) statistician David Berri tells us Andrew Bynum is better than Kobe Bryant right now. (Hat tip: TrueHoop.) He states (hopefully sarcastically) Bynum should be the guy upset over the lack of star power around him, and Bynum should be invoking Kobe’s name in vain in front of some dudes with a video camera in a market parking lot. Sigh.

My Response: As noted above, the comment on Byum at the end was meant to be humorous.  We cannot draw strong inferences from 17 games.  Still, a number of other metrics do agree with the notion that Bynum has played better than Bryant.  So even if you don’t like Wins Produced — but liked the other metrics — my evaluation of these 17 games would not be thought of as absurd.

TZ: Berri uses his proprietary, highly controversial metric (and little else) to make his assertion. Berri’s formula says Bynum’s contributes 0.376 wins for every 48 minutes of play, while Kobe offers only 0.268 wins per 48 minutes. (0.100 win/48 is the standard for average — both exceed it by Berri’s numbers.) In other words, Bynum has been quite a bit better than Kobe this year. There are problems with this, among them the inherent difficulties in comparing guards and centers, starters and subs.

My Response: Okay, Ziller is telling us we can’t compare centers and guards.  But Ziller will argue at the end of his column that Kobe is better than Bynum.  So he is comparing a guard to a center. This particular Ziller argument echoes many of the comments posted at the WoW Journal.  Many people posting comments told us that

1. Wins Produced, and/or stats in general, cannot tell us who is better or worse.

2. But I, the person posting this comment, can tell you that Kobe is better than Bynum.

If #1 is true, how did you arrive at #2?  What secret method did you use that told you that Kobe is better? At the very least, people need to tell us how they know Kobe is better than Bynum.

And I would add, there is a difference between saying Bynum out-performed Kobe on a per-minute basis across 17 games, and Kobe is better than Bynum for all time.  Those are not the same arguments.  So in addition to telling us how you know Kobe is better than Bynum, you also need to tell us which issue you are addressing.

TZ: Beyond that lies the crux of unanswered criticisms of Berri: How important is usage? It’s a question many folks much smarter than I have attempted to attack; needless to say, it remains one of the biggest concerns in advanced basketball metrics. Berri grades the importance of usage very lowly; ESPN’s John Hollinger, on the other hand, puts quite a bit stock in the ability to create shots (for oneself and others). And this is the regard in which Berri and the world Hollinger will differ — Berri believes Bynum’s higher efficiency on few shots is better for the Lakers than Kobe’s somewhat lower efficiency on many, many shots. Hollinger thinks otherwise — that there is value in a relatively efficient scorer (and Kobe is that) taking responsibility for so many shots. Like I said, it’s a debate. Few would state one side has achieved clear victory. But I’d venture more stats minds as well as cautious basketball fans would agree with Hollinger. I count myself among this group — I’d be worried if my argument assumed (as Berri’s does) Bynum could take over half Kobe’s shots and remain as efficient. I believe strongly in the efficacy of per-minute numbers — it’s rather evident (to me and others) a good player at 10 minutes per game can/will be a good player at 25 minutes per game. Usage is another beast entirely — I’m not willing to assume a 57%-shooting, 8 FGA/game third banana will shoot 57% when he becomes a 20 FGA/game superstar. There isn’t the evidence there for that… and that assumption seems to be a basis for Berri’s whole usage argument. Until he can (or attempts to) prove it, we should sincerely doubt his outlandish statements which provide little evidence beyond “I said so.” Like this one.

My Response: Okay, there are a few problems with this paragraph.

1. These are not unanswered criticisms.

Contrary to Mr. Ziller’s assertion, we have answered the usage criticism.  Here is a link to column Martin Schmidt posted (in June of 2006) on the link between shot attempts and shooting efficiency.

The Law of Diminishing Returns in the NBA

As the post indicates, Marty failed to find a negative relationship between shot attempts and shooting efficiency.  In other words, the “usage” story wasn’t found in the data.

And here is a link to my comment on John Hollinger’s PERs method.

A Comment on the Player Efficiency Rating

This post details the problems with Hollinger’s approach.  Specifically, Hollinger’s attempt to credit people for creating shots also ends up rewarding inefficient scorers.

2. More on the usage argument 

Let me also note something odd about Ziller’s usage comment.  He notes that the issue of usage has not been settled and people far smarter than him do not know the answer. Yet his argument presumes he does know the answer.  Kobe is better than Bynum because if Bynum took 20 shots per game he would shoot so poorly that his productivity would fall below Kobe.

Certainly this is possible.  Currently Bynum takes 7.5 shots per game and shoots 58%.  If he shot at the same pace as Kobe, and only hit 48.5% of his shots, Bynum’s per-minute performance would equal what we see from Kobe.  Now would Bynum’s shooting efficiency fall this far if he took more shots? I don’t know and I suspect Ziller doesn’t know either.

We could also look at this from Kobe’s perspective.  Would Bryant become a much more efficient scorer if he shot as often as Bynum? Currently Bryant shoots 49% from two point range and 37% from beyond the arc. If we lowered Bryant’s shot attempts per 48 minutes to the level we see from Bynum, Bryant would have to shoot 60% from two-point range and 45% from beyond the arc to match Bynum’s current per 48 minutes wins production.  This translates into an adjusted field goal percentage of 61.9%. 

Could Bryant make such a leap in shooting efficiency? Of the 3,816 players to play at least 1,000 minutes in a season since 1991-92, only 10 had an adjusted field goal percentage this high.  Ziller is telling us that if Bryant shot as often at Bynum, he would join this very small group.  This is certainly possible, but I think rather doubtful.

3. Arguing about hypothetical players

I would add, and perhaps this is the more important issue, that I have problems with this entire line of reasoning.   Ziller wants us to compare either a hypothetical Bynum (one who shoots more often) to Byant, or Bynum to a hypothetical Bryant (one who shoots less often).  Isn’t it easier and more relevant to just compare the actual Bynum to the actual Bryant?  And shouldn’t that comparison focus on more than just scoring?

4. How productive a player is? vs. Why the player is productive?

And here is one last issue.  The entire discussion of usage confuses the question of how a productive a player has been with why that player has been productive. This is a point I made in the following post:

What Wins Produced Says and What It Does Not Say

Even if usage was as big of an issue as Ziller claims, it would only tell us something about why a player is productive (or not).  It wouldn’t change how productive the player has been.  In the past I have argued that these questions should be examined sequentially.  First, let’s measure how productive the player has been.  Then let’s consider the impact of experience, coaching, teammates, and if you like, usage (although we already did look at usage and didn’t find much evidence supporting this story).

TZ: Furthermore, this is why many fans hate advanced metrics — ‘Nerdball’ — so much. Bynum versus Bryant (or Bynum versus Jermaine O’Neal versus Jason Kidd versus Alex Rodriguez versus Adriana Lima, for that matter) is a discussion for a sports bar, because there are myriad ethereal factors to weigh in such a discussion. Using statistics in sports to come up with a Holy Grail, two digit, indisputable ranking scheme is almost evil. I’d rather we use statistics to describe players and the game instead of using statistics to bastardize/militarize our viewing of the proceedings. Unzipping and whipping out Excel spreadsheets is not fun.

My Response: Again, Ziller tells us that we cannot know who is best.  But then again, he is telling us who is best.  And as I noted a few days ago, there are no Holy Grail or “magical” formulas. Here were my specific words:

“I sense, though, that people become frustrated with these metrics because they expect “magic.” In other words, people want a number that answers all questions and reduce the cost of thinking to zero. Models, though, help us explain the world we observe. Models are not “magical”, nor do they remove the need to keep thinking. And that is something to think about when you look at basketball measures, or any other models researchers offer to improve our understanding of our world.”

TZ: (To wit: I hated every second of writing this; Bynum — this year and last — has been much better than most give him credit for. He’s almost clearly the second best player on the Lakers, right? However, in my mind, there’s no way he’s more valuable to the Lakers this year than Kobe. And Berri left no room for question in his statement; to counter, you can likewise leave little air. Discussions like this SUCK.)

My Response: And then at the end of the column, we see Ziller actually thinks Bynum is very good. He just “knows” that Bryant is better.  And not just better across all time, but clearly better “this year.”  

Again, we only had 17 games when I wrote the original column on Monday night. With such a small sample we can’t say that Bynum (or Kobe) is better across all time.  So I did leave plenty of wiggle room on this question.  It appears that Ziller is the one lacking wiggle room.  We can see that he is sure Kobe is better (across this season and for all time).  We just don’t know how he reached that conclusion.

And let me repeat, I am not trying to pick on Ziller (or Dwyer).  This column, though, did re-iterate many of the observations made in the more than 100 comments posted on the Bynum-Bryant column.  Rather than try and respond to each of these (which many of the WoW regulars tried to do over the last couple of days) I thought it would be more efficient to just note some the inconsistencies in the Ziller story.

Oh, and one last point.  The part about Bynum demanding Kobe be traded was meant to be humorous.  Maybe I should put up a sign when that happens.  At times it does seem people who write about sports — which are supposed to be fun – don’t have much of a sense of humor.   

- DJ

And for those who want more information, I encourage you to read the following:

Our research on the NBA was summarized HERE.

The Technical Notes at wagesofwins.com provides substantially more information on the published research behind Wins Produced and Win Score

Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:

Simple Models of Player Performance

Wins Produced vs. Win Score

What Wins Produced Says and What It Does Not Say

Introducing PAWSmin — and a Defense of Box Score Statistics

Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.