Why do advanced statistics hate Kobe Bryant? Part 2

Yesterday, I posed a question that couldn't be answered in one post:  Why is there such a large disconnect between Kobe Bryant's place in the game analytically (A top 2-3 player, with at least a fair claim on being the best player in the league) and his place statistically (an elite player, but not even one of the top 5 players in the league)?  That question apparently resonated with quite a few of you, and it spawned quite a bit of great conversation.  Allow me a quick recap, so as to better set the scene for today's continuation.

First, I posed the question itself: Why do advanced statistics consistently seem to under-value Kobe Bryant, compared to how "experts" (in this case GM's) value him?  Because a conversation about all advanced statistics would be more of a book than a blog post, I limited the discussion to the most well known advanced statistic, PER.  I described, in depth, exactly what PER is and what it hopes to accomplish.  Finally, I presented Kobe's PER, broken down by individual component, and provided the simplest of analysis regarding those components, with a promise to go more in depth by comparing Kobe's individual components to other players.

Before I follow through on that promise, I want to clarify the intentions of the post, based on some of the comments that yesterday's entry received.  Quite a few people were quick to either dismiss or defend the role of advanced statistics.  It was pointed out that, while statistics themselves are cold, hard numbers, their interpretation is not set in stone.  Defenders pointed to the fact that these statistics are simply meant as tools to develop a deeper understanding of the game.  Many people asked for stats that quantify intangible qualities that we all know are unquantifiable.  Even more people were ready to talk about PER's limitations (and believe me when I say I will get into those myself later).  But very few latched on to the specific question I'm trying to address.

If you were to ask an NBA GM who the top 5 players in the league are, I'm guessing you would get a consensus that would look something like this (more or less in order): 

LeBron James, Kobe Bryant, Dwyane Wade, Dwight Howard, Chris Paul. 

Of the 5 players on that list, 4 of them were in the top 4 in PER.  Only Kobe is not approximately where he should be in the PER rankings.  So, one could say that PER and NBA GMs are pretty close to complete agreement when it comes to player evaluation (at least at the top of the league).  The one exception is Kobe.  In Kobe, GMs have a much higher appraisal of his value than PER does.  We are left with one of three conclusions:  1) GMs over value Kobe Bryant (or more simply, the GMs are wrong), 2) Something about Kobe Bryant's game doesn't translate statistically (PER is wrong), or 3) My imaginary "GM rankings" are full of shit and GMs don't really value Kobe that high (I'm wrong).

Option one is fairly pretentious.  Assuming that you know better than the people who are paid a great deal of money to know this stuff is quite arrogant, even if you have the "numbers" to prove it.  Option three could be true, but I feel like my imaginary list is an accurate portrayal of things, based on the answers to various questions from the aforementioned GM survey.  So I'm left to believe, and attempt to prove, option 2, that something specifically related to Kobe Bryant's game fails to translate statistically. So, now that my intentions are laid bare, let's get into it and start comparing Kobe to some other, pretty decent, players.

As has been pointed out, using PER to compare players across positions can be a bit tricky.  No matter if it is PER's purpose to give us a way to compare Chris Paul to Tim Duncan, the fact remains that PER is an arbitrary value, consisting of a collection of arbitrary values, and so any comparison of CP3 to TD is based on the assumption that the arbitrary values involved are all correctly assigned.  In layman's terms, we have to assume that Hollinger's formula correctly assigns the proper value to completely different stats, like assists and rebounds, in order to spit out a number that properly compares the normal production of a point guard to the normal production of a center.  In order to avoid that assumption (which, in my opinion, is iffy at best), we're going to stick to comparing apples to apples.  Which is to say, we will be comparing Kobe Bryant's PER values to other players who a) play the same position as Kobe and b) play roughly the same role for their teams as Kobe.  If you were looking forward to the PER comparison of Kobe to LeBron, I'm sorry to disappoint.

The two players we will use each serve a separate purpose.  The first is Dwyane Wade, who is an example of a player that should (based once again on my imaginary list) be considered roughly equivalent to Kobe, but who's PER is significantly higher.  The second is Brandon Roy, who is an example of a player who should be considered well beneath Kobe, but who's PER is strikingly similar to Kobe's.

The first table shows a comparison of Wade's and Bryant's separate PER components.  The 2nd table shows a comparison of their actual stats, per 36 minutes.  All numbers are courtesy of Basketball-Reference.com.  So, what do the numbers show us? 

• Wade picks up two PER points in assists, which correlates to roughly the same amount of additional assists in the real stats. 
  
• Wade picks up a full two PER points from FG, even though he only makes .4 more FG than Kobe per 36. 
  
• Wade loses about a point to Kobe due to increased turnovers. 
  
• Wade gains a full point from free throws.  This difference would be a lot higher (closer to two full points) if Wade converted free throws at the same high rate that Kobe does.
• The two are roughly equivalent in rebound PER points, despite the fact that Kobe pulls down a half rebound more per 36 than Wade
• Wade gains a full two points from "defensive" categories, blocks and steals.
  

I'm going to save my opinions regarding these results for the end.  For now, I'm simply pointing out the differences.

Kobe Bryant vs. Brandon Roy

Next, we have Kobe Bryant and Brandon Roy.

• Kobe picks up a full PER point in total PER FG points (3PT + FG - FGA)
• Kobe picks up 1/2 point from free throws.
• Kobe loses 1/2 point in assists
• Kobe loses 3/4 point due to increased turnovers
• Kobe slightly loses PER points in rebounding, despite pulling in .5 additional rebounds per 36 minutes
• Kobe picks up 1/2 point in defensive statistics (blocks, steals), but gives up most of that margin due to extra fouls

So what does it all mean?

I wish I could tell you that this is the part of the article where I hit you with some revolutionary way of thinking that will change the world surrounding PER, but the numbers don't support such an argument.  All I can do is repeat truisms that have long been known, except that I can actually provide the numbers to back up why previously mentioned limitations regarding PER are actually true.  That said, it should come as no surprise that there are two main areas that I feel PER mis-represents Kobe Bryant.  Assists and Defense.

We'll start with defense, because it is a limitation that is universally known for PER.  Hollinger himself admits that PER does not accurately reflect the contributions of a strong defender, especially a defender who's defense doesn't show in the stat sheet (i.e. with steals and blocks).  Hollinger uses examples like Bruce Bowen and Quinton Ross as defensive specialists who have quite poor PER, despite an obviously valuable contribution on the defensive end of the court.  But what of Kobe Bryant, one of the league's premier perimeter defenders.  Isn't PER similarly limited in quantifying his defensive contribution?

Let's take a step back for a second and talk about Kobe's defense.  There's no question, in anyone's mind, that when Kobe puts his mind to it, he is one of the best defenders in the NBA.  The problem with assessing Kobe as a defender is that he doesn't put his mind to it all the time.  Therefore, statements actually proclaiming Kobe as one of the best defenders in the NBA are rife with controversy.  But, the fact remains, Kobe is a very good defender.  Even when he's not trying to be "The Doberman", he's still above average.  Once again, don't take my word for it, ask NBA GMs.  The GMs list Kobe as the 4th best defender in the league, behind only Howard, KG, and Ron Artest.  They list Kobe as the premier perimeter defender in the NBA, ahead of Shane Battier and Ron Artest.  Now, I think we can all agree that the GMs are talking about defensive capability here, instead of actual defensive performance.  I certainly don't believe that Kobe, on an every game basis, plays better defense than Shane Battier.  But I do believe that Kobe is capable of playing defense on the same level as Battier.  Circling back to the main point, any conversation regarding Kobe and defense is going to be equal parts capability and actual performance.

You know a name that isn't appearing on anybody's list of top NBA defenders?  Dwyane Wade.  Wade's defensive reputation is known across the league.  He gambles too much for the defensive "highlight".  He looks for the steal or the block.  I have yet to hear one person say that Wade is anywhere near Bryant's league as a defender.  But, in a PER comparison, Wade picks up a full two PER points on Kobe for his defensive stats.  That's 33% of the difference between the two players, in Wade's favor.  Qualitatively, you'd be hard pressed to find a single GM, or any other knowledgeable NBA personnel, who would rank Wade ahead of Kobe as a defender.  But PER has Wade miles ahead of Bryant defensively.  In fact, PER rates Wade as one of the best defenders in the league, because he has such good stats for blocks and steals.  PER rewards Wade for exactly the type of defensive behavior that most experts criticize him for.  So, not only does PER not correctly evaluate good defense, it also incorrectly rewards stats that can actually be the result of bad defense.

Brandon Roy provides a less flashy example of the same point.  Roy doesn't play specifically for the stats that would increase his PER defensively.  His defensive PER is virtually the same as Kobe's.  But, once again, Roy has miles to go before he and Kobe are considered to be in the same boat as defenders. Earlier this season, talk surrounding Roy was that he needed to make only one more jump to become a truly elite player in this league.  The jump being discussed?  To become a better defender.  Nobody's talking about Kobe needing to make a jump as a defender.  But PER values Roy's defensive contribution and Kobe's defensive contribution as even.

Finally, we come to assists.  I don't want to get into a conversation about assists as a statistic.  There are plenty of people out there who point out how subjective the assist can be.  There was a story about the Hornets employee who actually admitted to "cooking the books" to increase Chris Paul's assist statistics.  There were rumors for years about how "gracious" John Stockton's bookkeepers were in granting dimes in Utah.  But that's not the argument I want to make.  I don't want to denigrate the assist as a statistic. Instead, I want to use the assist to make an argument so cliche that it almost pains me to put it in print:  Kobe Bryant's assists are skewed because he plays in the Triangle Offense.

I know, I know, so much for the original content, right?  First, PER is limited because of its evaluation of defense, and then Kobe's assist statistics are too low because he plays in the triangle?  If you come back next week, Captain Obvious will be able to prove to you that Ray Allen is a good shooter and Tony Parker runs fast.  Look, sometimes you launch into an experiment looking to find something new, only to find that the experiment confirms what was known (or suspected) all along.  That's just the way science works.  It doesn't de-value the experiment.  What would de-value the experiment would be for me to post these numbers and then try to use them to convince you of something that isn't true.  I won't do that, so we're left with what we've known all along. 

And it has been known all along that playing in the triangle offense is not great for a guard looking to pump up his assists statistics.  However, I can add something new to the conversation. 

Quick, off the top of your head, name Kobe Bryant's best season in terms of dropping dimes.  Doesn't matter whether you are going by the more traditional Assists Per Game, or the more advanced Assist % (which measures the number of your team's possessions that end up in an assist by your hand).  You might be thinking it was Kobe's MVP season, the one where he finally "got" it and started using his teammates properly.  You might think about last year, surrounded by enough talent to finally win a championship without Shaq.  You might even go back to one of the Shaq years, the one where Kobe put it all together to become a really great teammate on a championship squad.  If you thought about any of these years, you'd be dead wrong.  Kobe's best assist season came in 2004-2005.

Wait, what?!?  That was Kobe's first year without Shaq.  The year he was finally able to give his selfish tendencies free range.  The year that he was allowed to hoist shot after shot with nary a care in the world.  The year the Lakers went 2-19 to close the season, finishing with a record of 34-48, and missing the playoffs for only the fifth time in franchise history (and the only time in Kobe's career).  The year where he didn't have to pass it to anybody, and didn't really have anybody to pass it to.  How in the hell could Kobe have possibly posted his best assist numbers in that season? 

Because it was also the only season since 2000 that Kobe has played in an offense that wasn't the Triangle.  In 2004-2005, Phil Jackson left and the Lakers brought in Rudy Tomjonovich to run the team.  Rudy T didn't even last the whole season, the Lakers brought back Phil Jackson the next year, and the rest is history.  But, for that one season, the Lakers ran a much more traditional offense, filled with Kobe Bryant pick and rolls.  As a result, Kobe ended up with more assists than in any other season in his career.  It's not a significant increase, but considering the circumstances, and the assumptions most people have about that season, it still speaks volumes.  You want more proof?  Look no further than the king of PER himself, Michael Jordan.  Jordan's career assist numbers are very marginally better than Kobe's.  His best assist seasons all came outside of the Triangle offense.  Jordan may have been a PER legend, but assists were not the difference between his PER and Kobe's.

But, with respect to PER and the Triangle, the assist statistic hurts Kobe more than once.  Sure, his individual assists statistics are lower than they might be if Kobe were in a more traditional offense centered around the pick and roll.  But, while the Triangle is a bad offense in terms of one man getting a high assist count, it is also a great offense in terms of the team as a whole getting assists.  That's why the Lakers were 2nd in the league last year in assists, behind only the Utah Jazz.  Why does a high team assist total, combined with a low individual assist total, hurt Bryant so much?  Because PER punishes you slightly for being a member of a team that gets a lot of assists. 

The logic used is that, if your team gets more assists, it means that more of your baskets are likely assisted, and a basket you make which is assisted isn't worth quite as much as one that is unassisted, in the PER formula.  If you don't like that assessment of things, think of it the other way around.  If an assisted field goal were worth the same as an unassisted field goal (for the scorer), then an assisted field goal would be worth more overall, because of it's worth to the scorer, and to the passer.  Where PER fails isn't in the assumption that an assisted field goal is worth less than an unassisted one.  PER fails because it assumes that, if 60% of a team's baskets are assisted, it is safe to assume that roughly 60% of each individual player's baskets on that team are assisted.  For players like Kobe, we know this to be untrue.  Kobe creates his own shot far more than anyone else on the team, with the possible exception of Pau.  However, the same can be said for Dwyane Wade, Brandon Roy, and any other star.  So, while this is a valid criticism of PER's overall evaluation of star players, it doesn't add a whole lot in determining why Kobe specifically is under-valued.  Regardless, each of Kobe's field goals are worth slightly less than Brandon Roy's, because the Lakers as a whole were a better passing team than the Blazers.

We could spend some time talking about the importance of pace.  Pace adjustment is the reason that Kobe's rebounding stats (slightly higher than both Wade and Roy per 36 min) don't translate to a slightly increased PER rating.  But to pass judgment on PER's pace adjustment is to pass judgment on utilizing pace as a whole.  The idea that Kobe's baskets are worth less than Roy's, because the Lakers play at a fast pace and the Blazer play at a slow pace, might not sit well with you.  But the simple fact remains that, as long as that pace adjustment is held across all statistics (and it is, which is why Wade's turnovers are more costly to his overall score than Kobe's are), it's tough to criticize pace adjustment.

Sadly, in terms of finding loopholes and problems with Kobe's valuation in PER, assists and defense pretty much sums it up.  Any other weaknesses in Kobe's formula are a result of him not being the most efficient scorer in the world.  The truth of that matter is what it is.  Kobe does have the lowest shooting percentage of the players mentioned in this piece.  Since PER is, at its core, a measure of efficiency, it shouldn't be too surprising that Kobe doesn't quite measure up. 

Whether you equate efficiency with greatness or value is an entirely different conversation. 

[Author's note:  I wanted to give a hat tip to our resident stat expert, Dexter Fishmore, who was instrumental in helping me to gain a better understanding of PER so as to write this post.  He also contributed a small portion of the writing to the first post, used verbatim, to help describe PER]