A Substitute for War

Basketball philosophy

Howard is the DPOY, but he’s no Garnett

with 12 comments

Kevin Garnett led the league in defensive rebo...

Image via Wikipedia

Dwight Howard just won his 3rd straight Defensive Player of the Year, and I don’t disagree with the choice. In fact, I don’t disagree with any of the three votings that gave Howard the award. However, with Howard becoming the first player to win the award three times in a row comes discussion of how Howard stands compared to the best defenders of other eras, and there is a problem with this: Howard isn’t even the most impactful defender of this era, that would be Kevin Garnett.

Let’s start out by acknowledging that Howard and Garnett don’t play defense the same way. When we think of the great defensive big men, we tend to think of blocked shots. Howard fits that bill significantly more than Garnett. Although lets pause and consider that Howard’s 2.3 BPG this year, is only just ahead of Garnett’s 2.2 BPG peak – and that Howard’s never actually averaged 3 blocks per game in his career. If Howard were putting up these blocking numbers in other eras, we wouldn’t look at him as anything like the shotblocking ideal he often gets talked about today.

Still, thinking about defense in these terms, it’s not at all hard to see why people think Howard’s the superior defender when comparing peak to peak, and especially now as Garnett ages. Add in Howard’s current rebounding edge, and the fact that Orlando always ranks well on defense despite Howard’s supporting cast not having a stellar defensive reputation, and the debate is over before it begins in a lot of people’s minds.

Garnett, Russell & the Horizontal Game 

Garnett’s defense is significantly harder to sum up. Everyone who has watched him knows that his string bean frame combines great length with great quickness. There is some appreciation for how this allows him to be more versatile in who he guards than anyone else in the game. There is clear and glowing appreciation for Garnett’s religious intensity both in his effort on the floor, and the clear emotional push he gives his team.

More recently he’s been getting attention for what I call his “middle linebacker” role for his team’s defense. Meaning, he’s the floor general of his team’s defense, picking up on the opposition’s moves and strategems and directing his teammates.

I don’t think anyone would disagree that these are all good things, or that Garnett is a great defender, but most are reluctant to elevate Garnett too definitively based on these observations and I completely understand such hesitation. It’s not easy to compare two defenders who don’t have identical strengths and weaknesses – so most just go with what’s traditionally understood to be the gold standard.

I think it may be helpful if I could knock down the idea of the gold standard to some degree. Take a look at this quote from Bill Russell: A Biography:

Bill understood that Wilt’s game was more vertical, that is, from the floor to the basket. Wilt’s game was one of strength and power…Bill’s game was built on finesse and speed, what he called a horizontal game, as he moved back and forth across the court blocking shots, running the floor, and playing team defense.

I realize that it talks about Bill Russell blocking shots, but understand Wilt Chamberlain blocked shots like crazy too. What the book is referring to is Russell’s quickness edge, his ability to race out and around to challenge the opposition and recover back to the interior without getting burned, as well as the intelligence with which he played defense.

In short, that which gave Russell the edge over Chamberlain is exactly what Garnett has over Howard. Now of course, Russell has the additional edge that Garnett does not in his shotblocking, but let’s keep in mind that with the rise of the outside game shotblocking is not as powerful a force as it used to be and that I’m not saying Garnett’s defensive impact should be considered the equal of Russell’s.

Garnett lapping the field

Still not enough to sway you? Understandable, and this where these new fangled advanced stats become so revolutionary. You’ve seen me wax optimistic about the benefits of +/- statistics before, here is where I find them extremely powerful. Below are the top 10 players by 3 different studies among players who most of the years in the study and did so playing more than 24 minutes per games:

The 3 studies are:

1) Ilardi’s 6 Year Adjusted +/- from ’03-04 to ’08-09.

2) Engelmann’s 6 Year Regularized Adjusted +/- from ’05-06 to ’10-11.

3) Engelmann’s 4 Year Regularized Adjusted +/- from ’07-08 to ’10-11.

Now don’t worry about the shift in scale of the numbers. The “regularization” technique Engelmann uses means you can’t actually look at the numbers as literally representing the number of points a player is lifting his team – but it makes the rankings and proportions between the players more precise.

As far as why I cut off the ranking based on playing enough, I do that because I think that gives us the best picture of who would be in the running for defensive accolades based on the metric, and because I there’s a lot of distortion applying the metric to non-big minute guys. If you want to knock this approach as cherry-picking that’s your right, and we can debate. I’ll certainly agree it’s arbitrary.

Looking analyzing these metric, the first thing you’ll see is that Garnett leads all 3 of them. That’s nice, but there are two more “Holy Crap!” things to consider:

1) On all 3 metrics, the gap between Garnett and the #2 guy in the league is bigger than the gap between the #2 and the #10 by a VERY large margin.

2) Garnett switched teams before ’07-08, and it didn’t change this in the slightest. And in case you’re thinking that all 3 studies include time in Boston, understand that Garnett’s +/- peak was when he was in Minnesota, and that even Ilardi’s study doesn’t cover all of that.

That kind of consistent domination in the metric – how the heck did Garnett pull that off if wasn’t due to him actually having more impact than other defenders?

Enlighten me if you’ve got an answer, I don’t see one. Yes +/- has a reputation for being fragile to low sample size, but we’re talking about 8 years of sample size here, and Garnett’s defense remains off the charts throughout it.

Coming back to this Howard. You’ll note Howard’s absence by the first two metrics. On the ’03-04 to ’08-09 study, he’s nowhere near the top. On the ’05-06 to ’10-11 study, he’d have made the top 15. He reaches the #4 spot when we just go by the last 4 years, so you can see he’s rising up these lists just as we’d expect – but still not in Garnett’s league.

Didn’t I say I was fine with Howard winning the past 3 DPOY’s though? Well of course, in the past 4 years, Garnett’s started playing a lot less. Less minutes per game, and less games period. Gotta factor that in, and once you do, Garnett certainly can’t be considered a runaway winner for the award. The unfortunate aspect, is that in all those previous year in Minnesota, Garnett never won DPOY. I’d say he should have won several.

The Powers of Now

Looking at this year, the DPOY voting had Howard at 1, and Garnett at 2. Probably the only time the two will finish in the top 2 spots together, and hence perhaps the only time the voting truly got it right.

Below is a table of several different metrics that have a claim on measure per minute defense:

DRtg is Dean Oliver‘s defensive rating attempts to measure a player’s defensive impact by using that player’s blocks and steals, and the general defensive success of the player’s team. By Oliver’s own admission, this is very much a flawed stat because it doesn’t have any method of directly measuring the rest of a player’s defensive game.

The lower a player’s DRtg rating the better, so Howard has the edge in this stat just barely (Howard’s #1 in the league, Garnett’s #2). Consider though, this is a stat that overrates blocks & steals, and underrates man defense, sharpness of defensive rotation, and floor generalship, which are all areas where Garnett has the edge over Howard, and Garnett *still* almost beats him out.

Second, we have 82games.com‘s opponent’s production, in which again the smaller number is better. Here Howard has a more substantial lead – although again, this is based on traditional box score stats which gives Howard an edge.

Third, D RAPM is Engelmann’s study of defensive regularized adjusted +/- for this year. Note that Garnett once again takes the top spot by this metric, but that Howard has now risen to the #2 spot.

For good measure I’ve included defensive Expected Value as tracked by ElGee over at Back Picks, which estimates value added by tracking the good and bad things a defender does, and let’s the opposing team do. I like that this metric goes because it really starts to get into track smart man and team defense. On the other hand, the data is far from complete. I would not feel comfortable relying too much on this, but it is worth noting that it comes the closest to filling out the weak points of Oliver’s and 82games’ metrics, and it ends up giving the nod to Garnett pretty decisively.

With all of this in mind, and what we know about the bigger picture, it seems quite clear that a strong argument can be made for Garnett as the best defender in the game when he’s on the floor. This in of itself is astounding. To be in this position in his 16th year in the league is completely unprecedented.

With his limited play though, I do think Howard is the right choice for DPOY again.

I just hope though, as Howard likely continues to win DPOYs and breaks the record for the award held by Dikembe Mutombo and Ben Wallace, there will be a decent contingent of us who continue to remind the basketball cognoscenti of how extraordinary Kevin Garnett was.

Written by Matt Johnson

April 23, 2011 at 12:07 am

12 Responses

Subscribe to comments with RSS.

  1. Nice writeup. Totally agree. Howard doesn’t look as good in the 05-11 ranking because much of the defensive credit goes to StanVGundy though, since it’s an analysis with coaches factored in

    Eric Cartman

    April 23, 2011 at 5:28 am

  2. Wow. Great post. I need to think about this for a while. Is there any way to explain away KG’s huge edge pre-Boston?

    Always thought of him as an excellent defender — and an impressive looking one ITO of versatility — before coming to Boston. But the team DRtg’s were never spectacular. KG was 2nd in DPOY voting (26 first place votes) in 2001 and 2nd 10 years later.

    Now, I wonder how different San Antonio’s results on that end would have been if KG were swapped in for Duncan (and I have enormous respect for Duncan’s D).

    For someone who constantly rambles about how much team results bias our estimations of individuals, I’m wondering if I haven’t undersold Garnett on some of those Minny teams. Or, at the least, I’m wondering where these numbers are coming from.


    April 23, 2011 at 5:00 pm

  3. I’ve never thought much of Garnett as a man-defender, and yet one of the main arguments against individuated DRtg is through position defense, as you state.

    Garnett was a player that excelled through his athleticism and length in a team structure, as a team defender and anchor.

    But man to man? Well, he’s no Dennis Rodman. And a number of other names, past and present, come to mind as well.

    As a team captain, I thought Garnett was close to miserable in Minnesota. He didn’t take over games — observance, rather than pure stat-tracking, I admit — offensively, he wasn’t a volume monster from the 4 slot through scoring, and defensively there was never the thought that Minny was an intimidating team.

    The DRtg for his career there supports this. That was a mediocre defensive team, year and year out; about the only exception came in his MVP season, where they finally cracked the top 10.

    Now Howard? His teams have consistently been in the top 3 over the last 4 seasons — basically what we know of his prime or peak thus far — and even dating back to ’06 Orlando’s DRtg is in the top 6 for the league, matching Minnesota’s peak relative to average.

    Garnett’s career with Minnesota, minus his rookie campaign, gives us a picture that is less than impressive as far as team DRtg: a yearly rank of 14.5 over 11 seasons. A league mediocrity on defensive enforcement by team-DRtg.

    Dwight Howard, minus his rookie season, has been on teams that post an average rank of 8.6, and a finish in the top 6 or better for 4 of those seasons, including three at 3 or better for league average; Garnett didn’t achieve a top ten or better finish until his 8th season!

    If this was a question of position defense, a guy taking his man — assumptively the star player — and stopping him or hugely hindering him, I’d be more sympathetic.

    But that was never Garnett’s game. No, his was as a rim-protecting anchor of the Mutombo/Olajuwon mold, yet the stats don’t match up with the assumption on impact; certainly not the team stats, or their results.

    And the argument finding its base in +/- assumptions? It’s really hard to take, as this is one of the most notoriously…questionable stats in the league.

    Plus/minus might have made John Stockton a better offensive player than Michael Jordan in the 90’s. But what does that prove? What?

    It’s also a largely insular standing, standard and finding from that; meaning that perhaps Garnett played with lacking teammates — as an explanation for his teams’ lacking DRtg? — but this would likewise be an argument that means his plus/minus stats from that very foundation were overrated.

    If his plus/minus was of such value while his teams were putting up DRtgs like Howard’s, I’d be more inclined to agree with your narrative.

    But not as it is. And not when I consider my own findings on Garnett’s defensive worth, both statistically and in on-court contexts. His supporters use this as a standard that must elevate him in some way, especially when considering his timidity on offense, yet the result never matched the hype; not in his prime, not at his peak.

    As an addendum, Garnett gets great credit for Boston’s defensive turnaround and success over the last few years, yet it’s very interesting as a corollary that his best defensive squads have been those on which he plays less than 35 minutes a game.

    The fact that he gets a shout-out so often on leadership at this point in his career, certainly does pose questions of his viability in that role when with the T’Wolves as a logical extension. And his play on that level was simply compounding.

    I’m not completely sold on DRtg, as it has flaws that are deeply embedded, but I think it’s rather hard to escape it as a standard when something like +/- comes into the picture.

    So far as Mutombo’s name, I don’t recall any series that Garnett dominated as a primetime player the way Mutombo did in the mid-90’s. Not one.

    Bill Russel? Um…

    Team DRtg, in this instance, certainly does match what my lying eyes told me about Garnett’s impact on Minnesota.

    And I don’t understand how advocates for ORtg’s primacy can be so dismissive of DRtg in this instance.

    An evaluation of what advanced metric are truly worthy, and why, would go a long way toward solidifying arguments in general from the poster and the message of this blog.

    I do appreciate the effort, however.

    Though I can’t agree with it, and I do wonder if it’s altogether agreeable with other arguments provided up to this point as far as team-wide impact and the metrics chosen to define it.

    Nonetheless, the blog is proving interesting, even when disagreeable on whatever level, be that within or without.


    April 24, 2011 at 6:28 am

    • It is good to get a dissenting view-point to Matt’s findings, but here I must note that many of your rebuttals don’t seem to stand up to further examination as far as specificity goes. Said another way, your rebuttals read reasonably as general findings, but further scrutiny of your case doesn’t seem to support your conclusions. For instance, the only quantitative evidence you use to support your defensive rebuttals is that the Wolves’ team defensive rating was worse than the Magic’s. But by definition, team defensive rating is heavily dependent on a player’s teammates and coaching. It isn’t an individual measure. Even individual defensive ratings will be heavily skewed by factors outside of the individual.

      But all of that said, if one were looking to isolate individual impact from defensive ratings data the logical first step would be to compare an individual’s defensive rating to the team’s defensive rating. And if you do that, you see that Garnett’s defensive ratings are DRAMATICALLY lower than his team’s through the years. In fact, the difference between Garnett’s defensive ratings and his team tends to be larger than the gap between stalwarts like Duncan or Howard compared with their respective teams. And what’s more, that difference is consistent no matter the quality of the team’s defense (i.e. it’s not just a case where he looks good on average defenses, but not so much on great defenses). Again, defensive rating isn’t ideally suited as an individual measure, but even starting there the reasonable first-blush conclusion would be that Garnett was pulling lesser defensive players to form defenses much stronger than they could have produced without him. And it’s really hard to form a much more concrete conclusion based only upon defensive ratings.

      What defensive APM does is follow that level of brute-force analysis to its logical end-point. It adds up all of the individual contributions to a team’s defense from each of the players that plays and contributes over the time period (some of the recent APM efforts also attempt to factor in coaching). Obviously this is a complex mathematical problem, and over short periods (even a year or 2) this can lead to comparatively large errors. But as Matt points out, we’re looking at 8 years worth of data here in overlapping 5 – 6 year chunks. Noise becomes a negligible issue, especially when the difference in results is so large and so consistent.

      In other words, this isn’t just a case of “my stat or yours” with regards to defensive rating vs APM. If you actually go into reasonable depth from either angle, the story is similar and the way the two stats are calculated makes one inherently more suited for individual comparison than the other.


      April 25, 2011 at 4:29 pm

  4. As one other thought, the replacement stats aren’t necessarily helping their own cause, or Garnett, as a career aggregate versus — as opposed to a complementary piece — that of career shift and improvement of team DRtg through teammates.

    For if Garnett’s teammates improved, then why is it not showing in the +/- stats? If they didn’t improve, then what does that say about his time in Minnesota?

    Boston is obviously a much better team, but these stats don’t seem to be pointing that out when looking at Garnett’s presence or individual standing relative to that of his teammates, which makes the entire thing rather questionable numerically and as logical finding from that.

    The question you posed above — why didn’t the +/- stats shift with his role and team? — may be more damning of the method than enlightening as far as Garnett individually.

    I will say this: it’s fairly possible that things have shifted, and that comes with the fact that Garnett is not necessarily the team’s number 1 guy the way he was in Minnesota, particularly in an offensive role that he was unable to fill.

    By taking off that pressure — by putting Pierce and others in that role — perhaps he became a better defensive player, as that was now allowed to be his key focus.

    If that’s the case, it makes him a questionable number 1 — a stable facilitator and scorer — of the mold he’s often argued as being equal to.

    In other words, for all the flack Barkley got for his defense, it’s very arguable that Garnett deserved and deserves the same as far as his worth on offense.

    Through that? I think questions arise as to his viability, at any point, as a team’s clear Number 1 in a contending context.


    April 24, 2011 at 6:47 am

    • Maglight, lots of good thoughts in here. It’s honestly good to have them here as counterpoint to what I’ve written. As you say, I appreciate your efforts, and I’m glad you find the blog worth a read even if only as a launching point for you to make your counter-tome.

      Responding to the points that most resemble questions:

      “Plus/minus might have made John Stockton a better offensive player than Michael Jordan in the 90′s. But what does that prove? What?”

      You’re using a “What if?” statement about +/- against its credibility. I don’t see the sense in that. I mean I get where you’re going with the question: Maybe complete faith in +/- would lead to absurd opinions if we had all the information.

      I hardly have complete faith in +/- though. It’s merely one factor I use, and I technically side against it quite a bit. The Garnett case is special for all the reasons I’ve given above, and it’s also not like I’m elevating a B-lister here. Dude’s won DPOY and is about to tie the record for All-Defense 1st team.

      “It’s also a largely insular standing, standard and finding from that; meaning that perhaps Garnett played with lacking teammates — as an explanation for his teams’ lacking DRtg? — but this would likewise be an argument that means his plus/minus stats from that very foundation were overrated.

      If his plus/minus was of such value while his teams were putting up DRtgs like Howard’s, I’d be more inclined to agree with your narrative.”

      We’re using adjusted +/- here. You can’t knock it simply by saying he had weak teammates. The bias, if any, is far smaller, and far more complicated than that.

      As far as you being more likely to buy my narrative if his team’s defense was better – of course you would, because my narrative would less informative to hear. The story I tell is jarring, I’m quite aware of it, and I don’t expect it to convince everyone. I justhope it sticks in your mind as something to consider.

      Regarding Mutombo & Russell, y’know I didn’t actually compare him to these guys right? The closest I came was to say he had some things in common with Russell before making clear that Garnett didn’t have Russell’s shotblocking and that I wasn’t saying Garnett’s impact was on Russell’s level.

      “For if Garnett’s teammates improved, then why is it not showing in the +/- stats? If they didn’t improve, then what does that say about his time in Minnesota?”

      Huh, I made a response about +/- above, but now I don’t think I can assume you are familiar with the basics of the stat. Forgive me if this I underestimate your knowledge here, I just want to make sure everything’s out on the table.

      When we use adjusted +/- (APM), the idea is to factor in the player’s teammates and opponents and measure how well the team does with him compared to what one would expect with an average player. Ideally then a player’s APM numbers should be identical on different teams of vastly different qualities, and thus that actually happening says GOOD things about the stat.

      To be clear though, players have different value in different roles, so I would often expect a player’s APM to change when they move to different teams.

      “I will say this: it’s fairly possible that things have shifted, and that comes with the fact that Garnett is not necessarily the team’s number 1 guy the way he was in Minnesota, particularly in an offensive role that he was unable to fill.

      By taking off that pressure — by putting Pierce and others in that role — perhaps he became a better defensive player, as that was now allowed to be his key focus.”

      The point that Garnett has a different, more defense-oriented role in Boston is a good one to consider. The fact that Garnett has been such an extreme outlier on both teams though makes that hard for me to see as damning.

      Now let’s note, that Garnett’s offensive APM has fallen to the floor in Boston, just like one might expect as he’s gotten older and gone to a team with other offensive stars.

      On the first metric, Garnett’s offensive APM ranked 6th among all players, better than Pierce or Allen, and he was easily the best among big men. His overall APM utterly dwarfed the entire rest of the league.

      On the last two metrics, Garnett’s offensive APM has fallen below Pierce and Allen. Though his overall APM remains superior to his teammates and top 5 in the game.

      Matt Johnson

      April 24, 2011 at 12:35 pm

  5. Great post. I applaud you for putting in the effort to compile these numbers all into one easy-to-read table in one eminently reference-able post. I’ve used these numbers in debates, but it is so much more powerful to have it as a standalone post with informative and potentially explanatory context-writing support as opposed to embedded within random message board posts.

    And I really like the Russell horizontal-defense comparison because, much like your bulls-eye example from the APM post, it does a great job analogizing a difficult concept. I’ve been trying to describe that concept in various discussions through the years, as I’ve tried to point out that the traditional notion of a dominant shot-blocker being the best possible influence on a defense doesn’t mean that this is the ONLY way to be a dominant defender. And your results here suggest that if the dominant shot-blocker were ever the pre-eminent defensive archetype, in today’s game that tends to be shifting a bit.

    This was an informative post, and I hope it generates a lot of discussion.


    April 25, 2011 at 2:44 pm

  6. Matt. Write a book. If a hack like Simmons or a flawed polemicist like Berri can get published, then surely your (and Julien’s) quite sophisticated writing and analysis could find SOME sort of paid audience.


    April 25, 2011 at 11:54 pm

    • Really appreciate the kind words. Super stressful week right now, wouldn’t be driven to write more without the encouragement of my comrades.


      Matt Johnson

      April 26, 2011 at 9:56 pm

  7. […] Howard is the DPOY, but he’s no Garnett (asubstituteforwar.com) […]

  8. […] for someone that height. While it might work to have someone a bit taller here (Kevin Garnett comes to mind here), the agility is absolutely essential. And we’ve never seen a super-super-tall guy come […]

  9. […] These last 3 sortings are similar to what I did in my article on Kevin Garnett being the best defender of this era. […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: