Howard is the DPOY, but he

[Doctor MJ]

I wrote this piece on my blog Howard is the DPOY, but he's no Garnett.

I talk about various points in it, including making the connection between Garnett and Russell's "horizontal game". The focal point though is +/- statistics and this table:



(Details about methods are in the piece)

I'll direct you to Garnett not merely beating the field by utterly destroying the field in studies over 8 years and 2 teams. This is where I just have a tough time not siding with the +/- stat, but I'm really curious to know the thoughts of the stats board on this.

[eyriq]

Howard has anchored a top 5 defense for the past three years, KG never anchored a top 5 defense until he was traded to Boston. A Howard anchored defense is going to be top 5 in limiting opponent FGA at the rim, the most efficient shooting area on the court, the same can't be said for KG. Dwight turns opposing offenses into jump shooting offenses, KG doesn't. He is no KG, he is better.

[NO-KG-AI]

KG makes a bigger difference to his defense than Howard does, and everything backs that up, even at 35, in his 16th season, and multiple knee surgeries. Dwight hasn't sniffed peaked KG yet.

[eyriq]

KG makes a bigger difference to his defense than Howard does, and everything backs that up, even at 35, in his 16th season, and multiple knee surgeries. Dwight hasn't sniffed peaked KG yet.

Everything except for the two huge points I brought up you mean? you've been wrong about Dwight for some time so I get the resistance, but spare me the hyperbole.

[NO-KG-AI]

Except that everything points to the Magic being decent defensively without Dwight. People look at the names and assume they can't play defense, but the Magic never fall off a cliff defensively without Dwight like people claim they would, so there is really nothing backing the statements.

Doesn't really matter if you take away "the most efficient spots" as long as you destroy teh offenses efficiency totally, which KG does better.

Since being in Boston, the Celtics have been better with KG than the Magic have been with Dwight, and when you take both guys away, the Celtics have become the worst of the two defenses. Why is that?

[drza]

More-so than just a Garnett vs Howard comparison, one of the most interesting aspects of your findings is that they challenge the basketball truism that dominant shot-blocking is inherent in being a dominant defender. I like the comparison between Russell and Wilt...both were dominant shot-blockers, but (one) difference was that Russell could do more on the horizontal game than Wilt could. I think that most think of Russell's defense and start/end with the shot-blocking (and maybe bring in rebounding as well), but there was much more nuance to it than that.

I agree that dominant shot-blocking and interior defense are usually characteristics of what we consider to be the dominant defenders of all-time. But those aren't the ONLY ways to be dominant defenders, and there's more than 1 way to skin a cat. I like that your findings give a logical support to the idea that we should continuously re-examine what we "know" about basketball, especially as new findings and methods of analyses help us look at the game from different angles.

[drza]

Just trying to generate some of the conversation that I think Dr. MJ may have been looking for...

...for those that don't find this type of multi-year APM calculation convincing...can you give a reason why?

And I'm not talking things like describing Howard's perceived strengths or KG's perceived weaknesses.

What I mean is, the arguments that I've generally heard against using APM are (in no particular order):

1) The large error margins due to the complexity of the calculation
2) The potential issues with colinearity, trying to separate one player's impact from another's
3) That team quality might skew it. Some argue that being on a bad team can make a player look better, while others have argued that having good teammates can cause a non-impactful player to look more valuable (more of the collinearity argument).
4) That there is a potential system bias (essentially, the Steve Nash argument).

Disregarding, for the moment, whether these are valid criticisms or not... in this example, all four of those are pretty soundly debunked anyway. With 3 different studies, all of them at least 5 years long, and overlapping over an 8-year stretch the noise issue is nil. Over those years KG played with a whole slew of different teammates, under 5 different head coaches and several different defensive schemes. So collinearity isn't a problem, and neither is it a scheme effect. And he played for a pretty varied mix of good defenses, average defenses, and bad defenses...so team caliber (in either direction) also can't explain this.

So, are there any other reasons to distrust this result? I mean, besides it not fitting with your previous expectations.

[DSMok1]

Primary issues with multi-year APM (which I like to use, personally):

1: Aging is not accounted for.
2: It only gives a broad average for the player, weighted toward the years they played more.
3: Context effects are variable if players change teams (not sure of what effect this would have.)

However, if we just want to know how good a player was during a period, it's excellent. (Other than being weighted by Playing Time.)

[Doctor MJ]

So, are there any other reasons to distrust this result? I mean, besides it not fitting with your previous expectations.

Appreciate the effort you're making.

Yeah, I just can't get past these numbers, and I'm looking for real conversation on the subject. Tough to have though. Most people just want to throw out the stat, always based on some kind of thinking I reject.

[Doctor MJ]

Primary issues with multi-year APM (which I like to use, personally):

1: Aging is not accounted for.
2: It only gives a broad average for the player, weighted toward the years they played more.
3: Context effects are variable if players change teams (not sure of what effect this would have.)

However, if we just want to know how good a player was during a period, it's excellent. (Other than being weighted by Playing Time.)

Pretty much agree with everything here, though I'd use "valuable" instead of "good".

One other thing I'd add that I've yet to see anyone attack the standard error component effectively, but I think they could.

Standard error essentially assumes that we're seeing the same player with the same capability every night, and that there's just some randomness thrown in. In reality much of that error isn't randomness at all but player inconsistency and role/matchups differences - and there's certainly no way to precisely state how big those factors are.

Along these same lines, the notion that it is only multi-year studies that are essentially saying "all of this was the exact same player", and are thus doing so in error - and this is an issue that applies to all statistics.

[penbeast0]

LeBron James was individually more effective when Boston defended him one-on-one this season, averaging 86 points per 100 possessions. When the Celtics provided help, James averaged 73 points per 100 possessions; his turnover rate went from 17 to 22 percent and his shooting percentage declined from 48 to 41 percent.

Kevin Garnett was the help defender that Boston used the most; when he helped on James, both the Heat as a team and James were more efficient than with other help defenders (or without any help).

The Heat scored 112 points per 100 possessions with Garnett as the help defender, but just 80 with other Celtics helping. Garnett's value as a defender against the Miami big men was compromised when he helped out on James.

Via Pete Newmann & Dean Oliver/ESPN.com


Read more: http://basketball.realgm.com/wiretap/21 ... z1L7qlUWTz

If this limited sample bears itself out league and career wide for Garnett, does it significantly change your opinion of him as a defensive force?

[Doctor MJ]

LeBron James was individually more effective when Boston defended him one-on-one this season, averaging 86 points per 100 possessions. When the Celtics provided help, James averaged 73 points per 100 possessions; his turnover rate went from 17 to 22 percent and his shooting percentage declined from 48 to 41 percent.

Kevin Garnett was the help defender that Boston used the most; when he helped on James, both the Heat as a team and James were more efficient than with other help defenders (or without any help).

The Heat scored 112 points per 100 possessions with Garnett as the help defender, but just 80 with other Celtics helping. Garnett's value as a defender against the Miami big men was compromised when he helped out on James.

Via Pete Newmann & Dean Oliver/ESPN.com


Read more: http://basketball.realgm.com/wiretap/21 ... z1L7qlUWTz

If this limited sample bears itself out league and career wide for Garnett, does it significantly change your opinion of him as a defensive force?

I don't think so, though I'd be open to hearing the argument.

I look at what they are saying there, and it appears to be:

1) Help defense against a particular Heat player effectively decreased that player's ability to score.

2) When other defenders left their men, the rest of the team was able to keep that man under control.

3) Using Garnett on help defense against LeBron left a gaping hole elsewhere that compromised the defense completely.

With that information, you can just as easily say that Garnett's role defending big men is so valuable, that taking him out of that role kills the defense as say something bad about Garnett.

That's before you even consider that the whole basis for saying those numbers are meaningful is +/- statistics. And if you believe in +/- statistics for a small sample size featuring only one scenario, why wouldn't you consider a more complete study all the more meaningful? And of course, in the complete study, Garnett looks ungodly.

[J08]

howard is the best defender in this decade

[Doctor MJ]

howard is the best defender in this decade

Well, I guess that settles it.

[Vinsanity420]

Garnett is somewhat of an outlier in +/- defensive stats. He's obviously a great defender, but does anyone really think he's just that much of a higher impact player than the rest of the league? Illardi's table especially turns him into some sort of a God of impact basketball. Forget Howard, but looking at those tables you could also say Duncan's no Garnett as well.

I personally like ElGee's defensive stat better ( KG comes out to be ahead of Howard this year in defensive EV, but at least he looks the stat makes him look more human compared to the rest of the league ).

[Doctor MJ]

Garnett is somewhat of an outlier in +/- defensive stats. He's obviously a great defender, but does anyone really think he's just that much of a higher impact player than the rest of the league? Illardi's table especially turns him into some sort of a God of impact basketball. Forget Howard, but looking at those tables you could also say Duncan's no Garnett as well.

I personally like ElGee's defensive stat better ( KG comes out to be ahead of Howard this year in defensive EV, but at least he looks the stat makes him look more human compared to the rest of the league ).

See, I don't know what to say when someone says "I just don't believe it". I mean, you are entitled to your opinion, that's fine, but what's there to say?

Also, when you talk about personally like a stat better because of the results, well what's the point of using stats if you just pick the one you agree with? I do understand that there's a base level of credibility involved when a stat deviates too far from your opinion, but one should be careful how far one goes with that type of thinking.

Something I would point your attention to though: In this year's RAPM study, Garnett leads the league over Howard, and the gap is actually smaller than the gap between Garnett & Howard in ElGee's stat. So while +/- stats do give Garnett a big boost over other defender compared to more traditional metrics, the only evidence we have comparing apples to apples between APM and the stat you prefer shows no freakish Garnett-God gap on the +/- side of things.

[Vinsanity420]

See, I don't know what to say when someone says "I just don't believe it". I mean, you are entitled to your opinion, that's fine, but what's there to say?

Also, when you talk about personally like a stat better because of the results, well what's the point of using stats if you just pick the one you agree with? I do understand that there's a base level of credibility involved when a stat deviates too far from your opinion, but one should be careful how far one goes with that type of thinking.

Something I would point your attention to though: In this year's RAPM study, Garnett leads the league over Howard, and the gap is actually smaller than the gap between Garnett & Howard in ElGee's stat. So while +/- stats do give Garnett a big boost over other defender compared to more traditional metrics, the only evidence we have comparing apples to apples between APM and the stat you prefer shows no freakish Garnett-God gap on the +/- side of things.

What really supports the idea of Garnett having that kind of defensive advantage over the rest of the league for this long? Eye-test? Some other stat?

You have DWS or D Rating, which he is great at, but not absolutely dominant over everyone else like APM suggests. You have team D ratings, and Minny didn't exactly excel in this. Forget team D ratings, Minny didn't exactly excel in points allowed in the paint area either for a number of years.

If you look at just this year, there's also the EZPM stat, and KG doesn't come out to be on the Top, or even Top 5 - http://thecity2.com/2011/01/31/ezpm-v-2 ... he-biggie/

Or using DWARP, which blends EZPM with RAPM

http://thecity2.com/2011/04/13/the-city-2011-dpoy/

I don't have a problem with completely throwing out the eye test if every stat out there suggested the defensive dominance that APM is. I was mainly noting that APM has this tendency of painting Garnett's defense as if it were leaps and bounds ahead of the rest of the league.

[Doctor MJ]

What really supports the idea of Garnett having that kind of defensive advantage over the rest of the league for this long? Eye-test? Some other stat?

You have DWS or D Rating, which he is great at, but not absolutely dominant over everyone else like APM suggests. You have team D ratings, and Minny didn't exactly excel in this. Forget team D ratings, Minny didn't exactly excel in points allowed in the paint area either for a number of years.

If you look at just this year, there's also the EZPM stat, and KG doesn't come out to be on the Top, or even Top 5 - http://thecity2.com/2011/01/31/ezpm-v-2 ... he-biggie/

Or using DWARP, which blends EZPM with RAPM

http://thecity2.com/2011/04/13/the-city-2011-dpoy/

I don't have a problem with completely throwing out the eye test if every stat out there suggested the defensive dominance that APM is. I was mainly noting that APM has this tendency of painting Garnett's defense as if it were leaps and bounds ahead of the rest of the league.

Clearly we start from two very different sets of assumptions.

You I think look at +/- stats as one more of many metrics. If it, and it alone deviates from the other metrics, and you think the other metrics generally are pretty good, that makes the deviator suspect.

Where I come from is looking at the theories involved. Box score base stats always miss the more subtle parts of the game, and arguably most of those parts are on defense. This is why I started looking for alternatives to consider.

+/- doesn't have that kind of weakness. To my mind, it's only real relative weakness in inaccuracy is that it is more susceptible to small sample size issues than the other stats. That weakness is significant enough that when I rate players over a given season I don't necessarily give it even as much wait as some other metrics. However we are talking about 8 years worth of data here with Garnett on 2 different teams showing Garnett's impact way ahead of everyone else.

At that point, I need a plausible argument for why I should discount this data, and mentioning other stats goes nowhere because I'm looking at +/- specifically because of how flawed I know the other stats are.

Let's also be clear. Pretty much any list of the top defenders of the past decade would have Garnett in the top 5, and Garnett plays defense with notably different strengths and weakness than the other top defenders. We aren't talking about a random player here.

I look at the top 10s listed above, and they look pretty dang reasonable. To me that's how you should judge the sanity check. If the results in general look like what you'd expect, and then you see one unexpected result, that exactly what you hope a new stat will tell you.

[EvanZ]

Garnett is a great help defender, and those kinds of guys are going to be underrated by box score or PBP stats (like mine). On the other hand, Howard is basically a huge deterrent to players driving to the rim. According to Hoopdata, ORL had the lowest % of opponents taking shots at the rim (24.8%) and the lowest opponent XeFG%, simply meaning that they forced more difficult shots. My assumption has been and still is that Howard is primarily responsible for that. They both help their teammates tremendously, but in different ways.

[drza]

Garnett is a great help defender, and those kinds of guys are going to be underrated by box score or PBP stats (like mine). On the other hand, Howard is basically a huge deterrent to players driving to the rim. According to Hoopdata, ORL had the lowest % of opponents taking shots at the rim (24.8%) and the lowest opponent XeFG%, simply meaning that they forced more difficult shots. My assumption has been and still is that Howard is primarily responsible for that. They both help their teammates tremendously, but in different ways.

I tend to agree with this, but I think where some of the disconnect betwen perception and reality sets in when it comes to Garnett is that even his help defense is unique. We've seen lane/rim protectors like Howard before, and in fact his style of defense is what we've come to expect from defensive anchors. Even if Howard doesn't block every shot, we can watch a game and tell exactly where his effect lies. We can see folks hesitating to drive the lane, then if they do drive changing their shots. As you point out, you can see in the percentage of opponent shots in the lane where Howard's effect is.

The problem with Garnett's brand of help D is...how do you find it while watching the game? Much like the subtleties of offense (off-ball movement, hockey assists, picks, correct ad-libs, etc.) it is difficult to really track EVERYTHING that occurs on defense. The human attention span can only encompass so much. But from watching Garnett through the years, his specific defensive contributions aren't always so easy to point out. It's not like he's always trapping, or always blocking shots at the rim, or always doing any one specific thing at any one specific location. The closest one-phrase definition like that for Garnett is "best pick-and-roll defender in the NBA", which I've heard many ascribe to him. And though it's likely true, it still doesn't IMO encompass KG's weird defensive impact.

The best way I can describe it ties in something that one poster...I believe it's NO-AI-KG, but not sure...used to point out all the time: part of the reason that KG didn't have crazier blocked-shot totals in his stupidly athletic youth was that he tended to rotate to the offensive player BEFORE they could get the shot off. So in the pick-and-roll that people have noticed, KG might show hard on the dribbler to stop his shot/drive, then recover to his own man and prevent that player from shooting/driving, which essentially forces the play to reset. Only...what would you possibly credit KG with in the stat sheet? He didn't steal it, or get a rebound, or block the shot...heck, there WAS no shot to block. But he still disrupted the play and likely forced the offense into a lower percentage look.

I've never seen anyone attempt to chart "shot prevention"...I'm not really sure how you could, but I would bet that if such a project were ever undertaken Garnett would lead this generation. And he does it in different ways. I talked about the pick-and-roll, but he can also do it in 1-on-1 situations. For example, in the Knicks series game 1 Amare was just on fire, then he didn't touch the ball in the final 2:30. After the game everyone blamed Melo and D'Antoni for not getting Amare the ball, but hardly anyone really mentioned how difficult Garnett was making it to get Amare the ball except D'Antoni, who said "Kevin was draped all over him". Now, one could easily say that the onus was still on the Knicks to get Amare the ball, but the point as far as this thread goes is that Garnett's length and agressiveness made it much more difficult for the Knicks to get Amare the ball than normal, which changed the odds on what the Knicks wanted to do, and lowered their team scoring likelihood in a way that doesn't show up in the stat sheet or really in most people's conscious thoughts.

And that was just one recent example of 1 stretch of game. But last year in the '10 playoffs, despite Garnett's obvious mobility issues, you can see a similar shot prevention effect...if you look. Last year I made a point of checking the play-by-play info to see what Garnett's main assignment did offensively when KG was on the court and assigned to him (i.e. I didn't keep track of switches or fast breaks, just raw what did that player produce when KG was assigned on him). I tracked the first 22 playoff games (5 for Beasley, 6 for Jamison, 6 for Shard Lewis, and 5 for Gasol) and my money shot at the time was that those four players averaged:

17.4 points/36 minutes on 47% shooting in the regular season
18.6 points/36 minutes on 56% FG in the postseason when not playing the Celtics
20.6 points/36 minutes on 53% FG against the Celtics when KG wasn't assigned to them (234 min)
9.0 points/36 minutes on 36% FG during the time that KG was assigned to them (523 min)

The massive drop in scoring and field goal percentage were what attracted my notice at the time, but as I wrote this I went back and looked at those players field goals attempted and...sure enough...they averaged 14.1 FGA/36 in regular season, 12.8 FGA/36 in postseason not Celtics, 13.5 FGA/36 in postseason against Celtics/no KG, but only 9.7 FGA/36 in postseason with KG assigned to them. Again, shot prevention.

One last anecdotal evidence point from recent playoffs, this time dealing with help non-pick-and-roll defense is Knicks game 2, when Melo set Jeffries up for the potential game-winning layup at the end. Again, what most noticed/remembered was that Jeffries didn't shoot and Garnett got the steal. But the reason why Jeffries hesitated to shoot was that by the time he caught the ball and turned Garnett had completely rotated to him, in enough time to have his own feet fully set to go for the block. People remember the steal, but the point I'm getting at here is the actual "shot prevention" effect from Garnett's quick rotation.

Conclusion. Again, the examples I gave are just that: examples. I haven't done any kind of long-term tracking on this "shot prevention" theory that kind of just popped into my head. It'd be a hard thing to track, and on top of that it's not like Garnett always has the exact same impact. Last postseason he lacked mobility and his impact seemed to be more on his 1-on-1 assignment, but in general he's known more for his help D and maybe his own assignment might have better numbers. But if you just go back to the physical tools...KG's defensive mobility and range are extremely rare among 7-footers, his length is uncommon even among 7-footers, his defensive aggressiveness/mindset is uncommon for any position, and his understanding of defensive sets and opponent tendencies also seem to be uncommon for defenders in general. Put that together, and based on the anecdotal examples I presented just from KG's time as a 34-year old I feel pretty confident that this "shot prevention" idea would likely pan out as something that Garnett would likely lead the league in for this generation without it ever really being capturable.

But if you combine tangibles like defensive rebounding, blocked shots and steals of which he still is AMONG the best of his generation and accounts for his strong ratings in the box-score-based defensive measures... with tangible-but-as-yet-uncaptured like this "shot prevention" concept or the "middle linebacker/verbal teammate coaching" that many credit Garnett with being the best at...with intangibles like Garnett's fanatic approach to practice (which has to in some way affect his teammates) or the "intensity/energy" that he provides on that side of the ball...eventually, you start getting into layers of impact in combinations that we've rarely if ever seen before. It's a package that's DIFFERENT than the traditional rim/paint protector which we usually think of, but could go a ways towards describing how Garnett could individually be having such a huge, measurable impact on opponents' ability to score without it showing up in the boxes or even always in our conscious thoughts as we watch the games.