Layups: Best Defenders in 2011
Posted by Neil Paine on April 13, 2011
At Back Picks, ElGee took a look at the top individual defenders of the 2010-11 NBA season.
Posted by Neil Paine on April 13, 2011
At Back Picks, ElGee took a look at the top individual defenders of the 2010-11 NBA season.
April 13th, 2011 at 11:00 am
"Top Individual Defensive Players from Top Teams that I Happen to DVR Their Games This Season and Analyze" therefore Mike Miller is a top defender.
April 13th, 2011 at 12:22 pm
I agree with #1. The sample sizes are laughably small.
April 13th, 2011 at 12:43 pm
Mike Miller has terrible counterpart per on 82games.com, so that should be incorporated into the analysis.
April 13th, 2011 at 1:14 pm
@Huevonkiller
Counterpart PER doesn't pass the laugh test - while Howard is #2, Garnett is #28.
Garnett is #2 in defensive win-shares, and #1 in regularized defensive plus-minus (7th in unadjusted Net defensive Plus-Minus).
But--Kevin Love is the prime reason why PER is so obviously wrong -- rebounds are NOT that important (as was highlighted by Engalmann's study. He is 4th in 82games' play-by-play PER and 6th in Net PER, yet his impact is somewhere between neutral and a little above neutral.
April 13th, 2011 at 2:01 pm
Oh, I agree with #1 too that sample size is a problem. The stat-tracking information should be balanced against other information -- that's why the sample sizes are provided. (And if one has a synergy account, that too.) The Defensive EV is merely a sorting tool here, it's not a definitive ranking.
My question is, would rather have none of the information, or 5-25 games of data to work with? (It's19 Heat games...I wouldn't call Miller's sample "laughably small.")
April 13th, 2011 at 2:04 pm
Okay, I'm glad to see that I'm not the only one who thought this was a fairly bizarre analysis.
April 13th, 2011 at 3:40 pm
#5 - No, it's not Miller's sample that's laughably small, his sample is actually laughably big. The problem is I'd rather this article not proclaim to be an analysis of the "NBA season's best defenders" when it's not actually looking at the entire NBA season. It says right there in the methodology that data is pulled from games recorded on DVR, and that last place teams didn't get enough data, so I can only assume the entire season wasn't actually considered, and probably only nationally televized games were recorded (hence Mike Miller's large sample). Like how can you only have 23 possessions of Iguodala, who would make a fantastic topic when discussion this year's best defenders?
April 13th, 2011 at 4:56 pm
In a Neil Paine special, I've added a D RAPM run to the final table (I thought I had that in originally but apparently not).
#7 That's 23 "used" defensive possessions for Iguodala. I'm including him because of all the defenders who are in the discussion, he's really the main guy who is really small-sampled for my stat. Wings have a lower defensive usage on average than bigs, so it takes more time to accrue large samples of "relevant" defensive events. I only have 4 76er games because, sadly, they aren't on TV much (and that will double in the playoffs, at minimum.)
Again though, this should be looked at as a cross-section of defensive information. It's my contention that we are grossly lacking defensive stats in basketball, and this is an attempt to look at the rough tools we have and paint a picture. Call me crazy, but I don't think overlooking Minnesota, New Jersey, Toronto, Washington, Cleveland and Sacramento is doing a disservice to this discussion. What good defender is left out there, anyway?
April 14th, 2011 at 2:58 am
Nathan your list is laughable, stop being a hypocrite. Your list is sure getting cute, don't know about reasonable though. Chris Bosh above Wade and Bryant? Steve Nash? I'll go with SPM if I want to incorporate plus-minus. Not adjusted plus minus nonsense.
Nathan you're painting a fuzzy picture aren't you? Rebounds aren't that important, yet you're citing the win shares formula that clearly demonstrates Kevin Love is an elite player? What is your point? The Timberwolves are surrounded by incompetent defenders, not named Kevin Love. This has led to their struggles, counterpart PER also reflects that.
Counterpart PER is the study cited in this article, and it doesn't do that correctly. So it should be mentioned.
April 14th, 2011 at 4:59 am
#4
1. There was no study, just numbers
2. If you cite someone's name, try to at least get it somewhat right
April 14th, 2011 at 9:20 am
#8: If Mike Miller and Brandon Jennings can make it to the tops of your lists, anything is possible!
April 14th, 2011 at 9:30 am
#8 - first of all, good on you for giving this a shot. I agree that there are problems with some of the method, but I like the effort. I know I wouldn't have spent the time required to make that post happen.
However, the thing about stats is that you never know what is hiding in there. I'd rather analysis skip Minnesota, Toronto, New Jersey, Cleveland, and Sacramento because of time/resource constraints than an assumption that there will not be valuable data.
April 14th, 2011 at 1:48 pm
@Huevonkiller
Study after study have shown that PER overvalues rebounds. Dean Oliver's ORTG on which Win Shares are based doesn't particularly have this issue. Why are you throwing out all these ad hominem attacks on me? I'm just presenting an argument.
April 14th, 2011 at 2:00 pm
@Huevonkiller
Ridge plus-minus numbers have two benefits that no other statistic can give us.
First, that it is extremely conservative: a player must show continued, repeatable performance in order to do well in ridge +/- (and vice versa) because of the penalty factor. One way of thinking about it is that the algorithm basically only changes the values (up or down) when it is 'very confident.'
Second, that it is extremely predictive: using these numbers to predict any given possession is by far the most accurate of any single system, as has been shown by its very low standard errors. Next, in terms of predictive accuracy, would be statistical plus minuses and non-regularized adjusted plus-minuses...next would be something that is shown to have decent equal predictive value like Win Shares per 48 or even PER.
A theory's ability to explain the past (i.e. Points Per Game, or, Games Won) is NOT the best explainer of 'true ability.' My philosophy thesis briefly touches on the subject, so here's a primer:
http://en.wikipedia.org/wiki/Predictive_power
I'm not saying this to be condescending, rather to just throw out there that explaining the past is a good guess, but predicting the future is a much, much better guess.
April 14th, 2011 at 2:01 pm
(when I say "no other statistic can give us" I mean to say "that few other statistics can give us)
April 14th, 2011 at 2:49 pm
http://www.basketball-reference.com/blog/?page_id=9221
That isn't perfect but appears far more reasonable. I don't rely on one metric but I prefer this one much more.
April 14th, 2011 at 4:40 pm
How is extending or ending a possession not important? That's what rebounding does for a team.
April 14th, 2011 at 6:17 pm
#12 - Good point. I have data on every team, just less than 5 games on those weaker clubs who are rarely on TV.