What's the best defensive metric

[The Beam King]

To help evaluate a player's defensive worth to a team? I understand metrics aren't the silver bullet to establish one player better than the other, but much like PER is to offense, a nice guideline, is there a similar statistic/metric to evaluate an individual's defense? Thanks

[NO-KG-AI]

I like to compare a guy's defensive rating to that of his teammates, but plus/minus and APM are the best to me, though it's hard to take it TOO seriously over one season, if you have a few years of data, it's pretty good evidence.

[nonemus]

A combination of apm and synergy

[parapooper]

Defensive win shares seems to give plausible results.
For instance last decade it was Duncan, Wallace, Duncan, Howard in the top spot and back in the 60s it was 10 years straight of Russell.
Defensive rating seems good too, but do they have it for the 50s and 60s?

[mysticbb]

Defensive win shares seems to give plausible results.

Defensive win shares is based on the individual boxscore-based DRtg. That makes it somewhat flawed regarding judging players who have the biggest contribution via non-boxscore stats related defensive elements like rotating well, taking charges forcing a offensive player into a bad shot, etc.
But still, it makes the best job regarding defense by all purely boxscore-based stats.

Defensive rating seems good too, but do they have it for the 50s and 60s?

ORtg and DRtg needs some informations which are not available for these years (before 77/78).

I agree with nonemus here. Synergy Sports is better than play-by-play extracted informations about the opponents player. We can see cleary how much a player is used on defense by those numbers and how well he defends certain areas. The defensive APM gives us a good impression about the effect a player has on defense. The combination of both tells us more about the individual defense.

[Ben]

Hope you guys don't mind me bumping this old thread, but I have some questions for the defense-stat crowd.

On the Bulls board we've had an ongoing debate about potential shooting guard targets when the labor issue is finally resolved. JR Smith (target via free agency) and OJ Mayo (target via trade) are two commonly discussed names. Both have defensive liabilities, but it seems to me that Smith gets maligned out of proportion to the progress that he's made in the last couple of years.

One person recently posted these numbers from Synergy. Caveat: I don't have a Synergy account and don't understand their measures terribly well.

JR SMITH Defense:

overall PPP 1.02 rank 434
iso 0.84 rank 170
P&R ball handler 0.85 rank 138
post up 1.04 rank 261
P&R roll man 1.36
spot up 1.28 rank 389 (!!!)
off screen 0.96 rank 123
hand off 0.78 rank 33

He drew the conclusion that Smith is terrible on defense.

I responded by admitting that I don't understand the Synergy stats all that well, and don't have access to them, but I posed the following questions. Never got a response from the person who posted the JR Smith stats. Perhaps some of you can help?

Where does Denver rank as a team in those stats? I think that they ranked very low. According to this blog post, written at the end of January, the Nuggets were very bad on a number of Synergy defensive metrics:

The Nuggets are horrid at defending the pick & roll – according to Synergy, they allow 0.87 points per possession defending the ballhandler (24th in the NBA) and a matador-like 1.10 points per possession against the roll man (28th in the NBA). They’re also poor at defending post-ups (.90 PPP, 23rd), spot-up shooters (1.01, 19th) and terrible on cutters (1.30, 26th).

Also, according to a different post that looks at Synergy's ranking of starting defensive wings, Kevin Martin is just as bad as JR Smith in defensive PPP, and Nic Batum is worse. Oh, and Monta Ellis is absolutely excellent. Here's a photo of the Synergy stats that the blog post linked:
http://www.depressedfan.com/img/ai9vsstarters022211.gif
You can make of that what you will.

Again, I don't have access to Synergy (or if there's free access available, I don't know how to get it, and if you could tell me I'd be grateful). But I'm curious to know where guys like Lawson and Gallinari ranked on those defensive categories. And I'd like to know how much worse JR Smith is than the rest of Denver's defenders, and/or whether his Synergy stats are affected by the kinds of defenses that Denver runs and the nature of their team play.
Also interested in knowing how OJ Mayo fares on those metrics, b/c last year he was really bad on statistical measures of defense.

Here's one of the things I'm trying to figure out:
If you look at the Nuggets' 2010-11 defensive ratings (DRtg), you'll see that JR Smith had the lowest (best) DRtg of any Nugget who played any kind of serious minutes except for Nene. Of course, we all know that DRtg is affected by team play and team defense. But if one explains away Smith's defensive rating by deferring it to the team's play, what role would one ascribe to the team's play in examining his Synergy defensive ratings?

I pose these questions to you guys b/c some of you were addressing the very things that I was mentioning: The player's DRtg compared with his teammates' DRtg and also adjusted +/-. Any insights into statistical analyses of Mayo's and Smith's defensive performances over the past couple of years?

[mysticbb]

I responded by admitting that I don't understand the Synergy stats all that well, and don't have access to them, but I posed the following questions. Never got a response from the person who posted the JR Smith stats. Perhaps some of you can help?

You can see those numbers as an expression of the individual 1on1 defensive ability of a certain player. JR Smith is indeed not good at that. But we also need to put those numbers into context here. Smith is pretty good at forcing turnovers, but he is bad at defending the 3pt shot, too often he gives his man the shot, by not aggressively closing out when he thinks he is too far away. That results into an awful 46.2 3P% against him. His individual defense against shots is bad, good defenders have their arm up and the hand in the face of the opponent, Smith has not. That is an issue which leads to such bad numbers for him.
BUT in the context of a wing defender he has also a lower defensive usage, resulting into a lower overall negative impact. We also see Smith increasing the offensive efficiency, which helps to get into a set defense earlier and is reducing the amount of fast break opportunities. But his overall bad shot defense nullifies any advantage here anyway.

For the last couple of seasons his defensive impact is around average (slightly below), he had seasons with a small positive impact and seasons with a small negative impact.

How does that look for Mayo? Well, Mayo is better at defending the shot (39.6 3P%), that makes him better as Smith regarding the PPP (0.96, ranked 367 overall), but he is hardly great in that area either. Mayo is forcing less turnovers (9.1 TO% vs. 13.2 TO% by Smith) and has an inefficent offensive games, which is an additional negative impact on defense. Overall Mayo effects the defense clearly in a negative way. For the 3 years he had in average a -2 DRAPM (Smith had -0.3), a clear difference in terms defensive impact here with the advantage for Smith despite the fact that Mayo defends the shots better.

On the offensive end it is no contest, Smith takes that easily, which makes him overall the better choice for the Bulls at SG. I also think that the flaw of Smith can be fixed easier on the defensive end, the forced turnovers are showing that Smith can play defense at a higher level, he just isn't doing it consistently. We also have to take into account that the defensive scheme of the Nuggets is much less sophisticated than that of the Grizzlies, which helps Mayo's numbers in 1on1 situations and will hurt Smith's numbers.

Any insights into statistical analyses of Mayo's and Smith's defensive performances over the past couple of years?

I hope it helped a bit.

[Ben]

Thanks a lot. Much appreciated.

[floppymoose]

A combination of apm and synergy

I'm not into synergy.

I like to look at +- and also on/off splits (for that player) of various team metrics, like defensive rebound percentage, turnovers, opponent efg, opponent PPP, etc.

[X-Factor]

Team Defensive Rating
Defensive +/-
Individual Defensive Rating from the WARP system
Defensive WARP, which is the wins above replacement player that can be credited to a player's defense based on rebounding, blocks, steals and personal fouls.
Each of the above rates for individual defensive statistics.
Net defensive plus-minus, via BasketballValue.com.
Statistics about individual defense from Synergy Sports, depending on the position: defense against pick-and-roll ballhandlers for point guards, isolation plays for wings and post-up defense for big men.

All combined.

[Fenris-77]

DRat based on PDSS data. It's the last word in defensive metrics.

[floppymoose]

Can you define PDSS data, and also provide a source for it? I hunted around and while I see some threads about it, I could find answers to neither of those questions.

[Wannabe MEP]

Can you define PDSS data, and also provide a source for it? I hunted around and while I see some threads about it, I could find answers to neither of those questions.

I did some sleuthing too...I'm pretty sure this is just Dean Oliver's defensive rating.

On the Raptors boards, there was a massive debate that ultimately pitted defensive rating (basketballvalue.com) vs. defensive rating (Oliver's method, which became nicknamed "PDSS" for incredibly stupid reasons). The Bargnani homers jumped on the latter as a counterargument to the former, because by any +/- related stat, Bargnani is epically terrible.

[Wannabe MEP]

I do believe this is the source of the nickname:
http://www.dukebasketballreport.com/for ... 463108886f

[EvanZ]

PDSS is something different from Oliver's defensive rating. His DRTG just uses box score stats. The PDSS was attempting to record more kinds of stats than are available in the box score (e.g. like field goals allowed), but as far as I know, that was never actually implemented in any way. It's certainly not available right now.

The closest that we can get to that idea is using Synergy or using ElGee's defensive rating, because he keeps track of some of the things Oliver proposed. ezPM uses counterpart data to calculate defensive stats, and while that's probably some improvement over DRTG, it's still far from perfect.

[Wannabe MEP]

PDSS is actually Oliver's language...ok, I was totally wrong.


So the defensive rating from PDSS, or Project Defensive Score Sheet, IS in fact Oliver's defensive rating, just on steroids. It involves a much more complex compilation of individual defensive stats: basically, you have to work incredibly hard to assign "credit" for who gets stops (and who doesn't).

But complex does not equal good. I'd take APM/RAPM--combined with healthy levels of unit data caveats--100% of the time. PDSS Drtg involves a huge subjective component because help defense impacts ALL plays, and help defense eliminates pure individual credit.

[EvanZ]

The point is nobody's actually compiling PDSS stats right now (maybe Oliver did when he worked for the Nuggets, but it wasn't made public obviously). The DRTG that basketball reference gives is a modified form of DO's DRTG, and just uses box score stats.

[Wannabe MEP]

Right, but you could compile it, at least for a team. An individual/team of individuals could track it for a team for a game/season/series/whatever. Been done. Just really hard work, and subjective.

What I'm saying though, is even if you had skilled stats teams tracking these stats for every NBA team with well-developed statistical checks to limit subjectivity . . . it still wouldn't be an improvement over APM/RAPM.

[EvanZ]

Right, but you could compile it, at least for a team. An individual/team of individuals could track it for a team for a game/season/series/whatever. Been done. Just really hard work, and subjective.

What I'm saying though, is even if you had skilled stats teams tracking these stats for every NBA team with well-developed statistical checks to limit subjectivity . . . it still wouldn't be an improvement over APM/RAPM.

RAPM is just a single number. It can't break down the strengths and weaknesses and specific skillsets of various players. You need that to build a team, evaluate trades, etc. RAPM might be the most useful stat for the fan, but for GM's and coaches, there's more to it than that single number. I definitely use RAPM, but I also use Synergy quite a bit to bring in that orthogonal data set.

[Wannabe MEP]

RAPM is just a single number.

Which is what the OP was asking for.