The DRAYMOND defensive ranking metric by 538.com
Goto page 1, 2  Next

 
Post new topic    LakersGround.net Forum Index -> LA Lakers Lounge Reply to topic
View previous topic :: View next topic  
Author Message
Annihilator
Star Player
Star Player


Joined: 02 Jul 2001
Posts: 3222

PostPosted: Fri Jul 12, 2019 5:05 am    Post subject: The DRAYMOND defensive ranking metric by 538.com

fivethirtyeight.com just came out with a new defensive ranking metric that they call the DRAYMOND. It essentially tries to measure how many points the players keeps the opponent from scoring based per 100 shots. A player who is a bad 3-pt defender is especially hurt by these ratings. I will leave the analysis of how good of a metric it is to the article and those who like to discuss such things.

In the article is a list of the players with 2000 possessions. The data and scores for each player in the league since the 2013-14 season can be found here. Those of you who wished that the Lakers had drafted Mitchell Robinson or not traded Zubac will not be happy.

The 2018-19 scores for the 2019-20 Lakers were:

Cousins 2.18 (just missed the 2000 possession limit)
McGee 2.16

Davis 1.15
Kuzma -0.25

James 0.40
Dudley 1.40

Green 0.08
KCP -2.49 (fifth worst in league of those with 2000 possessions)
Daniels -2.57

Rondo -1.79
Bradley 0.41
Caruso 4.17 (but didn't have 2000 possessions)
Cook -2.11
_________________
Get your facts first, and then you can distort them as much as you please. --Mark Twain

Scientists say that the universe is full of protons, neutrons and electrons. They forgot to mention morons.--Unknown
Back to top
View user's profile Send private message Reply with quote
LandsbergerRules
Star Player
Star Player


Joined: 29 Aug 2004
Posts: 9351
Location: The Other Perspective

PostPosted: Fri Jul 12, 2019 5:12 am    Post subject:

Is this one of those weird things like the CARMELO metric that states Kemba is much more valuable than Kawhi?
_________________
"Chick lived and breathed Lakers basketball…but he was also fair and objective and called every game the way it was played."
-from Chick: His Unpublished Memoirs and the Memories of Those Who Knew Him
Back to top
View user's profile Send private message Reply with quote
Annihilator
Star Player
Star Player


Joined: 02 Jul 2001
Posts: 3222

PostPosted: Fri Jul 12, 2019 5:21 am    Post subject:

LandsbergerRules wrote:
Is this one of those weird things like the CARMELO metric that states Kemba is much more valuable than Kawhi?

All metrics on basketball players of this type are weird in that they try to project a situation that is multidimensional onto a single number. In any such system there are usually outliers that show that the system doesn't work perfectly. A book that shows how this fails when evaluating results in business, education, etc is Weapons of Math Destruction (link). Having said that, these metrics do provide **some** useful information--but they absolutely need to be carefully scrutinized when deciding on such things as performance reviews, hiring, firing, etc.
_________________
Get your facts first, and then you can distort them as much as you please. --Mark Twain

Scientists say that the universe is full of protons, neutrons and electrons. They forgot to mention morons.--Unknown


Last edited by Annihilator on Fri Jul 12, 2019 5:25 am; edited 1 time in total
Back to top
View user's profile Send private message Reply with quote
ringfinger
Retired Number
Retired Number


Joined: 08 Oct 2013
Posts: 26363

PostPosted: Fri Jul 12, 2019 5:25 am    Post subject:

Definitely a weird stat.

If i close out on a defender and force him to pass up the shot, or strip the ball from him, or deflect the ball in the passing lane, it doesnt seem like this metric gives credit to that. (And I would argue that is good defense).

You have to have allowed the offensive player to get the shot off it seems?
Back to top
View user's profile Send private message Reply with quote
Annihilator
Star Player
Star Player


Joined: 02 Jul 2001
Posts: 3222

PostPosted: Fri Jul 12, 2019 5:34 am    Post subject:

ringfinger wrote:
Definitely a weird stat.

If i close out on a defender and force him to pass up the shot, or strip the ball from him, or deflect the ball in the passing lane, it doesnt seem like this metric gives credit to that. (And I would argue that is good defense).

You have to have allowed the offensive player to get the shot off it seems?

Good point.

Whenever someone tries to project a multidimensional problem onto a single number, information is either lost or ignored.

A quote from the article:
Quote:
But there are others like Trevor Ariza, who gets lots of steals but has been rated poorly by DRAYMOND in recent seasons. That doesn’t mean that Ariza is a poor defender, just that you need to take the good (steals) with the bad (allows opponents to convert field goals at a high rate) when evaluating him.

_________________
Get your facts first, and then you can distort them as much as you please. --Mark Twain

Scientists say that the universe is full of protons, neutrons and electrons. They forgot to mention morons.--Unknown
Back to top
View user's profile Send private message Reply with quote
Aeneas Hunter
Franchise Player
Franchise Player


Joined: 12 Jul 2005
Posts: 22910

PostPosted: Fri Jul 12, 2019 5:45 am    Post subject:

There's already a thread about this in General.

This is another stat that validates itself by comparing itself to RAPM (now RPM). Okay, so DRAYMOND correlates with DRPM. Given that DRPM is readily available, what's the point?
Back to top
View user's profile Send private message Reply with quote
epak
Retired Number
Retired Number


Joined: 30 Aug 2005
Posts: 31472

PostPosted: Fri Jul 12, 2019 8:13 am    Post subject:

Aeneas Hunter wrote:
There's already a thread about this in General.

This is another stat that validates itself by comparing itself to RAPM (now RPM). Okay, so DRAYMOND correlates with DRPM. Given that DRPM is readily available, what's the point?



Did you get a chance to read the article? I dont think it's about DRPM.

Quote:
the idea of minimizing openness. The main goal of shooting defense, especially in today’s spacing-centric, ball-movement-forward offensive era, is really to minimize the chance of an open shot.


It's more about guarding possessions more than points per possession no?

I think it's a very interesting way to look at defense. I recall Pop talking about contests and the value of that. And generally how shooting %s go down when contested. Now we know that all contests are not the same, but still it's very interesting. There will definitely be edge cases and exceptions.
_________________
👊 Los Angeles Lakers 👊

For us anything short of a championship is not success.
Back to top
View user's profile Send private message Reply with quote
ribeye
Franchise Player
Franchise Player


Joined: 10 Nov 2001
Posts: 10312

PostPosted: Fri Jul 12, 2019 8:43 am    Post subject:

Aeneas Hunter wrote:
There's already a thread about this in General.

This is another stat that validates itself by comparing itself to RAPM (now RPM). Okay, so DRAYMOND correlates with DRPM. Given that DRPM is readily available, what's the point?


The point, as was explained, is that dRPM is noisy -- meaning that it fluctuates greatly until there is a large sample size. As such, where RPM has great validity over a span of several seasons, it is less so for a single season. DRAYMOND seems to be a decent single season metric.
_________________
If ever there were a Mount Rushmore of liars, it would be Trump, Trump, Trump, and Trump.
Back to top
View user's profile Send private message Reply with quote
Sentient Meat
Franchise Player
Franchise Player


Joined: 04 Jul 2014
Posts: 12978

PostPosted: Fri Jul 12, 2019 8:56 am    Post subject:

Appreciate the link.

Interesting reading, but like all advanced stats... seem to be very vulnerable to biased interpretation and abuse.

A lot of random people that no one would sign at the top of the list.
Back to top
View user's profile Send private message Reply with quote
Annihilator
Star Player
Star Player


Joined: 02 Jul 2001
Posts: 3222

PostPosted: Fri Jul 12, 2019 9:38 am    Post subject:

ribeye wrote:
Aeneas Hunter wrote:
There's already a thread about this in General.

This is another stat that validates itself by comparing itself to RAPM (now RPM). Okay, so DRAYMOND correlates with DRPM. Given that DRPM is readily available, what's the point?


The point, as was explained, is that dRPM is noisy -- meaning that it fluctuates greatly until there is a large sample size. As such, where RPM has great validity over a span of several seasons, it is less so for a single season. DRAYMOND seems to be a decent single season metric.

First, I apologize for starting a thread for which one already exists. My only excuse is that I rarely ever go into the the General Forum. I checked this forum to make sure nothing was here. This thread, however, has taken a slightly different turn in that we are discussing how this metric is good or bad.

Second, ESPN has never published the methodology for how DRPM is computed. From this link,
Quote:
[The] metric isolates the unique plus-minus impact of each NBA player by adjusting for the effects of each teammate, opposing player and coach. ... The RPM model sifts through more than 230,000 possessions each NBA season to tease apart the "real" plus-minus effects attributable to each player, employing techniques similar to those used by scientific researchers when they need to model the effects of numerous variables at the same time.
Maybe they are using some variation of partial correlation. Who knows? Oh well, unlike the DRAYMOND method, which allows us to go in and see its strengths and weaknesses, DRPM is essentially a black box that requires that we trust whatever number comes out. Overall, though, its stated goal is to measure the same thing that DRAYMOND does--how much does a player being on the court affect how many points are scored against the team.

Third, I was hoping that people would actually compare the two methods to see if the scores told a story. For example, KCP has a -2.17 DRPM and a -2.49 DRAYMOND. If both metrics tell a reasonable story--which we really can't tell because we don't know the DRPM methodology--most of KCP's defensive problems are that he doesn't contest shots very well. On the other hand, Caruso has a 0.77 DRPM and a 4.17 DRAYMOND. Again, if both metrics are telling a reasonable story, Caruso contests shots really well but has some other deficiencies on which he needs to work.
_________________
Get your facts first, and then you can distort them as much as you please. --Mark Twain

Scientists say that the universe is full of protons, neutrons and electrons. They forgot to mention morons.--Unknown
Back to top
View user's profile Send private message Reply with quote
epak
Retired Number
Retired Number


Joined: 30 Aug 2005
Posts: 31472

PostPosted: Fri Jul 12, 2019 9:45 am    Post subject:

I just go w/ the "all contests are not the same" thought.
A 6'0" wingspan guy contesting a shot vs a 6'5" wingspan guy contesting a shot would have different effects imo.

I like the combination of both "contesting" + "fg% against" to see how good a defender someone is.
_________________
👊 Los Angeles Lakers 👊

For us anything short of a championship is not success.
Back to top
View user's profile Send private message Reply with quote
Aeneas Hunter
Franchise Player
Franchise Player


Joined: 12 Jul 2005
Posts: 22910

PostPosted: Fri Jul 12, 2019 10:16 am    Post subject:

epak wrote:
Aeneas Hunter wrote:
There's already a thread about this in General.

This is another stat that validates itself by comparing itself to RAPM (now RPM). Okay, so DRAYMOND correlates with DRPM. Given that DRPM is readily available, what's the point?



Did you get a chance to read the article? I dont think it's about DRPM.


The article uses RPM (which is the same as RAPM) to validate DRAYMOND. Here is what it says:

Quote:
But does DRAYMOND measure something that’s actually meaningful? To test this — and I have to apologize because there are a lot of similarly named statistics here with confusing acronyms — we regressed DRAYMOND and defensive BPM5 against five-year defensive Regularized Adjusted Plus-Minus (RAPM). What the hell does that mean? BPM is based on conventional box score statistics — most importantly rebounds, blocks and steals for the purposes of measuring defense. DRAYMOND is based on opponents’ shooting. RAPM, meanwhile, measures how much better or worse a team plays when a player is on or off the floor. In the long run, RAPM is basically6 the “right” way to measure player value, since it can account for all the direct and indirect contributions a player makes that may or may not have a corresponding statistic attached to them. In the short run and even the medium run, however (remember all that misleading data people cited about how the Warriors were so good without Kevin Durant?), RAPM can be extremely noisy. So RAPM is great if you’re looking back over a five-year sample, as we’re doing here, but on-court/off-court statistics need to be treated with extreme caution over small samples.


The stuff about short term effects is certainly true. You need a decent sample size for RPM to be meaningful, which is why ESPN typically does not release it until about 20 games into the season. However, I don't buy the argument that RPM is only useful over a five-year sample. In fact, I think that's a load of nonsense. It's one thing to say that RPM generates noise over a small sample of games (the Durant example he uses), but that will be true for any statistical measure, including DRAYMOND.
Back to top
View user's profile Send private message Reply with quote
ribeye
Franchise Player
Franchise Player


Joined: 10 Nov 2001
Posts: 10312

PostPosted: Fri Jul 12, 2019 10:35 am    Post subject:

Annihilator wrote:

Second, ESPN has never published the methodology for how DRPM is computed. From this link,
Quote:
[The] metric isolates the unique plus-minus impact of each NBA player by adjusting for the effects of each teammate, opposing player and coach. ... The RPM model sifts through more than 230,000 possessions each NBA season to tease apart the "real" plus-minus effects attributable to each player, employing techniques similar to those used by scientific researchers when they need to model the effects of numerous variables at the same time.
Maybe they are using some variation of partial correlation. Who knows? Oh well, unlike the DRAYMOND method, which allows us to go in and see its strengths and weaknesses, DRPM is essentially a black box that requires that we trust whatever number comes out. Overall, though, its stated goal is to measure the same thing that DRAYMOND does--how much does a player being on the court affect how many points are scored against the team.


What is now RPM is some evolution of RAPM, both of which are the products of Jerry Engleman (was or is an analyst for the Mavs), who, at least, popularized the metric. There are explanations of RAPM that you can find by using "the Google machine", but unless one has a solid understanding of statistics regressions, which I don't, it is like trying to figure out Trump's appeal.

Year's ago, I noticed that Pau and Kobe played together in all but two or three of the many five man groups, as shown by 82 games, and these had very few minutes of play. So, for some 95+% of the time, they played in the same five man groups. But Pau had a much better defensive rating in RAPM. I emailed Jerry and he responded that one of factors he used was a player's height--since it almost always correlated to a greater defensive efficiency. I don't know if he uses height as a factor in his RPM. I suspect Jerry is a bit protective of his RPM formulas.
_________________
If ever there were a Mount Rushmore of liars, it would be Trump, Trump, Trump, and Trump.
Back to top
View user's profile Send private message Reply with quote
Aeneas Hunter
Franchise Player
Franchise Player


Joined: 12 Jul 2005
Posts: 22910

PostPosted: Fri Jul 12, 2019 10:36 am    Post subject:

Annihilator wrote:
Second, ESPN has never published the methodology for how DRPM is computed. From this link,
Quote:
[The] metric isolates the unique plus-minus impact of each NBA player by adjusting for the effects of each teammate, opposing player and coach. ... The RPM model sifts through more than 230,000 possessions each NBA season to tease apart the "real" plus-minus effects attributable to each player, employing techniques similar to those used by scientific researchers when they need to model the effects of numerous variables at the same time.
Maybe they are using some variation of partial correlation. Who knows? Oh well, unlike the DRAYMOND method, which allows us to go in and see its strengths and weaknesses, DRPM is essentially a black box that requires that we trust whatever number comes out. Overall, though, its stated goal is to measure the same thing that DRAYMOND does--how much does a player being on the court affect how many points are scored against the team.

Third, I was hoping that people would actually compare the two methods to see if the scores told a story. For example, KCP has a -2.17 DRPM and a -2.49 DRAYMOND. If both metrics tell a reasonable story--which we really can't tell because we don't know the DRPM methodology--most of KCP's defensive problems are that he doesn't contest shots very well. On the other hand, Caruso has a 0.77 DRPM and a 4.17 DRAYMOND. Again, if both metrics are telling a reasonable story, Caruso contests shots really well but has some other deficiencies on which he needs to work.


On the RPM methodology: It's a regression based computer calculation, so there isn't going to be a nice formula like you would find for PER and the like. I've looked at some of the technical literature on the subject, but my math and statistical skills are not sufficiently advanced. When they start talking about things like ridge regression, my eyes glaze over. Here is a guy who attempts to explain it:

https://fansided.com/2014/04/21/calculating-rapm/

Here is a more technical piece:

https://squared2020.com/2017/09/18/deep-dive-on-regularized-adjusted-plus-minus-i-introductory-example/

On your second point, you should always look at multiple metrics. If you have watched any of the Thinking Basketball videos I have been linking over in General, you will see that Ben Taylor (a geek) looks at multiple "single number" metrics. There is no magical metric that answers all questions and that cannot be challenged.

My reaction to DRAYMOND, once again, is that I don't see the point of a stat that is validated by its correlation to RPM when we already have RPM. DRAYMOND strikes me as more of a specialized stat that highlights the importance of one particular aspect of defense.
Back to top
View user's profile Send private message Reply with quote
Annihilator
Star Player
Star Player


Joined: 02 Jul 2001
Posts: 3222

PostPosted: Fri Jul 12, 2019 11:05 am    Post subject:

Aeneas Hunter wrote:
On your second point, you should always look at multiple metrics. If you have watched any of the Thinking Basketball videos I have been linking over in General, you will see that Ben Taylor (a geek) looks at multiple "single number" metrics. There is no magical metric that answers all questions and that cannot be challenged.

That was my point--so I am agreeing with you on the need to use multiple metrics.

OK. I will continue down the topics laid out here. Which are the most valuable defensive metrics (in your opinion) and why?
_________________
Get your facts first, and then you can distort them as much as you please. --Mark Twain

Scientists say that the universe is full of protons, neutrons and electrons. They forgot to mention morons.--Unknown
Back to top
View user's profile Send private message Reply with quote
Car54
Franchise Player
Franchise Player


Joined: 10 Aug 2005
Posts: 12443

PostPosted: Fri Jul 12, 2019 11:22 am    Post subject:

LandsbergerRules wrote:
Is this one of those weird things like the CARMELO metric that states Kemba is much more valuable than Kawhi?


I think so lol@ us worried about not drafting Robinson. I could have sworn we traded for AD lol
_________________
Coach Vogel, Kidd, Hollins
Max slot : Kawhi
Back to top
View user's profile Send private message Reply with quote
ribeye
Franchise Player
Franchise Player


Joined: 10 Nov 2001
Posts: 10312

PostPosted: Fri Jul 12, 2019 11:30 am    Post subject:

LandsbergerRules wrote:
Is this one of those weird things like the CARMELO metric that states Kemba is much more valuable than Kawhi?


Yes, but who knew what to expect from Kawhi after his injury and layoff.

I would imagine that CARMELO will see this differently next year.
_________________
If ever there were a Mount Rushmore of liars, it would be Trump, Trump, Trump, and Trump.
Back to top
View user's profile Send private message Reply with quote
Aeneas Hunter
Franchise Player
Franchise Player


Joined: 12 Jul 2005
Posts: 22910

PostPosted: Fri Jul 12, 2019 11:38 am    Post subject:

Annihilator wrote:
OK. I will continue down the topics laid out here. Which are the most valuable defensive metrics (in your opinion) and why?


DRPM is the best single number metric in my opinion, and in any event it is the most accessible because it is on ESPN.com. DBPM isn't really a plus/minus stat, but it is decent. I have heard of some other versions, but they are hard to find or are behind pay walls. I've been hearing about some stat called luck-adjusted RAPM, or something like that.

But no single number stat is ever going to capture everything, and they will all have their quirks. For example, DRPM is no good for comparing players who have different roles. A big man who plays in the paint will almost always have a higher DRPM than a guard. For example, last season the highest rated defensive guard was Chris Paul, who was 35th in overall DRPM. This is probably accurate -- big men have a much greater defensive impact than guards. But you need to keep this in mind when people throw around DRPM stats.
Back to top
View user's profile Send private message Reply with quote
ribeye
Franchise Player
Franchise Player


Joined: 10 Nov 2001
Posts: 10312

PostPosted: Fri Jul 12, 2019 11:57 am    Post subject:

Annihilator wrote:

OK. I will continue down the topics laid out here. Which are the most valuable defensive metrics (in your opinion) and why?


In addition to dRPM, I like the Opponent Production stat from 82games. (I also like the Own Production stat and the on/off as well.) The Own/Opponent productions are similar to PER but the Opponent Production is obviously from a defensive standpoint. Since it is proprietary data, I don't don't the specifics beyond. Also, it does not update often, sometimes not until the season has ended.
_________________
If ever there were a Mount Rushmore of liars, it would be Trump, Trump, Trump, and Trump.
Back to top
View user's profile Send private message Reply with quote
tox
Franchise Player
Franchise Player


Joined: 16 Nov 2015
Posts: 15139

PostPosted: Fri Jul 12, 2019 12:44 pm    Post subject:

Hold on a second. RAPM is not the same thing as RPM. RPM is an "evolution" over single-season RAPM. As the 538 article itself states, RAPM over a long period of time (5+ years) is generally the gold standard of NBA stats. However, the statistical model is prone to overfitting on shorter timespan with less data.

The idea of RPM was to incorporate some prior knowledge (including player height, box score stats, etc.) to help anchor (single-season) RAPM and make its projections more accurate. If I recall the 538 article correctly, it said that DRAYMOND correlates with the 5-year RAPM sample as well as BPM.

There is a sleight-of-hand going on here: DBPM itself is a pretty bad stat (as anyone looking at Westbrook's stats has noticed). Its correlation with single season RAPM was, iirc, 0.5. It's not terrible but it's not great. So DRAYMOND is also not great.

538 hints at this, but this stat is not actually a single-measure defensive contribution score. I like to see this stat as a "hustle/ activity monitor" for perimeter players and "rim protection monitor" for big men near the basket.
Back to top
View user's profile Send private message Reply with quote
tox
Franchise Player
Franchise Player


Joined: 16 Nov 2015
Posts: 15139

PostPosted: Fri Jul 12, 2019 12:47 pm    Post subject:

Also, give Caruso some real minutes. Would love to see his defensive numbers against real NBA players, not during the garbage portion of the season
Back to top
View user's profile Send private message Reply with quote
Aeneas Hunter
Franchise Player
Franchise Player


Joined: 12 Jul 2005
Posts: 22910

PostPosted: Fri Jul 12, 2019 3:11 pm    Post subject:

tox wrote:
Hold on a second. RAPM is not the same thing as RPM. RPM is an "evolution" over single-season RAPM. As the 538 article itself states, RAPM over a long period of time (5+ years) is generally the gold standard of NBA stats. However, the statistical model is prone to overfitting on shorter timespan with less data.


Well, sure, RPM is not the same thing as five-year RAPM. Any metric will benefit from more data, and this is especially true for a regression based stat. But five-year RAPM has limited utility for fan discussions because we’re interested in what a player is doing now, or is expected to do next year, not what a player has done over the past five seasons. The average career length of an NBA player is less than five years. That’s the point of RPM.
Back to top
View user's profile Send private message Reply with quote
The Juggernaut
Star Player
Star Player


Joined: 24 Aug 2017
Posts: 4572

PostPosted: Fri Jul 12, 2019 3:16 pm    Post subject:

Defense is the most complex and hardest area of basketball to judge with stats. This should just be used as a tool to add into other evaluations like film breakdowns and the defensive schemes coaches want players to adhere to.
Back to top
View user's profile Send private message Reply with quote
Dr. Laker
Franchise Player
Franchise Player


Joined: 12 Apr 2002
Posts: 12782

PostPosted: Sat Jul 13, 2019 5:37 am    Post subject:

Give me all of your lunch money, you Geeks.
_________________
There was NO COLLUSION with Russia. Ukraine, Australia, Italy, Britain and others are a different story . . .
Back to top
View user's profile Send private message Reply with quote
LandsbergerRules
Star Player
Star Player


Joined: 29 Aug 2004
Posts: 9351
Location: The Other Perspective

PostPosted: Sat Jul 13, 2019 5:58 am    Post subject:

tox wrote:
Also, give Caruso some real minutes. Would love to see his defensive numbers against real NBA players, not during the garbage portion of the season



_________________
"Chick lived and breathed Lakers basketball…but he was also fair and objective and called every game the way it was played."
-from Chick: His Unpublished Memoirs and the Memories of Those Who Knew Him
Back to top
View user's profile Send private message Reply with quote
Display posts from previous:   
Post new topic    LakersGround.net Forum Index -> LA Lakers Lounge All times are GMT - 8 Hours
Goto page 1, 2  Next
Page 1 of 2
Jump to:  

 
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum






Graphics by uberzev
© 1995-2018 LakersGround.net. All Rights Reserved. Privacy Policy. Terms of Use.
LakersGround is an unofficial news source serving the fan community since 1995.
We are in no way associated with the Los Angeles Lakers or the National Basketball Association.


Powered by phpBB