DRAYMOND, or 538's newest defensive metric

Sam Ray Not

Member
SoSH Member
Jul 19, 2005
8,848
NYC
https://fivethirtyeight.com/features/a-better-way-to-evaluate-nba-defense/
Unsurprisingly, the highest-rated defender over the past five seasons according to DRAYMOND is Draymond Green. Last season it was Derrick Favors.

Lowest-rated defender over the last five seasons according to DRAYMOND: Rajon Rondo. Lowest-rated last season: Collin Sexton (followed somewhat surprisingly by Trevor Ariza).

Haven't really had time to ponder their methodology — I'm generally somewhat skeptical of Nate Silver, and of defensive metrics that are based on things other than plus-minus over the largest possible samples (ideally regressed for quality of teammate but not for boxscore stats). But of course I love the name, and am always curious to hear what the MBPC crew think.
 

PedroKsBambino

Well-Known Member
Lifetime Member
SoSH Member
Apr 17, 2003
31,191
The one thing that jumped out at me was how many big guys were near the top. That may be because it is heavily focused on impact on sh%, which is relevant but imperfect tool to look at.

I suspect it will be helpful as an addendum to RPM or other plus minus-related metrics. Those have the gap that you can't actually correct for matchups....which this helps with, though this would appear to have other flaws.
 

Captaincoop

Member
SoSH Member
Jul 16, 2005
13,487
Santa Monica, CA
Trying to create a single metric to measure basketball defense is a fool's errand. There are so many variables involved, many of which are impossible to account for in a system like this.

All it does is provide people who don't really understand basketball another blunt instrument to misuse in arguments.
 

RG33

Certain Class of Poster
SoSH Member
Nov 28, 2005
7,199
CA
[QUOTE="Captaincoop, post: 3471949, member: / . . . another blunt instrument to misuse in arguments.
[/QUOTE]
Another reason it is perfectly named!
 

Devizier

Member
SoSH Member
Jul 3, 2000
19,467
Somewhere
The main issue is stuff like: if a big is contesting a layup then your scoring chances are much lower than if a guard is doing the same. But the fact that the offensive team is even attempting a layup is the problem in the first place. So yes, context matters a ton. You could take this metric one step further and have scoring expectations for each shot but you’re not going to be able to do that with play by play recaps alone. I’m sure teams have proprietary versions of this already.
 
Last edited:

Cellar-Door

Member
SoSH Member
Aug 1, 2006
34,460
To me the biggest problem is that it appears to be eliminating plays where you are out of position, eliminating when you are the closest defender but not actually close sounds good in theory, but in practice you're usually the closest defender because you fucked up. So like Whiteside gets rewarded for challenging lots of shots but not punished for all the easy looks at the basket teams get by taking advantage of his utter inability to follow the system.
 

DJnVa

Dorito Dawg
SoSH Member
Dec 16, 2010
53,840
I like it because it shows Jaylen is better than Kawhi.
 

NomarsFool

Member
SoSH Member
Dec 21, 2001
8,157
Interesting that Marcus Smart is rated only average. Of course, it doesn't even attempt to account for things like drawing charges, steals, deflected passes, etc. It also doesn't do anything for when an offensive player chooses to pass the ball away rather than attempt a shot.
 

Gdiguy

Member
SoSH Member
Jul 15, 2005
6,231
San Diego, CA
I find it odd that they didn't really release / discuss a version that adds both DRAYMOND and RPM, because (as they say) they seem to be more complementary rather than overlapping

Like, the analysis is useful to identify 'who's the best at things that current methods ignore', but they seem to then avoid the more interesting final conclusion of 'ok, so who's the best overall defender then?'
 

BaseballJones

ivanvamp
SoSH Member
Oct 1, 2015
24,375
The main issue is stuff like: if a big is contesting a layup then your scoring chances are much lower than if a guard is doing the same. But the fact that the offensive team is even attempting a layup is the problem in the first place. So yes, context matters a ton. You could take this metric one step further and have scoring expectations for each shot but you’re not going to be able to do that with play by play recaps alone. I’m sure teams have proprietary versions of this already.
And what if you have 3 fouls in the first half and your coach tells you to NOT pick up your fourth, and a guy takes it hard and you basically let him go in order to not pick up that extra foul? Or if you're guarding a bad shooter and they tell you to lay off him because the coach WANTS that bad shooter to take open shots (Hey guys, I'm OPEN! Uh, yeah, you're open for a REASON, you dummy!), but he happens to knock a few down on you.

That stuff actually happens in real life basketball, but I bet the defensive metrics all penalize defenders for things like this. After all, they don't "know" what the coach told his player to do.
 

lexrageorge

Member
SoSH Member
Jul 31, 2007
18,096
538's history with sports statistics is mixed. Just recall their so-called analysis of the Patriots fumbling tendencies (or lack thereof) during Deflategate as Exhibit A.
 

Jimbodandy

Member
SoSH Member
Jan 31, 2006
11,403
around the way
And what if you have 3 fouls in the first half and your coach tells you to NOT pick up your fourth, and a guy takes it hard and you basically let him go in order to not pick up that extra foul? Or if you're guarding a bad shooter and they tell you to lay off him because the coach WANTS that bad shooter to take open shots (Hey guys, I'm OPEN! Uh, yeah, you're open for a REASON, you dummy!), but he happens to knock a few down on you.

That stuff actually happens in real life basketball, but I bet the defensive metrics all penalize defenders for things like this. After all, they don't "know" what the coach told his player to do.
Of course, but shit like that normalizes over time. Why wouldn't it?
 

BaseballJones

ivanvamp
SoSH Member
Oct 1, 2015
24,375
Of course, but shit like that normalizes over time. Why wouldn't it?
Coaches have different philosophies. Maybe on one team they have enough depth that they don't mind guys getting into foul trouble, but on another team, they can't afford that, so they give their players different instructions. Maybe some teams really want to challenge EVERY shooter while on others they want to let certain guys shoot. We know that different philosophies exist. So I don't see why it necessarily should "normalize" for everyone. We're not talking about rolling fair dice here; we're talking about humans who have philosophies and perspectives and who get orders from coaches (who in turn get orders from front offices) and there's just more factors involved. So I suppose it could normalize, but I suspect what's "normal" is that we should actually see differences, not uniformity over time.
 

wade boggs chicken dinner

Member
SoSH Member
Mar 26, 2005
30,491
I find it odd that they didn't really release / discuss a version that adds both DRAYMOND and RPM, because (as they say) they seem to be more complementary rather than overlapping

Like, the analysis is useful to identify 'who's the best at things that current methods ignore', but they seem to then avoid the more interesting final conclusion of 'ok, so who's the best overall defender then?'
Doesn't the last chart more or less do this? They've taken their old CARMELO defensive ratings, which 2/3rds RPM + 1/3rd BPM and added the new shooting metric to get new defensive ratings.

Note that by this, the players that are most underrated by the old CARMELO defensive ratings are: Porzingis, Embiid, Bjelica, Josh Smith, Brook Lopez, Mozgov, AD, and Kemba.

Nice to see Kemba is way ahead of KI in this metric. :)
 

wade boggs chicken dinner

Member
SoSH Member
Mar 26, 2005
30,491
Of course, but shit like that normalizes over time. Why wouldn't it?
Question is how many games will it take to normalize it. Like if a coach is happy letting Rajan Rondo take 3Ps, and he happens to make 4-5 one game,

The article says that the average guard contests 15 shots per 100 possessions, which is about 12 shots a game. Seems like a couple of fluky games could really affect the season's statistics.
 

Cellar-Door

Member
SoSH Member
Aug 1, 2006
34,460
wait... did they not put it in the article, or did they really devise a system on the assumption that an open layup is only 8 percentage points more likely to be made than one against average defense?
 

Mooch

Member
SoSH Member
Jul 15, 2005
4,494
Any metric that tells me that Terry Rozier was a better defender than Marcus Smart last year deserves at least a small dollop of skepticism. Sorry, since 2013-2014.
 

bowiac

Caveat: I know nothing about what I speak
Lifetime Member
SoSH Member
Dec 18, 2003
12,945
New York, NY
I'm a little confused how (or whether) they're dealing with the "double counting" issue here.

Lets assume this defensive shooting data is meaningful (which it probably is in some regressed form). The impact of this shooting defense is already built into RPM (since it impacts your plus/minus), and is likewise already implicitly built into your BPM (since it impacts your team's net offense/defense). Ignoring the fact that the mix of RPM and BPM is already arbitrary and nonsense in the first place (since RPM already includes box-score data like BPM!, and the 1/3 weighting is an arbitrary eyeballed number it seems), I think you're going to end up overrating players with shooting defense through this metric since you're double counting the impact of that shooting defense (it first goes into RPM, and then goes into DRAYMOND as well).

What they should be doing here is layering in this defensive shooting data to predict stints in the first place, and essentially creating their own version of RPM. But it sounds like they haven't done that here. They haven't even gone so far as to calculate their own version of RAPM - they're just using Ryan Davis's data for that.

I may be misreading what they're doing here of course.
 

jon abbey

Shanghai Warrior
Moderator
SoSH Member
Jul 15, 2005
70,713
Always very dubious about 538 and anything sports-related, but any stat that has my man Mitchell Robinson #2 in the NBA I am a fan of.