New defense metric by 538

Home for all your discussion of basketball statistical analysis.
Post Reply
Posts: 18
Joined: Tue Aug 07, 2018 11:15 am

New defense metric by 538

Post by Snedecor » Wed Jul 10, 2019 8:39 am


This is another gold article by 538 guy, the issue here is we can’t use in europe or minor league we don’t have access to second spectrum or another tracking system. While US and NBA are leading the world in all fields, europe are years ago.

Two questions about this:

-What do you think about article and metric. ... a-defense/

-Any ideas to use similar DRAYMOND with boxscore stats or easy way? How match the tracking metrics to europe without tracking system?


Posts: 22
Joined: Tue Oct 17, 2017 7:04 pm

Re: New defense metric by 538

Post by dtkavana » Wed Jul 10, 2019 1:56 pm

Really enjoyed their very thorough explanation of the metric ... ry-twitter

While it does prove itself over time as a stand-alone metric ("In any event, what we found is that BPM and DRAYMOND basically do equally well in predicting long-term RAPM. What that means is that the opponents’ shooting data is basically as powerful as all box score defensive statistics combined in predicting how much value a player’s defense truly has over the long run.") they still believe that the best use of DRAYMOND is in collaboration with RAPM and BPM, as that is the way they calculate CARMELO.

Very exciting stuff, and can't wait to get it incorporated into the defensive formulas on the CraftedNBA site.

Posts: 6188
Joined: Thu Apr 14, 2011 11:10 pm

Re: New defense metric by 538

Post by Crow » Wed Jul 10, 2019 3:07 pm

The Draymond metric shows how much impact value boxscore metrics are missing. All boxscore metrics could be amended to add Draymond or something of that ilk. Most won't and most won't even consider doing so. PIPM does something in its own way.

It is useful to have and compare with DRPM, DRAPM and factors of them. What are the averages of each for point guards, wings and bigmen? Those averages could be used to better evaluate the results or create role specified adjusted versions of them.

But this article doesn't compare DRAYMOND to the efg% factor of RPM or RAPM. That should be done. And there is still an argument for looking at these factors which try to capture off the ball team defense impacts. How well does the efg% factor of RPM or RAPM do compared to DRAYMOND at predicting long term DRAPM? Better or worse?

If you individually adjust direct efg% defense, you probably should look at play level not the shot defender efg% defensive data by itself too and see what you can learn.

Probably should consider similar exercises with the other factors.

I wouldn't look more at any one cut over the others. We should look at them all. Local, global. Raw, regularized, luck adjusted, split,,, Players in lineups splits, lineup up data. What combinations unleash more or less good or bad data? What is a player's estimate impact value relative to others if he was optimized to some average, good or max level? By each level of metric.

Post Reply