TDL Archived: The 2nd Rnd Pick Yankee Swap

[mvossman]

But the stat I am sharing IS unique to him. 



Look at the defensive on/off for the Mavs through 41 games. 

Of the guys with 500+ minutes, SD is the team bottom at -7.6. The next closest is DFS with -4.6. Then Luka with -3.9. And McGee in 214 minutes is only -2.8.

In On/Off numbers the far extreme on either end tends to be really telling and easier to interpret and potentially easier to address (the middle numbers are way messier and harder to sort through when looking at context). Luka is the extreme on offense with a +13.9 (not surprising). Maxi is the extreme on offense negatively -6.3. Green is the extreme on D with +12.8 (not surprising to me). SD is the extreme on defense negatively. 

Again, that does NOT mean Maxi is THE problem with the offense or SD is THE problem with the D. Not at all. But somehow their role within the context of the games is not currently working for the team on that end of the court. So the coaches and GM should take a long look at it.

So my problem with the on/off numbers is that they introduce even more noise into an already noisy stat (single player net rating).  Din has similar D rating to the rest of the starting lineup, so why are his on/off numbers worse?  Is it because he has played less percentage of his minutes with Green (who currently is the defensive on/off God on this team) than the other guys?  I struggle with a stat that is just as influenced by time when the player is not even on the court.

That being said, I do see that his defensive on/off looks to be an outlier, and if this is what you are looking at it is not misleading to show just Dinwiddie's numbers in this context.

I will say that if we are looking at those numbers above without context, Powell has a significantly higher impact on both sides of the ball than Wood, and using just that information it would be a strong argument to start Powell and give him more minutes.  I know you don't advocate for that and neither do I, but my point is that context really needs to be looked at when we look at these kind of stats.