"Quality of Lift" in Relatively Analogous +/- Readouts

[D Nice]

So this is more theory, I haven't really seen anyone else but Doc mention it (and last time I did I think was in the Nash/Stockton thread a month or so back), but do you guys feel that a RAPM score of a guy lifting a stronger offensive/defensive team is more valuable than the lifting of a more pedestrian team, all else being equal? I guess the theoretical catchphrase for this right now is "scaling," but I haven't seen it quantitatively addressed much in the top 100 projects (the theoretical side might bear some similarity to WOWY but it's more simple than that).

Not sure if I'm being clear but here's an example.

The Toronto Raptors have an ORTG of 100 when Vince Carter is on the bench and his O-RAPM score for the season is 3.0.

The Boston Celtics have an ORTG of 104 when Paul Pierce is on the Bench and his O-RAPM score for the season is 3.0.

Did Paul Pierce have a higher quality offensive season than Vince Carter? Why or why not? Assume both players played all 82 games and averaged 36mpg with roughly similar box-score stats. The point here is to isolate "quality of lift" or a lack-there-of.

[Doctor MJ]

Yup, all things being equal, being able to apply X lift to a good team is a better thing than to a bad thing.

2 things:

1) In general in life, it's easier to make X progress form scratch than it is to improve a functional system by X.

2) It's possible that a guy can achieve X progress in a dysfunctional system in a manner which becomes far less useful once the rest of the people in the context are actually playing well together.

Now, all this is "all things being equal". There are plenty of intricacies to consider. But yeah, if I see a guy doing some kind of herculean effort where he takes all the shots inefficiently on a mediocre team, I think it's important to err on the side of "He's going to have to totally adapt how he plays if he wants to actually win" rather than "All his teammates suck. Give him a decent cast and he'll get rings."

[D Nice]

1) In general in life, it's easier to make X progress form scratch than it is to improve a functional system by X.

Yeah, this matches my experience in life. I was generally asking for affirmation/to see if I'm missing anything because you're really the only person I've seen mention it, and even then only in regards to KG on the Celtics and Nash vs. Stockton. I think it has a lot more top 100 implications than have been discussed, particularly with how much team lift and +/- is being referenced there.

Thanks for the response. Completely agree.

[colts18]

Lift only matters when it comes to Net +/- not RAPM. Thats why KG has the highest Net +/- ever recorded in 2003. RAPM doesn't care about that lift because it adjusts for the fact that your teammates suck offensively. A 3 RAPM is worth the same on a bad offense and good offense.

[D Nice]

Lift only matters when it comes to Net +/- not RAPM. Thats why KG has the highest Net +/- ever recorded in 2003. RAPM doesn't care about that lift because it adjusts for the fact that your teammates suck offensively. A 3 RAPM is worth the same on a bad offense and good offense.

What is the methodology for teammate quality adjustment? Is it just using Team ORTG/DRTG with [X] player off the court as a regression variable? Or a separate model entirely to use in Ridge Regression? I have no reason to believe it can what you claim it does accurately, and you yourself don't even seem to be aware of what the nuts and bolts actually are.

RAPM itself is a co-dependence metric, if it could properly model individual player "goodness" of the guys around the target variable (the player in question) it would directly apply said methodology to analyzing its target player.

[Chicago76]

Lift only matters when it comes to Net +/- not RAPM. Thats why KG has the highest Net +/- ever recorded in 2003. RAPM doesn't care about that lift because it adjusts for the fact that your teammates suck offensively. A 3 RAPM is worth the same on a bad offense and good offense.

In general, I agree with this, but there is still inevitably some "lift" when dealing with a good offensive player being added to an already good offensive team. Ex: a +6 offensive team probably won't become a +10 offensive team if you were to replace 67% of a position's floor minutes currently at a 0 with a +6 player (which is what the math would indicate).

There will be offensive redundancies that will scale down the contribution of each player. How much would depend upon what those redundancies might be given the roles of each player on the floor
. The game itself requires interaction between teammates. RAPM (in its various forms) attempts to turn something non-linear into a standard linear equation where we can simply add up the value of each player at both ends of the court and get a team approximation.