pancakes3 wrote:3 - APM is not only variable through lineups, but also matchups and teammates. in a league where PFs are beastly, and KG playing on crap teams, his APM would be significantly greater than say... Amare's simply because Amare could have either marion and nash on the court when he took a breather due to foul trouble or whatnot. you take KG out and you put in Rasho/Kandiman and you've got +/- problems out the wazoo. So yes, if you want to show consistency in the APM throughout the various TWolves teams... fine. you can draw the conclusion that KG was the best player on those teams, and that his teammates were consistently crap. however, i don't know if you can use that as a way to figure out "intangibles".
A few things.
1) You are oversimplifying APM to the point that you aren't even accurate. Raw +/- is, in fact, heavily team dependent. But the whole purpose of APM is to ADJUST for the factors that you name. In the example that you name, APM would be looking at how Amare performs with/without Marion and with/without Nash, as well as how Nash and Marion perform with and without Amare. So no, it's not an example where good teammates mean's lower +/-. That's incorrect.
2) The biggest problems with the "adjustments" I mentioned are noise and collinearity. Non-technical terms aside, that's saying that you have to have a lot of data from many different combinations of players for the "adjustments" to be solved accurately. Those issues can be present even in a year or two of data. Which is why multi-year studies are the best for APM results. They provide a very large sample size, and we get to see the player with and without a large group of teammates. Which means that an individual player's impact can be isolated to a great extent from his teammates, and that those results are much more accurate. So again, your protests above are just wrong when applied to the APM results that have been presented in this project.
3) The 2004 - 2009 APM study widely referenced here was NOT just from KG's poor Wolves teams. In that 6 years, KG played for a 66-win team, a 61-win team, a 58-win team, a 44-win team, a 33-win team, and a 32-win team. AND HIS HUGE APM MARKS STAYED CONSISTENT! So no, you can't just draw the conclusion that because KG was the best player on crap Minnesota teams his APM would be higher. He had the huge marks on both crap teams and great teams, consistently. On the flip side, back to your example, Amare had poor APM marks whether with Nash in Phoenix or on his own in New York. Again, the entire purpose of APM is to try to isolate a player's impact from his teammates as much as possible. And in KG's case, APM indicated that he had a similar impact on his 32-win team in 2007 and his 66-win team in 2008, and in both cases that his impact was HUGE.
Both your complaints, and the logic behind them, are just factually incorrect here.