tsherkin wrote:drza, excuse my ignorance, but has anyone bothered to look if it's possible to exert a larger margin of impact on crappier teams? E.G. are there diminishing returns on player impact as presented via RAPM?
I had seen this question and hadn't gotten to it yet, but my answer is similar to the ones that therealbig3 and acrossthecourt gave. That I don't know of a study yet on the diminishing returns question you asked, but that there is a lot of anecdotal evidence that such an effect isn't significant. They mentioned KG doing well in Boston, but there's more examples than that. For example, if you look at Doc MJ's normalized RAPM spreadsheet from 1998 - 2012, there aren't a lot of players in the top-20 that are associated with a lot of bad teams. KG obviously had some bad teams. LeBron in Cleveland. Kobe had his mid-00s Lakers. Paul Pierce at #19 and Chris Bosh at #20 also had some poor teams.
But for the most part, the top of the list is full of guys that played primarily with good teams. And for those that did play on bad teams, there's no indication that the bad teams boosted their quality. KG was #1 in 2003 with below average support, dropped to "only" top 5 in 2007 with ridiculously terrible support, and was #1 again in 2008 with great support. Kobe was solid in 2006 and 2007 with his meh support, but his best years were from 2008 - 2010. Pierce's best results were on the 2008 Celtics. LeBron's results were better in Cleveland, but he still hit #1 in Miami. This isn't rigorous, but it seems unlikely that a potential poor-team-RAPM-inflation, if it even exists, is explanatory in this case.
tsherkin wrote:Right, but the problem with 08 is that the Celtics had Thibs as well, a guy who would go on to make some remarkable defensive impact on the Chicago team for whom he is now the head coach, yes?
It makes the full measure of KG's impact hard to properly assess, though obviously his injuries do help correlate his presence with maximum defensive efficacy. His impact on D was obviously great, but it was obviously more than just him, so that means he can't take full credit, IMO. It's something to think about, I believe, worth consideration.
I wonder, as always, are people conflating greatness and individual player ability, which are related, but not exactly the same thing. KG was clearly a remarkable player: a very good overall offensive player (second-tier offensive star) and an A-list defensive presence who did some great things and was probably miscast as a first-option offensive star due to the bungling inadequacy of Minnesota's management team.
As ever, I also question the value of single-number metrics as a Grail stat used to justify serious change in a player's ranking. KG was obviously very good across a variety of analytical POVs, but his case seems entirely made on the basis of RAPM and that is not a comfortable thing for me. Boston had a remarkable defense, but it behooves people to remember that in 2008, they added KG, Thibs had a wicked and keen scheme, we aren't talking enough about James Posey, or the coaching impact of KG playing more limited minutes as a way to keep him fresh, right? Pierce and Ray were both playing basically the best defense of their careers even when Garnett wasn't on the floor, in part because of reduced offensive load, in part due to coaching... in part due to intangible leadership elements from KG (he's a great communicator on D, surely).
I was going to comment on your post about KG having a smaller offensive role in Boston, but it was directed more at therealbig3 and he did a great job addressing it. So instead I'll jump in here.
You raise some great discussion points here, because I think they address issues that many of us have. And ironically, these very reasons are a big reason why people like looking at +/- results. So, let's start with the first part I underlined. Because you're right. The Celtics' team defense was amazing, and Garnett shouldn't get all the credit. Ironically, that is kind of the point that we're stuck with in the absence of any "impact"-style stats. Without those, everyone is left to only guess how much credit to apportion to a particular individual for a team result.
However, here we have some data to work with. The RAPM data doesn't say that Garnett is the best defensive player in the NBA, or that he gets all of the credit for the defense. Instead, it says that Garnett's presence on the court correlated with an amount of defensive scoring margin improvement that was larger than the amount of correlation that any other player had with defensive scoring margin. That does NOT prove that Garnett CAUSED that defensive improvement nor does it say how good Garnett's defense was. It simply is a data point that Garnett's presence was more associated with great defensive scoring margin results than anyone else in the NBA. It is up to us, through other analysis and observation, to interpret whether we believe that correlation to be causal.
And to your point, it's important to note that interpreting that correlation to in fact be causal is NOT a zero sum solution. This doesn't say that Thibs' schemes aren't great, or that KG was out there by himself. You could note, for example, that Perkins also had positive defensive results and the best of his career. Pierce also had positive defensive results that were better than he had in years. Rondo had positive results. The defensive credit is shared among the Celtics players, but KG had the highest mark. And though I haven't done it, perhaps if you analyzed the individual defensive RAPM marks of all of those players and compared those results to years when they weren't in Thibs' defensive schemes, you might see an improvement in the Thibs years that would also help to quantify the potential impact of his schemes.
But the point is, RAPM isn't a holy grail stat for goodness. It simply is a measurement of how much a player's presence correlates to given scoring margins. It doesn't give all of the credit to one individual. But it does give a result that can help to quantify how much different individuals are contributing to the whole. The very problem that you posed in the first underlined section above, is the very thing that RAPM data can give evidence to help answer.
tsherkin wrote:I see the "lightning strikes 14 times" thing and it's great, but what makes me stop and pause is that KG's impact is supposed to be so remarkable, but we still haven't resolved whether that should matter in the absence of team results or not. This is about greatness of career, right? Not about who took crappy teams to moderate success, so theoretically there should be some element of winning bias inherent to the very concept we're discussing, and that should have some weight as far as KG's career is concerned. If not that, then his inability to fill the totality of the role which he was assigned in ways which others of his peers were able to manage. And again, we're looking primarily at RS RAPM, if I'm understanding this correctly, when most of the criticisms of Garnett pertain to his playoff offense, which is considerably worse than his regular-season offense, yes? There's a gargantuan difference in his performance on that end of the floor once the playoffs begin, so I'm given to wonder just how much that affects an impact stat like RAPM. Perhaps there's data at which I've not looked?
These are the questions floating through my mind, since I"m admittedly not really deep into the +/- stuff.
I kind of lost you a bit on the first underlined. Maybe your question was relating to your previous thoughts about a poor-team bias and suggesting that this is what has helped KG? If so, as I mentioned above, there is no evidence that I've seen that a "poor team bias" would be in any way explanatory here.
As for the bolded/underlined part, a few things. One, most (all?) RAPM studies that I know of use postseason results as well. The old APM studies used to double-count the postseason to really give it more weight, and I don't know if the RAPM results do as well. Second, I don't see this "gargantuan" difference in offensive performance that you note. In this project I've posted the unadjusted TS% for KG in the postseason vs the regular, and it's like 2 or 3% which emphatically is not a gargantuan difference. Then, I also broke down his career into different parts (e.g. Minnesota vs Boston), and looked at each playoffs individually from 1999 - 2004 and his scoring efficiencies against similar competition were very similar to Duncan's. Then, Acrossthecourt took that one step further and analyzed his offense in the postseason specifically against level of competition, and there was no significant difference at all in his performance.
So all told, in review, RAPM isn't a holy grail stat of goodness. It tracks how a player's presence correlates with scoring margin, which is a useful data point. The postseason results are included in the calculation. And there doesn't appear to be a poor team bias, and if there is it doesn't appear explanatory to KG's results.
