sp6r=underrated wrote:I think for Kobe I'd be fascinated just 00-13. He had some garbage years with decent minutes at the end and the beginning. That said I'd can't imagine those minutes are what his driving him this low as his peak years are very strong but not spectacular.
The closest analogues we have available are the 2001-2012 and 2001-2014 datasets released by Daniel Myers (aka the inventor of BPM).
2001-2014 RAPM:
https://public.tableau.com/views/14YearRAPM/14YearRAPM?%3Aembed=y&%3AshowVizHome=noThis dataset places Kobe at 15th overall (+5.3), placing 4th in ORAPM (+6.2) and 745th in DRAPM (-0.9). The three players above Kobe offensively were LeBron (+7.2), CP3 (+6.6) and Nash (also +6.6). This should be a fair comparison to 2001-2013, since Kobe only played 6 games at the end of his prime in 2013-14, so those values shouldn't sway things too heavily.
The 2001-2012 RAPM dataset is located here:
https://public.tableau.com/app/profile/dsmok1/viz/RAPMvsMPG2001-2012/Dashboard1This dataset places Kobe at 12th overall (+5.5).
I do agree in that Kobe seems low, and this seems to hold even in playoff only datasets that I've seen, but there are a few things to note/a few blind spots regarding RAPM that might be beneficial to Kobe -
1) RAPM considers impact within a player's role, not overall holistic impact/goodness. Ergo, if somebody thinks there is more wiggle room for Kobe's role to change, or for the timeframe in which his playstyle is more effective to change, one might consider him underrated. This can also lead to fun (or not so fun) collinearity errors based on how both the player in question and other players might change their roles within the relevant lineups.
2) RAPM may not account for scalability for offence, i.e. the ability for somebody to "get buckets" when needed correctly (alternatively, the frequency in which these situations may undermine the actual usefulness of a player of this skillset). Similarly, it may not account for effort-based improvements in defence in the playoffs. We've seen Kobe go on some ridiculous scoring streaks that some of the more lower volume scorers, or higher volume, but more "metronomic" scorers are incapable of.
3) RAPM may slightly overrate defence, on account of the fact that coaches influence defence more than they influence offence, ergo, the fact that Kobe's defence is negative via DRAPM might be slightly overblown in a non-coaching RAPM dataset (such as the 28 year dataset linked above). This doesn't actually really affect Kobe that much in the 28 year dataset since Phil Jackson is
the top coach in coaching RAPM over the same time period, but is worth noting for smaller datasets such as the ones I've linked above.
4) RAPM, being a per-minute stat, doesn't factor in minutes played, nor does it factor in fouls very well, which may limit the ability for players to maintain their impact/scale their minutes up in closer games as their playstyle may change defensively in accordance to this. Alternatively, they may just plain get fouled out
(note how basically all of these can be explanations as to why Amir Johnson ranking so highly in the 2001-2014 dataset is not sudden proof that Amir Johnson is an undercover All NBA player or something)
5) As a corollary to point 4), higher foul drawing players such as Kobe may reap some of the benefits of foul drawing in terms of penalty time, where Kobe may bring the Lakers
to penalty time, but see that the benefits of penalty time materialise when Kobe's on the bench.
6) RAPM is a
regression, and the regression may struggle to parse player variance out effectively. One example that's commonplace in older RAPM discourse is Kevin Durant:
In 2001-2014 RAPM, he was only +1.3 (ranked 159th) which is below his On court and On/Off ratings of +2.7 and +2.4 in the same time period (Durant had horrible on/off his first two seasons in the league). Westbrook has an RAPM of +4.2 with an On rating of +4.3 and an On/Off rating of -0.9. Those raw On/Off ratings would generally lend themselves to a somewhat RAPM. Similarly, Serge Ibaka was +2.2 in RAPM with +6.2 in On and and -0.7 in On/off respectively (and he wasn't there in the first two seasons of Durant being lousy).
If we look at Durant in 2010-2014, his On is +7.5 and On/Off is +6.5. Westbrook is +6.4 and +0.2. So, Durant has the most impressive signals of the 3 when they played alongside each other. This is corroborated further through Cheema's 2010-2014 RAPM, where Durant is +3.85, Westbrook is +2.76 and Ibaka is +1.17. I suspect that whilst Durant should obviously be "punished" for his 2008 and 2009 samples, some of his positive impact (and the sheer improvement of OKC) is being attributed to other players such as Westbrook and Ibaka joining the team whilst Durant is improving as a player. In other words, the regression may be attributing some of Durant's improvement to other players.
Now, Kobe had a few years in his prime where he had a down season (e.g. 2005) and so the regression might view the Lakers' team performance in an oddball way, where, for example, the departures of 2004 might have been interpreted as Kobe "holding the team back" rather than Kobe actually having a down season.
I do suspect this may have happened to Kobe defensively, where his "lone gunner" years in the middle of his prime might be coalescing with his stronger impact years in a way resulting in him having some
very low DRAPM numbers.
I'd advise that looking at RAPM in 3-5 year blocks can help inform interpretations of RAPM, as shorter blocks can limit the effects of point 6) above, as well as potentially mitigating the effects of changing role in point 1). I'd look at both large
and smaller blocks of RAPM, but I generally find the smaller blocks to suit my purposes more.