trex_8063 wrote:Owly wrote:Haven't got time to read fully but am pretty confident that if you go back to the post it will be in the context of "value above X" that I refer to Baron disadvantaged by having smaller samples by lesser opportunity.
Though fwiw in this comp if one believes playoff performance is more real than noise then I would think Baron getting more playoff minutes should boost his RAPM given he's a high end improver in box and impact side measures.
Fair enough, that's my bad: I missed the context of "value above X". Though fwiw, do we know for positive those specific calculations include BOTH rs and ps? It would seem odd to include the ps for a cumulative stat, for what I think are obvious reasons. I'm thus wondering if it actually does.
And I guess mentally erased the "if" in the second statement. I apologize for mischaracterizing what you said.Owly wrote:Player: to 14 RAPM (rank); to 22 RAPM (rank); to 24 RAPM (rank)
TP: 2.15 (68th); -0.3 (559th); 0.7 (718th)
BD: 2.66 (40th); 6.1 (30th); 5.8 (26th)
I meant to ask a question about these figures.
In a sample of '97 to '14 he's 40th........then somehow he jumps to 30th (with an RAPM more than twice as large) in a '97 to '22 sample (even though he retired in 2012)????
Then his rank jumps still further in a '97 to '24 sample??
That seems a pinch fishy to me (does this again speak to the wild variability in weighting of priors, etc??).
I understand there will be a few players whose careers started maybe between 2005-2010 (or thereabouts), whose career RAPM dropped by 2022 or 2024 (because of their twilight years now within the sample)--->and thus fall behind Baron.
otoh, I'd expect that to be MORE THAN compensated by new [elite] players whose whole or most of careers have come along after 2014 (e.g. Jokic, Curry, Embiid, Jimmy Butler, Tatum, Lillard, Draymond, Paul George, Kawhi, Jrue Holiday, and likely Harden or Durant pulling ahead during those years).
idk, it just seems odd that he CLIMBED by FOURTEEN places from '14 to '24, without playing a game, with all of the above [12!] names pulling ahead within that same time period.
Is it possible that 26 others fell behind? And in many instances---if they did----is it not because they played WELL past their primes (maybe 17-20 seasons)? And if the latter is true, this is sort of punishing guys for having some lingering usefulness in the league (where Baron could not do so)?
Anyway, more likely what happened is simply a change in the formula.......for I notice there are other guys who ALSO didn't play another game after 2014 [or barely did], who were AHEAD OF Baron in the '97 to '14 sample, yet suddenly are NOT in the later samples (e.g. Gheorghe Muresan [15th in '14, tied for 176th!!! in '24], Terry Mills [13th in '14, tied for 126th in '24], Byron Scott [16th in '14, falls to a tie for 357th in '24 sample!!!! (what?!?)], Alonzo Mourning, Clyde Drexler, Detlef Schrempf, Bo Outlaw [21st in '14, falls to 92nd in '24], Mookie Blaylock, Arvydas Sabonis, Nate McMillan [25th in '14, tied for [b]146th in '24], [possibly Kobe Bryant: I'm skeptical '15 and '16 could have dropped him so far; ditto Metta WP], Ron Harper, Steve Nash, Josh Howard).
This is perhaps "whatever", as maybe the newer formulations are "better". But again, the sometimes WILD (see Byron Scott) differences we can see, apparently based upon nothing but a change in formulation is a little disheartening.
And idk what to think sometimes; there's not a ton of transparency on these things.
As ever not an expert. My first glance thoughts
Regarding “doubling” I wouldn’t think that a concern. Understand scaling isn’t consistent or tied to real points in the way on-off is so rank and separation within metric would be what matter.
I terms of movements … the (I think) easy to understand
Byron Scott: (to a lesser extent others) it’s a one year 1440 minute sample. I’d probably just use the yearly variant but certainly this isn’t getting the level of data where the notional benefit of these larger samples comes from. Movement in players he’s playing presumably against but primarily with will move him. To be honest his +3.4 on-off (+6.1 on) on a small sample leading to an exceptional RAPM was probably always a bit unusual. Perhaps too much assuming rookie Kobe was what he would be later?
Looking at the other stuff
Looking at the rising at the players dropping well the 39 initally above were
Spoiler:
I’d can kind of see a bunch of names I’d expect to drop.
There’s some that aren’t thought of as exceptional, some who perhaps didn’t have blow you away numbers in the on-off (certainly Scott) and are working off small samples.
Some then-active players who would add more weaker years.
Several players close enough in list one or the other that the exact rank doesn't matter so much.
Fwiw Harden was already ahead for ’14.
Is it possible that 26 others fell behind? And in many instances---if they did----is it not because they played WELL past their primes (maybe 17-20 seasons)? And if the latter is true, this is sort of punishing guys for having some lingering usefulness in the league (where Baron could not do so)?
I mean if one were ranking solely on RAPM. Though that might also depend on how one feels about being harmful on a big salary. I also think the framing here is … generous … to the extent, for instance, that Kobe’s awful last three years do harm … it’s because he wasn’t providing any “lingering usefulness”
So yes, minutes always matter. And squeezing a career into one figure is always going to cause some issues. And where that’s an average that will always punish longevity and impact-y stuff probably gets more tricky.
At the same time … Baron had a pretty solid showing in “value added” variations off his weakest projection so I’m not sure I see this concern. So yeah we probably should shift a full career David Robinson back above him for RAPM. Or it makes sense that we still like those true prime Dwight years and aren’t going to hold the back half of his career against him.
It’s case by case contextual but I’m not seeing a shocking new wrong here.
The inactive fallers are mostly, I think, what is covered above (i.e. small samples regressed back more aggressively, mostly probably rightly so off the limited information available … Robinson, Blaylock and Outlaw look good in 94-96 but that’s not available to it).
Though fwiw, do we know for positive those specific calculations include BOTH rs and ps? It would seem odd to include the ps for a cumulative stat, for what I think are obvious reasons. I'm thus wondering if it actually does.
I assume the reason it would be odd to do a cumulative stat with postseason included is what I’ve been pointing out … the uneven opportunity. I think some others may care less that.
Do we “know for positive” … I know probably as much as you, mostly what I’m told here. That has led me to believe it’s all combined from that source.
Is there a lack of clear labeling, as others have said, yes. I would assume curious casuals on non-subscription platforms aren’t the primary focus of people good at doing this stuff’s output which may be one reason there isn’t a definitive number. IDK. It’s not ideal. It probably just is what it is at this point.
