penbeast0 wrote:lessthanjake wrote:OhayoKD wrote:... The difference here is ben's method is demonstrably more predictive. In other words, it ties into what actualy matters(making offenses better).
Am curious how Ben's era-adjustment works, but older players certainly aren't struggling with it(magic and nash dominate Lebron). Granted, I realize it's inconvenient for certain priors(cough Bird cough)
1. Ben’s methods are “demonstrably more predictive,”
as against Ben’s own subjective assessments. That’s the point! ....
Actually, if what OhayoKD says is true, it should be demonstrably more predictive, not against subjective assessments, but against team/lineup points per possession numbers since that is what making offenses better means at least to me. If he can show this or you can disprove it, then the arguments settles out at least for one reader.
There was an article that compared box-oc, pr, team-assists, ast:tov in terms of correlation with o-rating, but I'll need to look for it(hopefully it's not locked behind a paywall now -_-).
Went box:oc -> ast:tov -> pr -> raw assists from memory but feel free to discount that for now
Using the links provided on this thread Ben actually tests his regression goal(oppurtunities created) against assists using "expected value" which he also checks against offensive and defensive rating(And naturally box-oc maps much better to that than assists do):
Shouldn't be surprising oc looks better here given that
A. It is tracking missed shots
B. It is tracking creation that comes
before the final pass
C. Tracking creation that occurs
without a pass(steph, reggie, bird!)
Ben also directly checks o-rating against boxoc and pr here too(no comparison to ast:tov though he literally spends a bunch talking about them -_-)
https://youtu.be/yoLgSWA7n6g?t=470Maybe someone like uh...
moonbeam wrote:
...would be interested enough to consider doing a direct o-rating correlation check?
And if possible, I guess "adjusted passer-rating" could also be given a whirl?
lessthanjake wrote:OhayoKD wrote:ijspeelman wrote: These stats are formulas made to fit a wholly subjective analysis.
the idea that "the pass before a made basket created said basket" is also "wholly subjective". The box-score is
inherently subjective, that is why we look at things look at results, impact, and winning in the first place. The ball can objectively go into the hoop. How you credit everyone else for the ball going into the hoop and what labels you ascribe is subjective.
The actual point of the "subjective analysis" was to help see who is helping their teammates score more, aka,
helping improve offense more. That is literally the point.
2. Also, correct me if I’m wrong, but I’m fairly sure that there’s no such thing as “Ben’s era-adjustment.” Rather, as I understand it, that was something that someone here did to “adjust” for the fact that Box Creation is lower in prior eras. And it’s an adjustment that I don’t really think Ben would even approve of, since he specifically mentions that in his hand-counted analysis, there actually *was* less creation in those earlier eras. But I don’t have the source of this “adjusted” data, so I don’t actually know for sure where it came from.
iirc this was sent over by someone from his discord:
https://docs.google.com/spreadsheets/d/1NcqcHXyV28OPJpXHz2eSiHZ98x-_WUrtBlb_Z2dcF2A/edit#gid=0Ben did do some "box-creation across history" whatever(or more specifically top 5 creation averaged), so that may be the basis of his adjustment. Does give older players a big boost from the vanilla thing. I do not think it was from realgm.
Regardless, raw creation being lower would not necessarily mean an earlier player was less valuable
relative to the league which is why you would do this if you were comparing players in an era-relative lens.
He does a similar thing with scoring-volume in vids but I do not have a patreon-subscription to check nor am i planning to get one.
For posterity here's the methodology for pr and box-oc:
Formula incorporates i) Layup assist percentage (since 2002), ii) Offensive Load, iii) Assist-to-Load ratio (per 100), iv) Non-3 Creation-to-Load ratio, v) Height, and vi) Turnovers
As of now, this seems way more useful to me than the alternatives(maybe someone can post whatever playval is based on though iirc it was some bpm/o-rating/box combo) and it doesn't seem to be low on off-ball guys or older players.
whatever approach you use(raw box, pr/oc, playval) Magic seems to come out at the top along with nash, lebron is at the next tier, curry is close, and then bird doesn't look all that(comparable creation to kobe in pr/oc and raw box but I don't know how the play-val compares).
I'm guessing that's probably because of Bird's limitations as a ball-handler, lack of volume 3-shooting, slashing, and rimited rim-gravity, but we don't have pbp or whatever for his prime, so I guess you can dismiss that all as biased if you want.
