johnnyvann840 wrote:MGB8 wrote:No, they don't. Their BPM has him at 2.4 for 2019-20 - with +3.2 on offense and -0.8 on defense. See below, under "advanced stats."
https://www.basketball-reference.com/players/l/lavinza01.htmlYou are misusing the "on-off" subset of data in that what you are pointing to is the "raw" plus minus, that is not adjusted in comparison to the average player (as a way of trying to adjust for opponent quality).
OK. You're using BPM. In your post, you stated a positive +/-. I assumed you were just talking about his raw +/- on/off numbers. Not a fan of BPM because of the way they adjust. I wish they just left RPM alone instead of changing into a offensive only stat. It's the same problem we have with Box Plus/Minus.
BPM does not use play-by-play data. it is not truly a plus-minus stat, so it's name is misleading. here is the formula:
Raw BPM = a*ReMPG + b*ORB% + c*DRB% + d*STL% + e*BLK% + f*AST% - g*USG%*TO% + h*USG%*(1-TOV%)*[2*(TS% - TmTS%) + i*AST% + j*(3PAr - Lg3PAr) - k] + l*sqrt(AST%*TRB%)
3PAr = percentage of FG attempts that are 3pt attempts
Tm = team
Lg = league
it is the best of the box score stats, but according to a study i saw it is only about 70% as predictive as an advanced plus-minus stat
as for RPM, the changes they made were basically an acknowledgement that defensive +/- stats are very noisy from season to season and thus create too many wacky single-season results. and RPM aspires to be a stat that the lay person can look at and have confidence that is is a reflection of how good a particular player was in a particular season. thus, by making the changes they did, those single-season results LOOK better (particularly to the many who focus on scoring ability), but it also sacrifices much of the unique value that advanced plus-minus stats are supposed to bring to the table. plus, there has always a box score component to RPM anyway. before this season's changes, i've read that it started each season as a box score stat and gradually mixed in play-by-play plus-minus data until there was a 50/50 split by the end of the season. so it was always a somewhat bastardized advanced plus/minus stat, and now even more so. i find it of little value at this point. probably not much better than BPM
the reality is that any pure advanced plus-minus stat IS GOING TO PRODUCE WEIRD OUTLIERS for a single season. the sample size simply isn't sufficient. i've read that 5000 minutes of play are required for the results to "stabilize" or whatever (2 full seasons for a typical nba starter), but that's obviously not a magic number. the more data the better. but if you chart a player's CAREER, it's a beautiful way to see the progression of that player's true impact on winning. of course, it doesn't factor in clutch performance, but anyway...
luckily, there is another regressed plus-minus stat that has come out in the past couple of years that seems to do a better job than RPM ever did: PIPM. it attempts to modulate some of the wild swings that come with this sort of stat. for example, it mutes the impact of opponent 3pt %, which the creator has determined is very much random and not particularly impacted by individual players on the defensive end. it is a pure RAPM stat with minor adjustments to reduce randomness, basically
anyway, at the risk of blowing gaskets, here are the offense/defense PIPM results for lavine/booker/trae (talented scorers who are defense-challenged) and sato (role player) over their careers:
sato2017 (2.58) 0.22 spot minutes as rookie
2018 0.33 (0.23)
2019 0.19 (0.16)
2020 0.42 (0.58)
consistently average nba player on both ends of court over career so far. borderline starter/backup
trae2019 2.33 (3.41) - slow start offensively rookie year, came on later in year
2020 5.02 (3.69) - became elite offensively, still one of the worst defenders in the league
almost certainly a net positive player with room to grow (offensively at least - maybe he can become smarter and more opportunistic defensively as well)
lavine2015 (2.00) (2.94) - truly awful as a rookie, but drafted as a talented project
2016 (0.41) (3.07)
2017 (0.25) (1.38) - season shortened by injury in minny, but clearly no longer one of the worst players in the league
2018 injured
2019 0.35 (1.37)
2020 1.49 (2.55)
still probably a net negative impact on winning despite significant offensive improvement and some degree of defensive improvement. great in the clutch this year, though
booker2016 (1.13) (2.17)
2017 1.02 (2.20)
2018 0.49 (2.05)
2019 2.49 (3.76)
2020 3.58 (2.45)
steady offensive improvement over career, now at least CLOSE to elite impact offensively. consistently very bad defensively. probably a positive impact player at this point, but not deserving of all-star consideration if impact on winning is a factor
https://www.bball-index.com/current-pipm/again, we should not be fixating on single season numbers with this sort of stat. it's value is mainly in providing a trajectory of a player's true impact on winning over the course of his career. role is also a significant factor. if for example, you task a guy like sato with shooting 20 times a game, his impact is going to deteriorate. because that's not the kind of player he is. and if you reduced the usage of a zach lavine, his numbers might improve. and probably would...assuming that indeed he is not, as most people feel, suited to be "the man" on a winning team
the donald, always unpopular, did worse in EVERY state in 2020. and by a greater margin in red states! 50 independently-run elections, none of them rigged