Page 1 of 3
A case where PER is NOT "good enough" to use
Posted: Wed Dec 28, 2016 3:41 pm
by Crow
According to BPM and RPM Joel Embid is giving a slightly above mean performance. Winshares per 48, slightly below. According to PER he is having the 23rd best season in the league (among guys at least in regular rotation) and thus probably is already All-Star worthy. One of these is not like the others because of the extra heavy reward of usage and exclusion of shot defense. RPM estimate has him as 26th best center on overall impact. Quite a difference between 23st best player on PER - HOL (hyped offensive line) and 26th best center on RPM (a "whole" metric).
(Attention certain podcasters, writers, the largest sports media company and maybe certain team execs who still heavily or at least regularly and too heavy for my tastes use this metric over more modern options. Not that I expect to sway you. There are more other options too, as good or better than these next most used options. Will you still be using 20 years old already PER in 5-10 years? Your call. Use a box score metric blind to 25% of the game in shot defense and supercharged on usage to make it "look right" in star crushed eyes but again partially blind to offensive impact on others or maybe something complete and aware of play by play results? Hmm... tough call... Easy to just not think / change. Don't want to stress people to move forward, or do you? I guess you can straddle if you want. Maybe, using a blend of old and new thinking, Embid is around the 75th best player? You could say that... but are you? Saying he is 23rd best and / or 130th or so isn't saying it directly.)
Re: A case where PER is NOT "good enough" to use
Posted: Wed Dec 28, 2016 5:44 pm
by rlee
Wouldn't that be "all cases"? Can't think of a time when I've seen or heard it and it actually was useful in a meaningful way. And often quite misleading.
Re: A case where PER is NOT "good enough" to use
Posted: Wed Dec 28, 2016 9:01 pm
by Mike G
In the possession-counting era, just a handful of players have averaged, per 100 poss., at least 35 points, 14 rebounds, and 4 blocks.
Code: Select all
per100 yr Pts Reb Blk 3FG TO TS%
Ewing 1991 35.1 14.7 4.2 .0 4.7 .561
Robinson 1994 39.2 14.1 4.4 .2 4.2 .577
Robinson 1995 36.9 14.5 4.3 .1 4.3 .602
A Davis 2015 35.4 14.9 4.3 .0 2.0 .591
Embiid 2017 37.6 14.8 4.8 2.4 7.5 .575
The turnovers are staggering, but he shoots 3's (currently at .403 accuracy)
What would be his WS/48 , if he played for the Warriors?
Re: A case where PER is NOT "good enough" to use
Posted: Wed Dec 28, 2016 10:13 pm
by Crow
I could have said all cases but I guess I tried to apply some restraint at least in the headline. PER is still used in podcasts, articles and has prominence at ESPN with special "advanced" billing and inclusion on player pages whereas RPM is not included on player pages.
His shot defense component on ws/48 would be better with GSW, deserved or not; so ws/48 has major issues too. There is plenty of reason to move to RPM or boxscore / RPM / player tracking blends. Especially if the latter got released more regularly.
Re: A case where PER is NOT "good enough" to use
Posted: Wed Dec 28, 2016 10:28 pm
by Crow
Embid is in rare company by that very demanding stat combo.
It is still fairly rare but not as elite on this one
http://bkref.com/tiny/mZfS7
Of players with 30pts / 10 rebounds or more per 100 possessions, Embid's offensive rating (including turnovers and the stuff) is 252nd of 254. Alternatively comparing his well below average 99 Offensive Rating to current players, he comes in at about the 33rd percentile (2/3rds better). ORPM and OBPM have him below mean performance too. So you can think of him as really special on offense or given really special prominence and doing quite underwhelmingly so far with it, most or all things considered over less. Bright spots are bright and important but still just spots.
Re: A case where PER is NOT "good enough" to use
Posted: Wed Dec 28, 2016 10:52 pm
by Dr Positivity
NBA.com's Race to the MVP has Derozan at 7th and Lowry not even in the top 15. Their rank in RPM, BPM, WS/48 and PER:
RPM:
Lowry - 4th
Derozan - 109th
BPM:
Lowry - 9th
Derozan - 83rd
WS/48:
Lowry - 11th
Derozan - 31st
PER:
Derozan - 13th
Lowry - 19th
Lowry's on/off is +14.5, Derozan's is -10.5
By RPM and BPM 7th is accurate for Derozan... on his own team
Re: A case where PER is NOT "good enough" to use
Posted: Wed Dec 28, 2016 11:07 pm
by Crow
Yeah, the metrics all raise concerns but they are not all the same level of concerns. Consistency or inconsistency with other metrics isn't objectively a winning / damning sign but it is something to stay aware of and think further about.
Re: A case where PER is NOT "good enough" to use
Posted: Thu Dec 29, 2016 7:00 am
by Mike G
A season somewhat similar to DeRozan'17 was put up by Allen Iverson in 2001
Code: Select all
. Pts/100 TS% TRb% Ast% Stl% TO% ORt DRt
Iverson 39.3 .518 5.2 23.0 3.2 10.0 106 99
DeRozan 39.3 .557 8.4 20.6 1.9 9.5 114 109
39 points per 100 possessions
and Ast% >20 is a rare combination. Before 1994, only Jordan had done it; then Robinson, Malone, Iverson (2 times), McGrady, Kobe (2), LeBron (4), Wade (2), and Durant (2). This year, Westbrook and DeRozan.
Jordan did it 9 times of the 24 before this season.
http://bkref.com/tiny/KYpRN
Lowry's on/off is +14.5, Derozan's is -10.5
Lowry is mostly backed up by Cory Joseph, while DeRozan is mostly replaced by Norman Powell.
Code: Select all
. Raps PER TS% WS/48 BPM RPM
Joseph 14.0 .500 .091 -0.9 -3.3
Powell 16.3 .619 .163 1.8 1.2
A few points at least from that 25 point difference may be due to their replacements in the lineup.
Re: A case where PER is NOT "good enough" to use
Posted: Thu Dec 29, 2016 3:18 pm
by AcrossTheCourt
Lowry typically crushes him in on/off or adjusted +/-. It's not something to be written off. It's a pattern.
Metrics like BPM or RPM have team adjustments, at least indirectly, where the collective rating equals the team's rating. If you do the same kind of adjustment BPM does with Philly, Embiid's PER drops, but not by much. But PER overvalues usage and blocks, so it's no surprise he looks best in PER.
Re: A case where PER is NOT "good enough" to use
Posted: Thu Dec 29, 2016 5:04 pm
by Mike G
By RPM and BPM 7th is accurate for Derozan... on his own team
Lowry typically crushes him in on/off or adjusted +/-. It's not something to be written off. It's a pattern.
Last year, Lowry was about 6 Pts/100 better in RPM; this year it's about 7 difference.
If DeRozan is really the team's 7th best player, why does he get 2nd most minutes? Coach Dwane Casey is in his 6th season, each year with a better Win%; so he doesn't seem to be an idiot.
For some of last year's stats, here are correlations with Min/G, broken into positions:
Code: Select all
pos WS/48 e484 PER BPM ORPM DRPM RPM
PG .54 .73 .67 .74 .68 .39 .67
SG .45 .64 .62 .69 .64 .00 .45
SF .40 .63 .54 .60 .68 .20 .57
PF .32 .58 .41 .65 .61 .26 .54
C .01 .63 .36 .42 .23 .34 .38
avg .34 .64 .52 .62 .57 .24 .52
At least one of these (eW/484) has a mpg component.
Some other correlations, per36 rates to MPG --
Code: Select all
pos TS% Sco Reb Ast PF Stl TO Blk 3s
PG .42 .59 .41 .48 -.11 .46 .41 .49 .59
SG .48 .64 -.04 .42 -.09 .28 .42 .21 .67
SF .21 .59 .32 .49 -.42 .25 .46 .05 .69
PF .24 .60 .09 .38 -.37 .19 .29 -.06 .47
C -.05 .47 .42 .30 -.34 .11 .43 .09 .16
avg .26 .58 .24 .41 -.27 .26 .40 .16 .51
That's just total 3FG for the year in the last column.
Scoring proficiency is still the biggest determinant of minutes.
Re: A case where PER is NOT "good enough" to use
Posted: Fri Dec 30, 2016 4:17 am
by Dr Positivity
Joseph has actually played slightly more minutes with Lowry than he has with Derozan, with the Lowry/Joseph pairing a lot more effective at +10.3 to -5.0 (Derozan and Joseph together is one of the Raptors worst lineups because of spacing problems with two non 3pt shooters)
I don't think Derozan has to be called the Raptors 7th best player, I don't have that much of a problem with putting him as high as 2nd or 3rd although Patterson has a good argument for #2... But I don't think there's a credible argument for him being the MVP of the team considering how much more their success correlates with Lowry and Patterson than him
Re: A case where PER is NOT "good enough" to use
Posted: Fri Dec 30, 2016 10:43 am
by Mike G
Here are guards and G-F of the last 20 seasons who have scored at least 30 Pts/100 poss, with less than 10% of their FG being 3's
Code: Select all
pp100 w <.100 3PAr (G/GF)
39+ Jordan'98, DeRozan'17
38 Wade'07
37 Wade'06
35 Jordan'02, Wade'15/'12, Parker'09
34 Kobe'02, Wade'08
33 Hamilton'03, Wade'05
32 Westbrook'11, Wade'13/'16
31 Hamilton'06/'02, Parker'13
30 Parker'07/'08/'06, Wade'14
So this is prime D-Wade territory. A full tier above Parker and Rip.
Most of these were not the main man for their team, but all were instrumental in great success.
Patterson takes 2/3 of his shots from the arc, is 80% assisted, finishes at the rim at 40%
DeRozan is assisted on about 25%, hits almost 70% inside 3'.
The Raptors' lineup has hardly changed since last season. Which players have improved to account for the team's improvement?
Re: A case where PER is NOT "good enough" to use
Posted: Tue Jan 10, 2017 5:16 pm
by Mike G
Crow wrote:According to BPM and RPM Joel Embid is giving a slightly above mean performance. Winshares per 48, slightly below. According to PER he is having the 23rd best season in the league (among guys at least in regular rotation) and thus probably is already All-Star worthy. One of these is not like the others because of the extra heavy reward of usage and exclusion of shot defense. RPM estimate has him as 26th best center on overall impact....
He's now #17 C in RPM, behind such luminaries as Mareese Speights and Dewayne Dedmon.
Among those going more than 24 mpg, he's 12th -- just ahead of Towns (1.46 vs 1.25)
http://www.basketball-reference.com/tea ... 17/on-off/
Here he's seen to have the following impacts, vs. opponents, relative to not-Embiid:
eFG% +.030
TReb% +4.6 (O and D both +1.5)
Stl% -0.8, Blk% +4.7
TO% nil
Pts/100 +10.8
That +10.8 is much better than 1.5, due either to RPM being shy about the small sample span, or due to weak alternatives in Philly, or both.
He's now at .105 WS/48 and 2.9 BPM, still with the 23.3 PER
Among those with 600+ minutes, he now cracks the top 100 in WS/48, he's 38th in BPM, and 19th in PER.
It might be argued that without PER, there was no metric indicating how good he was.
With PER
in the mix, we may get a better estimate.
Re: A case where PER is NOT "good enough" to use
Posted: Tue Jan 10, 2017 10:15 pm
by bchaikin
It might be argued that without PER, there was no metric indicating how good he was.
embiid is blocking shots at 5.3% BS, rebounding at 14.0 reb/48min, scoring at 19.4 pts/g (30.9 pts/40min), shooting a 60% ScFG% (2s, 3s, and FTs), and is committing 3.7 TO/g...
the only other players to even approach those numbers in combination in the same season - 5.0%+ BS, 13-15 reb/48min, 19+ pts/g (25+ pts/40min), 55%+ ScFG%, and 3.0+ TO/g - include patrick ewing, alonzo mourning, hakeem olajuwon, and david robinson...
Re: A case where PER is NOT "good enough" to use
Posted: Wed Jan 11, 2017 1:09 pm
by Mike G
That exclusive list is similar to post # 3 in this thread.
Now that Noel is playing, the Sixers are even more top heavy:
Code: Select all
Sixers'17 pos G MP PER WS/48 BPM
Nerlens Noel C 10 141 24.1 .229 2.9
Joel Embiid C 25 626 23.3 .105 2.9
Richaun Holmes PF 21 337 17.3 .123 .9
Ersan Ilyasova PF 32 894 16.5 .111 .2
Jahlil Okafor C 28 628 14.8 .060 -2.8
Sergio Rodriguez PG 31 815 12.1 .031 -2.9
T.J. McConnell PG 35 760 11.9 .060 -3.3
Gerald Henderson SG 31 764 11.6 .072 -2.4
Robert Covington SF 32 966 10.1 .026 -.4
Nik Stauskas SG 33 871 9.9 .049 -2.1
Dario Saric PF 35 847 9.9 .001 -4.0
Hollis Thompson SG 31 561 9.3 .064 -2.8
At least by ranking on this team, PER is mostly in agreement with the other 2 stats.
Noel has
1 TO in 140 minutes -- about 1/17 of Embiid's rate -- and this makes him twice as good in WS.