PER...Joke Stat

[LAZERUSS]

Westbrook leading the NBA at 33.3, and AD right behind at 32.5.

Neither anywhere close to historic seasons, and yet both are well ahead of the all-time record.

Just another of the MANY "mini stats" that can be thrown out today. RAPM, ORtG, DRtG, WS, WS/48, OBPM, DBPM...all WORTHLESS stats.

Reminds me of some baseball's ridiculous statistics, like how a player has batted against a certain left-handed pitcher, in a day game, after the seventh inning, with runners in scoring position, and with a two-strike count.

[SpanishACB]

LAZERUSS...Joke Poster


tl;dr some arbitrary numbers argument about why Wilt turns him on more than current players. Watch out, if challenged, will resort to redneckery and more dumb **** number shit, approach thread with caution

[LAZERUSS]

Forgot + / -

http://www.basketball-reference.com/boxscores/201102030ORL.html

Gilbert Arenas....better than Lebron in that game.


BTW, you can go thru the NIGHTLY box-scores and come up with these results.

[SpanishACB]

BTW, you can go thru the NIGHTLY box-scores and come up with these results.

no one asked about how you're spending these xmas

[LAZERUSS]

LAZERUSS...Joke Poster


tl;dr some arbitrary numbers argument about why Wilt turns him on more than current players. Watch out, if challenged, will resort to redneckery and more dumb **** number shit, approach thread with caution


I understand. You have reading disability, and any post over 5 words is too difficult to comprehend.

Please, go back to "reading" your coloring books.

[Legends66NBA7]

ORTG and DRTG are really team stats. Along with TRB%, AST%, TS%, eFG%, TOV%, etc...

[GimmeThat]

Forgot + / -

http://www.basketball-reference.com/boxscores/201102030ORL.html

Gilbert Arenas....better than Lebron in that game.


BTW, you can go thru the NIGHTLY box-scores and come up with these results.


right.

that's like saying a right hook, is a joke arsenal

because it's all about the jabs and the body shot that wins the game

[LAZERUSS]

right.

that's like saying a right hook, is a joke arsenal

because it's all about the jabs and the body shot that wins the game

This post makes ZERO sense.

Again... + / -...just as RIDICULOUS as PER...

How about this one, from just last night...

http://scores.espn.go.com/nba/boxscore?gameId=400578725

Mike Miller, by FAR-AND-AWAY, the best player on the floor.

:bowdown:

[LAZERUSS]

Westbrook's "efficiency" this year...

His eFG% is at .492...in a league that is shooting .499.

His TS% is .566...in a league where the entire Clipper and Warrior TEAMs are shooting above that.

PER = JOKE STAT.

[HurricaneKid]

Westbrook leading the NBA at 33.3, and AD right behind at 32.5.

Neither anywhere close to historic seasons, and yet both are well ahead of the all-time record.

Just another of the MANY "mini stats" that can be thrown out today. RAPM, ORtG, DRtG, WS, WS/48, OBPM, DBPM...all WORTHLESS stats.

Reminds me of some baseball's ridiculous statistics, like how a player has batted against a certain left-handed pitcher, in a day game, after the seventh inning, with runners in scoring position, and with a two-strike count.

This post is disappoint.

[LAZERUSS]

AD in 2014-15:

35.1 mpg, 24.4 ppg, 10.4 rpg, 1.6 apg, 2.9 bpg, 1.7 spg, 1.3 tov, .567 FG%, .792 FT%, .567 eFG%, .618 TS%, 16.4 TRB%... 32.5 PER

Keep in mind that the league eFG% is also .499 and TS% is .538.


A past-his-peak KAJ in 1979-80:

38.3 mpg, 24.8 ppg, 10.8 rpg, 4.5 apg, 3.4 bpg, 1.0 spg, 3.6 tov, .604 FG%, .765 FT%, .604 eFG%, .639 TS%, 15.4 TRB%,... 25.3 PER

in a league with an eFG% of .486, and a TS% of .535.


32.5 to 25.3. Yep...it all makes sense now.

[LAZERUSS]

This post is disappoint.

Why? Because there is virtually nothing to support the significance of those "mini" stats?

[Dr.J4ever]

AD in 2014-15:

35.1 mpg, 24.4 ppg, 10.4 rpg, 1.6 apg, 2.9 bpg, 1.7 spg, 1.3 tov, .567 FG%, .792 FT%, .567 eFG%, .618 TS%, 16.4 TRB%... 32.5 PER

Keep in mind that the league eFG% is also .499 and TS% is .538.


A past-his-peak KAJ in 1979-80:

38.3 mpg, 24.8 ppg, 10.8 rpg, 4.5 apg, 3.4 bpg, 1.0 spg, 3.6 tov, .604 FG%, .765 FT%, .604 eFG%, .639 TS%, 15.4 TRB%,... 25.3 PER

in a league with an eFG% of .486, and a TS% of .535.


32.5 to 25.3. Yep...it all makes sense now.

Even the devil can quote the Bible, some say.

You can use these stats to prove a lot of things that aren't real. For example, EFG% is skewed towards this era because of the 3. It used to be that defense was measured by no. of buckets that go through the hoop, literally. EFG% will naturally increase towards this era because of the preponderance of 3 point shooting, but bucket for bucket there were more shots being made during the 70s and 80s.

So the moral of the story is----do not compare eras. Each era has a story to tell, and they stand on their own terms.

[ralph_i_el]

Why? Because there is virtually nothing to support the significance of those "mini" stats?

except math :confusedshrug:

[Ariza4three]

Javale Mcgee > Prime Wilt

[Blue&Orange]

Again... + / -...just as RIDICULOUS as PER...


lol +\- is 100x more ridiculous. +\- try to pass it as a individual stat when it

[LAZERUSS]

A peak KAJ in 70-71:

40.1 mpg, 31.7 ppg, 16.0 rpg, 3.3 apg, .577 FG%, .690 FT%, .577 eFG%, .606 TS%, 18.9 TRB%.

Keep in mind that the league eFG% was .449, and the league TS% was .500. So, Kareem just BLEW AWAY the league in terms of "efficiency."


PER... 29.0


Oh, and even his per/36...

28.4 ppg, 14.4 rpg, 3.0 apg.

Of course, per/36 is just another "mini" stat that basically punishes players who play more than that.


Oh, and KAJ (Alcindor) anchored the best defense in the league that year, as well.

[LAZERUSS]

except math :confusedshrug:

Go ahead and give us all here a mathematical formula which supports individual WS's. Or DWS's. One that makes complete sense.

[K Xerxes]

I think it can be fairly useful in specific situations. For example if we want to compare the offense of a group of players that play in the same position, provided they are in the same era. It will probably give a decent indication of who is 'better'. LeBron and Durant would come out on top of a SF comparison with others in the modern day for example.

Comparing across eras or even positions is pretty silly. As is using it to compare old time players who didn't have some stats like blocks or steals recorded, so their PER would be misleading.

[LAZERUSS]

[QUOTE=Blue&Orange]lol +\- is 100x more ridiculous. +\- try to pass it as a individual stat when it

[HurricaneKid]

This post makes ZERO sense.

Again... + / -...just as RIDICULOUS as PER...

How about this one, from just last night...

http://scores.espn.go.com/nba/boxscore?gameId=400578725

Mike Miller, by FAR-AND-AWAY, the best player on the floor.

:bowdown:

+/- has MAJOR variance issues. For instance, if BJ Armstrong ONLY plays with MJ its gonna look like he is the second best player in the world. Which is why you want to isolate who someone is playing with and against. This is RAPM. And you still want to have a larger data set. But saying RAPM is a garbage stat is borderline stupid.

PER is a lazy man's amalgam. And it tends to reward chuckers.

As to them not being historic

Here is a list of players that has avg 28/5.5/7/2 (WB 28.6/5.5/7.2/2.2)

88/89 MJ

Thats the list. One MJ season. YES. ITS HISTORIC.

Here is the list of players (besides AD this season) that has gone 24/10/2.75/1.7 on .567:

Thats the list. NO ONE. YES. ITS HISTORIC.

[SpanishACB]

so you're really sitting at home looking at all these stats, adding them up, comparing them, and then you go out of your way to make a thread that apparently talks about something else, but the moment people start laughing and making fun of you you're quick to bring kaj, wilt et al and go on with your constant ramblings...

is this really what you do on a daily basis?

[LAZERUSS]

+/- has MAJOR variance issues. For instance, if BJ Armstrong ONLY plays with MJ its gonna look like he is the second best player in the world. Which is why you want to isolate who someone is playing with and against. This is RAPM. And you still want to have a larger data set. But saying RAPM is a garbage stat is borderline stupid.

PER is a lazy man's amalgam. And it tends to reward chuckers.

As to them not being historic

Here is a list of players that has avg 28/5.5/7/2 (WB 28.6/5.5/7.2/2.2)

88/89 MJ

Thats the list. One MJ season. YES. ITS HISTORIC.

Here is the list of players (besides AD this season) that has gone 24/10/2.75/1.7 on .567:

Thats the list. NO ONE. YES. ITS HISTORIC.

How many players have had seasons of 24-24-8 .683, and likely with 10 bpg?

How many players have had seasons of 50-26?

Those are HISTORIC seasons.

Not a 24-10-3-2 .567 season. That is PALTRY in comparison.

[Dresta]

+/- has MAJOR variance issues. For instance, if BJ Armstrong ONLY plays with MJ its gonna look like he is the second best player in the world. Which is why you want to isolate who someone is playing with and against. This is RAPM. And you still want to have a larger data set. But saying RAPM is a garbage stat is borderline stupid.

PER is a lazy man's amalgam. And it tends to reward chuckers.

As to them not being historic

Here is a list of players that has avg 28/5.5/7/2 (WB 28.6/5.5/7.2/2.2)

88/89 MJ

Thats the list. One MJ season. YES. ITS HISTORIC.

Here is the list of players (besides AD this season) that has gone 24/10/2.75/1.7 on .567:

Thats the list. NO ONE. YES. ITS HISTORIC.
This, not to mention that their numbers will in all likelihood fall off at some point in the season - they will have slumps.

OP is the biggest joke on this site tbh.

[LAZERUSS]

so you're really sitting at home looking at all these stats, adding them up, comparing them, and then you go out of your way to make a thread that apparently talks about something else, but the moment people start laughing and making fun of you you're quick to bring kaj, wilt et al and go on with your constant ramblings...

is this really what you do on a daily basis?

Of course, you never bring ANYTHING into these discussions/ Why? Because you simply have NO KNOWLEDGE of the game.

Next...

[HurricaneKid]

A peak KAJ in 70-71:

40.1 mpg, 31.7 ppg, 16.0 rpg, 3.3 apg, .577 FG%, .690 FT%, .577 eFG%, .606 TS%, 18.9 TRB%.

Keep in mind that the league eFG% was .449, and the league TS% was .500. So, Kareem just BLEW AWAY the league in terms of "efficiency."


PER... 29.0


Oh, and even his per/36...

28.4 ppg, 14.4 rpg, 3.0 apg.

Of course, per/36 is just another "mini" stat that basically punishes players who play more than that.


Oh, and KAJ (Alcindor) anchored the best defense in the league that year, as well.

You don't understand how stats work.

Since we don't have blocks and steals from that age there are fewer data sets for Lew to dominate. So statistically we can't demonstrate his superiority. Using PER from an age where several of the stats weren't even kept and attempting to compare it to the current age is insane. PER is only meant to compare the same year's statistical contributions. CERTAINLY not to compare seasons from when stats weren't even kept.

Any time you have limited data the hypotheses you base on that data are going to be worse.

[La Frescobaldi]

Westbrook leading the NBA at 33.3, and AD right behind at 32.5.

Neither anywhere close to historic seasons, and yet both are well ahead of the all-time record.

Just another of the MANY "mini stats" that can be thrown out today. RAPM, ORtG, DRtG, WS, WS/48, OBPM, DBPM...all WORTHLESS stats.

Reminds me of some baseball's ridiculous statistics, like how a player has batted against a certain left-handed pitcher, in a day game, after the seventh inning, with runners in scoring position, and with a two-strike count.

it's always been kinda useless. i like the concept of advanced statistics, because in theory you can apply them to game strategy..... but nothing beats reflecting on and studying the opposing team, hustle, rebounding, and great defense.

Simple fundamentals are the hardest thing to coach, and the easy way out is to submerge the hard work under mountains of statistics.

Advanced stats are more overrated than stupidly high level scoring from a player, but it's only by a matter of degree.

[Blue&Orange]

We don't need a "formula." The basic stats, per era, are all we need to know just how dominant a player is/was.

This "effciency" crap is ridiculous. I can tell how "efficient" a player is by FG%, FT%, eFG%, and TS%. And NONE of those, by themselves, tells us ANYTHING about how good or dominant a player is/was on the offensive end of the court.

DeAndre Jordan is currently shooting .707 from the field. So what? He only takes 5.8 FGAs per game (and in 34 mpg mind you), and makes 4 dunks. He has virtually zero range, and zero post game. Does anyone honestly believe that if he took 20 FGAs per game, that he would still shoot .707 from the field?
Context? boxscore will tell you Lebron dominated in last year finals, everyone knows he stat padded in garbage time. care to make a rant against boxscores?

Do you realize that nobody but you have pointed out to Westbrook and Davis PER? I've seen a couple of Westbrook and Davis threads and all were focuses on how they were dominating games and not PER.

[Dresta]

How many players have had seasons of 24-24-8 .683, and likely with 10 bpg?

How many players have had seasons of 50-26?

Those are HISTORIC seasons.

Not a 24-10-3-2 .567 season. That is PALTRY in comparison.
Another agenda-driven thread about Wilt i see. Considering you are old enough to have seen the man play, you are remarkably child-like.

[LAZERUSS]

Another agenda-driven thread about Wilt i see. Considering you are old enough to have seen the man play, you are remarkably child-like.

And yet, I also post KAREEM's stats earlier in this topic.

[HurricaneKid]

How many players have had seasons of 24-24-8 .683, and likely with 10 bpg?

How many players have had seasons of 50-26?

Those are HISTORIC seasons.

Not a 24-10-3-2 .567 season. That is PALTRY in comparison.

Likely with 10bpg? You get that since there is no data the data sets aren't there right? While old timey folk might like to sit around drinking their Arnold Palmers and talk about how many blocks Wilt had 50 years ago you are discussing data that DOESN'T INCLUDE THAT INFORMATION.

You are yelling at the numbers for not including information that WAS NOT KEPT.

[Dr.J4ever]

Everyone here is making some good points. Laz is right though. No single stat or even a series of stats should be used to prove a point or points since they can easily be used or abused.

I'm not a fan of George W Bush, but if you remember his debate in 2000 with Al Gore, he used the term, "FUZZY MATH". That's what advanced stats can be, at times.

[Dr.J4ever]

Everyone here is making some good points. Laz is right though. No single stat or even a series of stats should be used to prove a point or points since they can easily be used or abused.

I'm not a fan of George W Bush, but if you remember his debate in 2000 with Al Gore, he used the term, "FUZZY MATH". That's what advanced stats can be, at times.

Let me qualify to say that I believe advanced stats have their place, if you use them in tandem by ALSO OBSERVING PLAYERS CLOSELY IN REAL TIME.

[La Frescobaldi]

Let me qualify to say that I believe advanced stats have their place, if you use them in tandem by ALSO OBSERVING PLAYERS CLOSELY IN REAL TIME.

Bobby Jones is a crystal clear example of advanced statistics being worse than useless.

[LAZERUSS]

You don't understand how stats work.

Since we don't have blocks and steals from that age there are fewer data sets for Lew to dominate. So statistically we can't demonstrate his superiority. Using PER from an age where several of the stats weren't even kept and attempting to compare it to the current age is insane. PER is only meant to compare the same year's statistical contributions. CERTAINLY not to compare seasons from when stats weren't even kept.

Any time you have limited data the hypotheses you base on that data are going to be worse.

The problem is...PER is somehow considered an actual RECORD. It doesn't come with an asterisk stating that...hey, AD's all-time RECORD season, came in a year in which he only played 35 mpg, and against an "x" percent league average in efficiency, etc, etc.

Furthermore, I could argue that even the all-time FG% RECORD, held by Chamberlain, would be MUCH higher in TODAY's era. His .683 in a league that shot an eFG% of .441, translates into .773 in today's NBA. And his .727 in a league that shot .456 translates into a .796 FG% in today's NBA. How come those figures aren't used then?

[Dr.J4ever]

Bobby Jones is a crystal clear example of advanced statistics being worse than useless.

Absolutely.

Bobby Jones would always seem to get that key block or key steal in the last 2 minutes of a close ball game, it was uncanny. He didn't always score a lot, even though at his peak with Denver, he was a pretty good scorer.

[La Frescobaldi]

Absolutely.

Bobby Jones would always seem to get that key block or key steal in the last 2 minutes of a close ball game, it was uncanny. He didn't always score a lot, even though at his peak with Denver, he was a pretty good scorer.

He undoubtedly had one of the worst +/- in the history of basketball.

from an old post............ the guys that really get it jammed to them are like Bobby Jones on the old Sixers and before that on the Denver Nuggets in the ABA and then he was a Gold-Miner for a season or maybe two in the NBA too.

He was a great basketball player - not good, not just All-Star - but his mentality was pure team. Whatever it takes to win. Make your teammates better. So he's basically unknown by anybody.

But what he would do, is go in a Philly-Boston tilt... when the Sixers were up by like 12 or 15 points but it was clear that the Doc had run out of gas and Boston was making one of those tremendous Celtics rallies.

Well there'd be Larry Bird going for one of his 40 point games, in the very fury of his early days.... but instead the Celtics rally would sputter, Bird would only get 32 on that game, McHale would only get 18 instead of 25, and Big Chief would get a technical for cheap-shotting Moses Malone because he wasn't smart enough to understand it was really the Bobby Jones double team that made him look like a fool.

So Bobby's +/- would look just like dirt but yet he had come off the bench, totally disrupted the Celtics rally, and kept the Sixers lead to 4 or 6 instead of going behind by 4 or 6.

You see? He would go in the game at the exact time where he would be made to look the absolute worst ever....

But yet there was no possibility of those Sixers teams getting to the Finals all those years without Bobby Jones.
Not 1 chance in 10,000 seasons of that lineup ever getting to the Finals without #24.... and who ever heard of him?

[LAZERUSS]

Likely with 10bpg? You get that since there is no data the data sets aren't there right? While old timey folk might like to sit around drinking their Arnold Palmers and talk about how many blocks Wilt had 50 years ago you are discussing data that DOESN'T INCLUDE THAT INFORMATION.

You are yelling at the numbers for not including information that WAS NOT KEPT.

I agree that doesn't include that information. If it did, he would likely have blocked 10 shots per game in '67.

As an example of his staggering bpg, we KNOW this much.

Julizaver has articles and research which will substantiate that Chamberlain had a KNOWN 570 blocks in 80 of his playoff games. Which is just HALF of them, and most of them are later in his career.

Two other posters here that have actually researched it, have claimed that Wilt averaged 5.4 bpg in his entire '72-73 season (his last BTW.) How significant is that? It was quite likely his LOWEST bpg season of his career. Furthermore, just 12 years later, Mark Eaton set the all-time record of 5.56.

We also KNOW that Chamberlain had a game with every block RECORDED (by SI) in which he blocked 23 shots (December 25th 1968.)

nbastats.net has Chamberlain with FOUR games of 30+ blocks, and another 15 more of 20+.

In their known career H2H totals, Chamberlain had FAR more blocks than Russell (I won't take the time to look them up now, but it was a staggering difference. Off the top of my head I know that he held a 29-8 margin in their '65 playoff series hH2H.)

I could go on, but there is simply no doubt that Chamberlain was the GOAT shot-blocker, and by a HUGE margin.

Incidently, blocking shots, and attempting to block shots, HURTS rebounding totals. Had Wilt just went for rebounds, his career rebounding totals would have been MUCH higher.

[SpanishACB]

Everyone here is making some good points. Laz is right though. No single stat or even a series of stats should be used to prove a point or points since they can easily be used or abused.



This blows my mind though.

Since that's all the old man does :biggums:

edit: look above

[ninephive]

Westbrook leading the NBA at 33.3, and AD right behind at 32.5.

Neither anywhere close to historic seasons, and yet both are well ahead of the all-time record.

Just another of the MANY "mini stats" that can be thrown out today. RAPM, ORtG, DRtG, WS, WS/48, OBPM, DBPM...all WORTHLESS stats.

Reminds me of some baseball's ridiculous statistics, like how a player has batted against a certain left-handed pitcher, in a day game, after the seventh inning, with runners in scoring position, and with a two-strike count.
I completely agree with you, especially on +/-. The other night Duncan was -34 and Baynes was +14.

[Dr.J4ever]

He undoubtedly had one of the worst +/- in the history of basketball.

from an old post............ the guys that really get it jammed to them are like Bobby Jones on the old Sixers and before that on the Denver Nuggets in the ABA and then he was a Gold-Miner for a season or maybe two in the NBA too.

He was a great basketball player - not good, not just All-Star - but his mentality was pure team. Whatever it takes to win. Make your teammates better. So he's basically unknown by anybody.

But what he would do, is go in a Philly-Boston tilt... when the Sixers were up by like 12 or 15 points but it was clear that the Doc had run out of gas and Boston was making one of those tremendous Celtics rallies.

Well there'd be Larry Bird going for one of his 40 point games, in the very fury of his early days.... but instead the Celtics rally would sputter, Bird would only get 32 on that game, McHale would only get 18 instead of 25, and Big Chief would get a technical for cheap-shotting Moses Malone because he wasn't smart enough to understand it was really the Bobby Jones double team that made him look like a fool.

So Bobby's +/- would look just like dirt but yet he had come off the bench, totally disrupted the Celtics rally, and kept the Sixers lead to 4 or 6 instead of going behind by 4 or 6.

You see? He would go in the game at the exact time where he would be made to look the absolute worst ever....

But yet there was no possibility of those Sixers teams getting to the Finals all those years without Bobby Jones.
Not 1 chance in 10,000 seasons of that lineup ever getting to the Finals without #24.... and who ever heard of him?

Good example. I mean, I loved the guy, and he was universally loved in Philly. We knew what was going on when Billy C. would put him on the best offensive player on the other team. We would always anticipate that key play somewhere defensively. He would deliver, more often than not.

He was all team, and yes Philly could not have won without him. When Doc was asked to name the top 5 players he had to go against, Jones was on that list.

[Dr.J4ever]

This blows my mind though.

Since that's all the old man does :biggums:

edit: look above

:lol Yes Laz does this at times. Hopefully he gets to realize the folly of it all.

[LAZERUSS]

I completely agree with you, especially on +/-. The other night Duncan was -34 and Baynes was +14.

THIS.

Why does the NBA even bother AT ALL, with this truly RIDICULOUS stat? It has absolutely ZERO value. In fact, it is clearly more of a "-" than a "+" in the grand scheme of NBA stats.

[SpanishACB]

THIS.

Why does the NBA even bother AT ALL, with this truly RIDICULOUS stat? It has absolutely ZERO value. In fact, it is clearly more of a "-" than a "+" in the grand scheme of NBA stats.

so tell us how do you rank stats in value?

where do you rank intangibles? #10? #15 behind FG%? what's the most important stat?

please do make a list of the top 10 stats in order of value

[LAZERUSS]

Chris Paul currently has the 6th highest career PER in NBA history...and above players like Kareem, Duncan, Hakeem, Magic, Bird, Oscar, West, Garnett...and WAY ahead of Russell (96th) and Pippen (108th.)

[LAZERUSS]

so tell us how do you rank stats in value?

where do you rank intangibles? #10? #15 behind FG%? what's the most important stat?

please do make a list of the top 10 stats in order of value


"intangibles?"

Tell me how you come up with that number?

[ralph_i_el]

How many players have had seasons of 24-24-8 .683, and likely with 10 bpg?

How many players have had seasons of 50-26?

Those are HISTORIC seasons.

Not a 24-10-3-2 .567 season. That is PALTRY in comparison.

historic seasons where all the big centers were averaging 15 rebounds a game :rolleyes:

Like we should give a crap about inflated rebound totals from that era.

If you sort individual players seasons by total rebounds, the first result that comes in this decade was Kevin love in 2010-11....and that's 91st on the list in best total rebounding seasons.

All stats on Wilt should be thrown out the window. Stats from that era are not comparable.

Ben Wallace is the only person to average more rebounds per game than Love after 2000. It was the 86th best rebounding season ever.

[Anaximandro1]

http://2.bp.blogspot.com/-9Q6sKxYXZw0/VJ7gxgMHvJI/AAAAAAAADnI/M4VJoYGpTiE/s1600/11.jpg


- PER summarizes a player's statistical accomplishments in a single number.

- PER represents statistical production per minute.

- PER is pace-adjusted -> players on a slow-paced team aren't penalized.

- Regular Season PER is useless.

- Playoff PER is strong indicator of dominance.

- It is really tough to compare PER across generations.

-The best players always put up dominant Playoff PER, relative to their era.


My advice :

- Combine performance in the NBA Finals with H2H, PER, USG% and ORtg/DRtg relative to their era.

[LAZERUSS]

historic seasons where all the big centers were averaging 15 rebounds a game :rolleyes:

Like we should give a crap about inflated rebound totals from that era.

If you sort individual players seasons by total rebounds, the first result that comes in this decade was Kevin love in 2010-11....and that's 91st on the list in best total rebounding seasons.

All stats on Wilt should be thrown out the window. Stats from that era are not comparable.

How about this then?

Russell was considered the second greatest rebounder of his era. In their 143 career H2H's, Wilt held a 92-43-8 margin. Including one game by a 55-19 margin. Overall, Wilt outrebounded Russell, per game, by a 28.7 to 23.7 rpg margin...or FIVE per game. And in their eight post-season series, Chamberlain outrebounded Russell in EVERY one of them...some by margins of 5, 6, and even 9 per game!

And how about this? Wilt played in 29 post-season series, and was the leading rpg in 28 of them (and he outrebounded his opposing starting center in ALL of them.) In the one in which he was not the leader, he was outrebounded by Jerry Lucas, 21-20 per game.

HOWEVER, the two would meet in the '72 Finals, and with Lucas as the starting center. A 35 year old Wilt, playing 47 mpg, outrebounded the 31 year old Lucas, who played 46 mpg...by a 23.2 rpg to 9.8 rpg margin.

THAT was his rebounding DOMINANCE.

[La Frescobaldi]

so tell us how do you rank stats in value?

where do you rank intangibles? #10? #15 behind FG%? what's the most important stat?

please do make a list of the top 10 stats in order of value

The best and most accurate statistic is to add the 3 major stats together and there you have it - what happened during a game, a season, or a career.
* scoring shows ability with the ball
* rebounding shows intelligence, desire, muscle power, algebra skills, and hustle.
* assists show game sharpness and teamwork.

If you wanna throw in the 2 minor stats too why by all means have fun with that.

[Dr.J4ever]

http://2.bp.blogspot.com/-9Q6sKxYXZw0/VJ7gxgMHvJI/AAAAAAAADnI/M4VJoYGpTiE/s1600/11.jpg


- PER summarizes a player's statistical accomplishments in a single number.

- PER represents statistical production per minute.

- PER is pace-adjusted -> players on a slow-paced team aren't penalized.

- Regular Season PER is useless.

- Playoff PER is strong indicator of dominance.

- It is really tough to compare PER across generations.

-The best players always put up dominant Playoff PER, relative to their era.


My advice :

- Combine performance in the NBA Finals with H2H, PER, USG% and ORtg/DRtg relative to their era.

This.

Not only is pace a major hindrance to comparing players from different eras, but you also have to factor in philosophies on offense and defense and rules changes, which PER doesn't account for.

So stick to PER for players of this generation and use it for scouting players.

[LAZERUSS]

The best and most accurate statistic is to add the 3 major stats together and there you have it - what happened during a game, a season, or a career.
* scoring shows ability with the ball
* rebounding shows intelligence, desire, muscle power, algebra skills, and hustle.
* assists show game sharpness and teamwork.

If you wanna throw in the 2 minor stats too why by all means have fun with that.

THIS. And to a lessor extent, FG%.

Unfortunately, we have no real accurate measurement of defense, or "intimidation." For instance, Chamberlain routinely held his opposing centers to a 10% lower FG% in their H2H's.

And we don't have enough block info, and certainly no "intimidation" number for the seasons before 1974.

[La Frescobaldi]

historic seasons where all the big centers were averaging 15 rebounds a game :rolleyes:

Like we should give a crap about inflated rebound totals from that era.

If you sort individual players seasons by total rebounds, the first result that comes in this decade was Kevin love in 2010-11....and that's 91st on the list in best total rebounding seasons.

All stats on Wilt should be thrown out the window. Stats from that era are not comparable.

Ben Wallace is the only person to average more rebounds per game than Love after 2000. It was the 86th best rebounding season ever.

Rebounds are a function of pace, to a great degree. But if instead of throwing out Wilt's era, throw out Wilt himself and see what it looks like.
If you are specifically talking about rebounding, throw out Russell too, who an all-time great rebounder living in a blistering fast League.

Sure the rebounding numbers will still be high - but you'll see that it is due to pace, not due to anything else.

[HurricaneKid]

The problem is...PER is somehow considered an actual RECORD. It doesn't come with an asterisk stating that...hey, AD's all-time RECORD season, came in a year in which he only played 35 mpg, and against an "x" percent league average in efficiency, etc, etc.

Furthermore, I could argue that even the all-time FG% RECORD, held by Chamberlain, would be MUCH higher in TODAY's era. His .683 in a league that shot an eFG% of .441, translates into .773 in today's NBA. And his .727 in a league that shot .456 translates into a .796 FG% in today's NBA. How come those figures aren't used then?

No. Its a same record the way Mark Eaton shows up far higher on the all time block list than Wilt/Russell, etc.

In coming years they will probably start tracking hockey assist. Who caused the defense to break down that led to the pass that led to the pass to the open shooter (as most good defenses can make one rotation in the worst of circumstances). When they incorporate THAT into PER this generation will be slighted. SO WHAT?

Yes, I would be right next to you if you argued the superiority of Wilt's .683 season was far better than DJ at .700 this year. Hell I'll be in front of you on that one.

[LAZERUSS]

No. Its a same record the way Mark Eaton shows up far higher on the all time block list than Wilt/Russell, etc.

In coming years they will probably start tracking hockey assist. Who caused the defense to break down that led to the pass that led to the pass to the open shooter (as most good defenses can make one rotation in the worst of circumstances). When they incorporate THAT into PER this generation will be slighted. SO WHAT?

Yes, I would be right next to you if you argued the superiority of Wilt's .683 season was far better than DJ at .700 this year. Hell I'll be in front of you on that one.

:cheers:

BTW, I might not always agree with you, but you are among the few posters here that I do respect.

I appreciate those that provide logic, research, facts, stats, etc., in their arguments and discussions.

[LAZERUSS]

MJ's '88-89 season...

40.2 mpg, 32.5 ppg, 8.0 rpg, 8.0 apg, 2.9 spg, .538 FG%, .850 FT%, .546 eFG%, .614 TS%, 1st Team All-Defense...

31.1 PER


Now THAT was truly a GREAT season.

[NBASTATMAN]

http://2.bp.blogspot.com/-9Q6sKxYXZw0/VJ7gxgMHvJI/AAAAAAAADnI/M4VJoYGpTiE/s1600/11.jpg


- PER summarizes a player's statistical accomplishments in a single number.

- PER represents statistical production per minute.

- PER is pace-adjusted -> players on a slow-paced team aren't penalized.

- Regular Season PER is useless.

- Playoff PER is strong indicator of dominance.

- It is really tough to compare PER across generations.

-The best players always put up dominant Playoff PER, relative to their era.


My advice :

- Combine performance in the NBA Finals with H2H, PER, USG% and ORtg/DRtg relative to their era.


WOW SOMEONE WITH A BRAIN.. :applause:


Per is a good stat but it is not the end all of stats. PER DOESN'T measure intangibles or a players mindset. It doesn't measure a player when the game is on the line either..

[Dr.J4ever]

WOW SOMEONE WITH A BRAIN.. :applause:


Per is a good stat but it is not the end all of stats. PER DOESN'T measure intangibles or a players mindset. It doesn't measure a player when the game is on the line either..

Yep, this is why stats can be very deceiving if you don't watch a player closely in real time.

[GimmeThat]

The player who sees himself and the player who sees others

and when one who sees others is far greater than how one sees itself



How do you measure the importance of a compound?

as much as how you would statistically measure:



A secret

[dubnation]

Westbrook with a +40% usage rate :wtf:. That would shatter Kobe's record from 2005-2006.

[HurricaneKid]

:cheers:

BTW, I might not always agree with you, but you are among the few posters here that I do respect.

I appreciate those that provide logic, research, facts, stats, etc., in their arguments and discussions.

Likewise Laz. I don't bother conversing with most folks around here.

[Shih508]

The biggest problem for PER is it rewarded players who play fewer minutes. That's why stars nowaday only play 32-35min instead of 36-40min before PER was used.

So when you use PER to measure players in the past it's pretty accurate to see how good a player performed. But when you measure it for any player who started playing after PER was out, it's always get twisted cuz players know what they can do or not do to get their PER higher. Like DWade, Lebron and etc.

So i think they should invent a metric = PER x How many minutes you play to get your total contribution to a team. So if player A with 30 PER but only play 15 min a night, would have 450 PERxMIN. And a player B with 20 PER but plays 35min a game, then he should get 700 PERxMIN. So when u comparing PER you would thought A is greater than B but when u comparing PERxMIN then B actually contribute more than A.

[Shih508]

MJ's '88-89 season...

40.2 mpg, 32.5 ppg, 8.0 rpg, 8.0 apg, 2.9 spg, .538 FG%, .850 FT%, .546 eFG%, .614 TS%, 1st Team All-Defense...

31.1 PER


Now THAT was truly a GREAT season.

This is like what i said. MJ had truly one of greatest individual season if not the greatest. But his PER got hurt cuz he played 40min a game.

Did MJ contribute less comparing to players like LeBron AD or Westbrook? NO, but he played too many min per game that hurts his PER

[Shih508]

http://2.bp.blogspot.com/-9Q6sKxYXZw0/VJ7gxgMHvJI/AAAAAAAADnI/M4VJoYGpTiE/s1600/11.jpg


- PER summarizes a player's statistical accomplishments in a single number.

- PER represents statistical production per minute.

- PER is pace-adjusted -> players on a slow-paced team aren't penalized.

- Regular Season PER is useless.

- Playoff PER is strong indicator of dominance.

- It is really tough to compare PER across generations.

-The best players always put up dominant Playoff PER, relative to their era.


My advice :

- Combine performance in the NBA Finals with H2H, PER, USG% and ORtg/DRtg relative to their era.

This is another example why PER is more accurate come playoff time. Players have to give all they got when it comes to playoff time so they cant just play only 30-35min like what they do in regular season.

PERxMPG is how you truly measure greatness

[Pushxx]

I think anybody with a brain already knows PER is an arbitrary stat.

[navy]

PER doesn't mean shit, however it is an 82 game stat.

Davis might keep it up, Westbrook wont.

[GimmeThat]

I think anybody with a brain already knows PER is an arbitrary stat.

Which is favored because of the high scoring nature of basketball, compared to other sports.

[DaRkJaWs]

This is like what i said. MJ had truly one of greatest individual season if not the greatest. But his PER got hurt cuz he played 40min a game.

Did MJ contribute less comparing to players like LeBron AD or Westbrook? NO, but he played too many min per game that hurts his PER
Rofl your comment about this hurting MJ goes even more so for Wilt. Playing 48+ minutes a game and still having the highest per ever, even excluding his double digit block average per game.

[fpliii]

LAZ - Have to strongly disagree with your inclusion of RAPM. Incredibly high predictive and explanatory value, derived using a peer-reviewed, universally accepted mathematical process, and some variant has been used by all the smarter front offices for over a decade.

In using it, one just has to be careful to keep in mind player role (play style and minutes played), and it's a terrific tool alongside watching games to evaluate and compare players IMO. Only problem is the digitalized play-by-plays used to derive it don't exist prior to 96-97 (though some of the other measures we have aren't bad...SRS with and without a guy, net on/off, etc.).

I don't bother with PER, win shares, individual ORtg/DRtg, and all the other tripe. All are based on limited information, and remove context.

[fpliii]

Bobby Jones actually rates extremely well in +/- and net on/off. I got a hold of the Harvey Pollack guides from 76-77 to the present awhile back, and a poster on another board was willing to compute net on/off from the raw +/-:


YEAR MIN ORTG DRTG NET
1978-79 2304 0,7 -2,7 3,4
1979-80 2125 5,3 -3,1 8,4
1980-81 2046 3,8 -7,0 10,8
1981-82 2181 1,1 -3,1 4,2
1982-83 1749 9,7 -1,3 11,0
1983-84 1761 5,5 -2,1 7,6
1984-85 1633 8,4 -2,0 10,4
1985-86 1519 1,1 -2,7 3,8

The poster drew comparison to Manu Ginobili based on the data, but noted that Jones seems more balanced between offense/defense.

[LAZERUSS]

LAZ - Have to strongly disagree with your inclusion of RAPM. Incredibly high predictive and explanatory value, derived using a peer-reviewed, universally accepted mathematical process, and some variant has been used by all the smarter front offices for over a decade.

In using it, one just has to be careful to keep in mind player role (play style and minutes played), and it's a terrific tool alongside watching games to evaluate and compare players IMO. Only problem is the digitalized play-by-plays used to derive it don't exist prior to 96-97 (though some of the other measures we have aren't bad...SRS with and without a guy, net on/off, etc.).

I don't bother with PER, win shares, individual ORtg/DRtg, and all the other tripe. All are based on limited information, and remove context.

Maybe you can explain it better...

http://www.gotbuckets.com/what-is-apm/


[QUOTE]One statistical method known to improve on raw APM is regularization, particularly ridge regression and/or Elastic Net regression. Standard Elastic Net regressions penalize any coefficients that are far away from 0. An alternative possibility is to use a Bayesian interpretation of this method, by providing the model an

[fpliii]

Maybe you can explain it better...

http://www.gotbuckets.com/what-is-apm/
Sure. Have any particular questions, or do you want me to just break it down?

[LAZERUSS]

Sure. Have any particular questions, or do you want me to just break it down?

Well, IMHO, it might as well have been in Bohemian. I'm not sure Einstein could have made sense out of it.

If you can conclusively prove to me the impact a player has, HIMSELF, while in the game, then by all means explain it.

If it is based on any approximations, or estimations,...forget it.


Let g(x,y) be a function such that:
I. −1<g(x,y)<1.
II.
ln(1+g(x,y)1−g(x,y) )+2ytan −1 (yg(x,y))=2(y 2 +1)x,
for x∈R,y>1.
Then
i. Show that g(x,y) is increasing in x.
ii. Find
lim y→∞ g(x,y).

iii. Show that g(x,y) is differentiable.
iv. Find lim y→∞ ∂∂x g(x,y).

All of which shows that Lebron is clearly better than Durant.

[buddha]

Westbrook nowhere close to historic season? he's averaging 29, 7, 6 with 2 steals.

[TheMarkMadsen]

Westbrook nowhere close to historic season? he's averaging 29, 7, 6 with 2 steals.

What's historic about that?

[ninephive]

Chris Paul currently has the 6th highest career PER in NBA history...and above players like Kareem, Duncan, Hakeem, Magic, Bird, Oscar, West, Garnett...and WAY ahead of Russell (96th) and Pippen (108th.)
PER has so much to do with your coaching lineups and rotations. For instance, because the Spurs have played Ginobili off the bench for several years, it will hurt guys like Parker's +/- significantly, because now you have an amazing bench who is going to destroy other team's second rotations, but is going to hurt your starting squad. However, the added gain on the second unit outweighs the decrease you get from your first unit and makes the Spurs a perpetual contender. It's a brilliant strategy, but the strategy makest he +/- and RAPM basically useless for comparison sake because other teams don't employ it (apples and oranges).

[chocolatethunder]

Westbrook leading the NBA at 33.3, and AD right behind at 32.5.

Neither anywhere close to historic seasons, and yet both are well ahead of the all-time record.

Just another of the MANY "mini stats" that can be thrown out today. RAPM, ORtG, DRtG, WS, WS/48, OBPM, DBPM...all WORTHLESS stats.

Reminds me of some baseball's ridiculous statistics, like how a player has batted against a certain left-handed pitcher, in a day game, after the seventh inning, with runners in scoring position, and with a two-strike count.
I think you're about as annoying as annoying can be but I'm not into PER either.

[Djahjaga]

Just because you don't understand or care to understand it doesn't mean it's not useful in the right context.

[fpliii]

Well, IMHO, it might as well have been in Bohemian. I'm not sure Einstein could have made sense out of it.

If you can conclusively prove to me the impact a player has, HIMSELF, while in the game, then by all means explain it.

If it is based on any approximations, or estimations,...forget it.
Well, basically that's what the goal is. It's still an approximation, but it's a damn good one. From year-to-year it's consistent, and it does a terrific job of modeling scoring margin. Regressions (the main process here is one) are used in science, finance, computing, economics among other fields, and are universally accepted in the mathematical community, as well as the scholarly communities at large that use them.

Basically, one starts with the play-by-play, and produces a list of all players on the floor, calculates how many possessions each side has had, and notes the points scored by each team. Every time there is a lineup change, there's a new row in the matchup file. There are between 30,000-35,000 5 vs 5 matchups over the course of an 82-game season plus the playoffs.

What we're trying to do is produce an estimator that will come closest to modeling the observations. Ordinary Least Squares (OLS) does so in the L2 sense. The problem is, there are some 400 players in the league each season, and nearly 100 times as many matchups. So the system equations is *severely* overdetermined. Ridge regression comes in handy here, as instead of trying to minimize something like (Ax-b)^2, we seek to minimize its sum with a second term, which helps improve the problem.

From the matchup file, we produce a data matrix consisting of 0's (meaning a player wasn't on the floor), 1's (on the floor for the home team), -1's (on the floor for the away team). Our dependent variable is the scoring margin of the matchups per 100 possessions. We weigh each of the 30,000-35,000 observations by the number of possessions (so if a certain unit sees the floor more often), and can proceed with calculations.

First we use cross-validation (testing estimates from one part of the sample against other parts of the sample, to find an optimal value) to find a lambda, which is the scalar used here in the equation which helps us identify solutions that have smaller error terms. Quality lambdas can be derived analytically, or using different computational methods (using software such as R, matlab, Stata, etc.). Once one has the lambda (when lambda=0, it's just OLS), direct computation of the coefficients is possible:

(X^T *(w*X) + lambda * I) ^ (-1) * (X^T *(w*y) + lambda*p0)

x is the matrix described above, w is the weights, I is the identity matrix of the appropriate dimensions, y is margin. The other term p0 corresponds to an optional prior argument. The text you posted above notes how that group computes its priors (which are guesses as to where a player will land in this year, based on information from previous years). I use a similar method (with a different aging term, and a different value than .85 for regression to the mean), but theirs is pretty good (they're a gambling syndicate, which makes a ton of money betting based on their estimates). When p0=0, it means we have no guesses for players, so they are all guessed to be 0 (league average) from the start.

Once we have the coefficients, we center the player values to zero based on weights, and depending on one's preference, throw out values for guys below a certain threshold (I don't consider guys below 375 minutes in a season, or 750 over one season and the previous one combined; if I'm using RAPM for player comparisons, I'll throw out any players who have played fewer than the equivalent of 24mpg over an 82 game season).

The offensive/defensive splits are computed slightly differently. Instead of margin for the y variable, we use a term called diffod (100*home points / home possessions - 100*away points / away possessions). We replace all the 1's with -1's, since scoring environment is a product of both teams (so there will be high positive diffod for instance if both teams have good offensive lineups, which are bad defensively). We then run the same regression as above. After this second regression is completed, we use the results of the two regressions to compute ORAPM ([diffod+margin]/2) and DRAPM ([diffod-margin]/2).

Generally we don't want to compare guys from season to season directly (since league environments are different), so we can calculate z-scores to do so (after removing players below the threshold). Multiplying the values by ~2 will give you the best predictive value, and multiplying by ~3 will help you best analyze performances while the season is still going on.

tl;dr - RAPM isolates player impact, taking into account quality of teammates and opponents. It's fine, as long as you note that a player's impact is tied to his role (so only players in similar roles should compared), and throw out observations over a small sample. It's impossible to get perfect estimates, but again, this is a process universally accepted in economics, science, and the like based on how precise it is. It's pretty much the highest quality estimate of player impact out there, which is why high-leverage gamblers and front offices of successful/smart teams use it as faithfully and often as they do (and it's stood the test of time; the original APM was released in 02 by Jeff Sagarin and Wayne Winston, the latter of whom worked for the Mavs for around a decade).

[fpliii]

Chris Paul currently has the 6th highest career PER in NBA history...and above players like Kareem, Duncan, Hakeem, Magic, Bird, Oscar, West, Garnett...and WAY ahead of Russell (96th) and Pippen (108th.)
Also just a note, PER can't be computed prior to 77-78 since stats are missing. They're just zeroed out on B-R, so Russell, Wilt, Thurmond, Hayes, and Kareem all have 0 blocks in calculations prior to 73-74, and guys like West, Oscar, Wilkens, Frazier have 0 steals for most of their careers (same with turnovers, so high usage ball-handlers who lost the ball over a ton aren't penalized).

[ralph_i_el]

Are all stats joke stats? I mean Larry Hughes and Nick Young averaged 22 and 18 ppg respectively, so is PPG a "Joke Stat" because those dudes scored more points than players who were way better scorers than them?

PER isn't made up, it's just a composite of stats, adjusted for the average pace and production of a year in the NBA. Nobody thinks it's perfect. It's just an easy stat to gauge a players production at a glance

And then OP goes comparing PER to eras where they didn't even record all the stats it takes into account :facepalm how stupid can you be?



Why bother comparing Wilt to modern day players anyway? He played in a league with no real international talent, and without the money to draw the best athletes in sports or motivate as many people to take basketball seriously as a career. Players smoked cigs in the locker rooms for ****s sake. They didn't even show the finals live....because not many people cared.

I don't compare Wilt to modern day players for the same reason Babe Ruth isn't in my top-10 baseball players. The level of competition just isn't even close to comparable. I'm ready for your pre-fab response to this laz. save yourself the trouble

[STATUTORY]

^^^RAPM is definitely valid for comparing player contributions on the same team playing in the same position

but beyond that there is severe selection issues. Lineups are not randomly generated and you simply never observe certain players on the court together and definitely those from different teams unless there is team change via trade or something.

I'm sure the predictive power is decent but identification is problematic

[fpliii]

^^^RAPM is definitely valid for comparing player contributions on the same team playing in the same position

but beyond that there is severe selection issues. Lineups are not randomly generated and you simply never observe certain players on the court together and definitely those from different teams unless there is team change via trade or something.

I'm sure the predictive power is decent but identification is problematic
Lineups aren't randomly generated, which is why you need to take role into account (especially for lower minute guys who only come in for defensive purposes/scoring against a second unit). Since the regularization is performed on all players simultaneously, no need for all players to play with/against all others. There's an abundance of lineup data (again, 30,000+ matchups), which is nearly 100x the number of players in the league.

I think we'll have to agree to disagree. I have no issue using RAPM for both its predictive and explanatory (which I think is the same thing as identification here) value. It's important to again remember that a player's impact is tied to his role.

RAPM (and when its not available, net on/off, or SRS with/without) is pretty much the only metric I use when evaluating players. I can ascertain whatever I need to know from the impact numbers, and from watching as much tape as possible (though I'm a fan of player synergy-type splits, and shot charts as well).

[Trollsmasher]

PER was alright when Wilt held the record:lol


What's historic about that?

He is doing it in 31.5 mpg

[STATUTORY]

Lineups aren't randomly generated, which is why you need to take role into account (especially for lower minute guys who only come in for defensive purposes/scoring against a second unit). Since the regularization is performed on all players simultaneously, no need for all players to play with/against all others. There's an abundance of lineup data (again, 30,000+ matchups), which is nearly 100x the number of players in the league.

I think we'll have to agree to disagree. I have no issue using RAPM for both its predictive and explanatory (which I think is the same thing as identification here) value. It's important to again remember that a player's impact is tied to his role.

RAPM (and when its not available, net on/off, or SRS with/without) is pretty much the only metric I use when evaluating players. I can ascertain whatever I need to know from the impact numbers, and from watching as much tape as possible (though I'm a fan of player synergy-type splits, and shot charts as well).

about 300 players in the league, lets say theres 60 player of each position, how many possible 10 men lineups are there? gotta be in the billions, the fact we only have 30,000 in actual sample just shows how much omitted data there is. for OLS to be consistent you need a random sample of that population of all possible 10 men lineups. The actual in-game lineups you have which are strategically chosen lineups from actual games make me hard pressed to believe they are random. This is a question of the sample and data, maybe ridge regression corrects for the selection somehow but I haven't really thought about its properties. Maybe the regularization mitigates the issue

But I agree this is the best possible metric for a dynamic game like basketball. Just playing devil's advocate

[fpliii]

about 300 players in the league, lets say theres 60 player of each position, how many possible 10 men lineups are there? gotta be in the billions, the fact we only have 30,000 in actual sample just shows how much omitted data there is. for OLS to be consistent you need a random sample of that population of all possible 10 men lineups. The actual in-game lineups you have which are strategically chosen lineups from actual games make me hard pressed to believe they are random. This is a question of the sample and data, maybe ridge regression corrects for the selection somehow but I haven't really thought about its properties. Maybe the regularization mitigates the issue

But I agree this is the best possible metric for a dynamic game like basketball. Just playing devil's advocate
That's why I noted RMSE (projected SD of the population), which is about .5 SDs of RAPM.

Nothing wrong with playing Devil's Advocate though. :cheers: I think asking the right questions is important, and a big part of the peer-reviewing process.

[PejaTheSerbSnip]

if theyre closely correlated to, or help explain wins, then no, theyre not joke stats. for example, Wins Produced does an excellent job accounting for how teams come about winning games. Same as Win Shares to an extent. Isn't that the chief purpose of any stat? To explain how games are won?

PER, however, is a joke stat because it doesn't really penalize missed shots...which is just ridiculous lol.

[PejaTheSerbSnip]

This post makes ZERO sense.

Again... + / -...just as RIDICULOUS as PER...

How about this one, from just last night...

http://scores.espn.go.com/nba/boxscore?gameId=400578725

Mike Miller, by FAR-AND-AWAY, the best player on the floor.

:bowdown:


....wait.....what?

Its not a ridiculous stat. It tells you how their team performed with this player on the floor, and from there you draw your own conclusions.

Of course, there are always single-game outliers. Sometimes a player will be lighting it up but the rest of the players on the floor wont be able to make the necessary contributions to make sure the redlining players effort isn't for naught. I don't think any "statistician" would claim otherwise.


+/- isn't really meant to be an all-encompassing stat, its just one of many useful metrics.


This is like comparing LeBrons game 6 in Boston in 2012 to his game 6 in 2010 and saying "WOOOOOW, LEBRON HAD 10 ASSISTS IN 2010 COMPARED TO 5 IN 2012, WOOOOW, ASSISTS SAY LEBRON WAS BETTER IN THAT 2010 DEBACLE, ASSISTS ARE USELESS"


Notice the logical fallacy?

[PejaTheSerbSnip]

AD in 2014-15:

35.1 mpg, 24.4 ppg, 10.4 rpg, 1.6 apg, 2.9 bpg, 1.7 spg, 1.3 tov, .567 FG%, .792 FT%, .567 eFG%, .618 TS%, 16.4 TRB%... 32.5 PER

Keep in mind that the league eFG% is also .499 and TS% is .538.


A past-his-peak KAJ in 1979-80:

38.3 mpg, 24.8 ppg, 10.8 rpg, 4.5 apg, 3.4 bpg, 1.0 spg, 3.6 tov, .604 FG%, .765 FT%, .604 eFG%, .639 TS%, 15.4 TRB%,... 25.3 PER

in a league with an eFG% of .486, and a TS% of .535.


32.5 to 25.3. Yep...it all makes sense now.


You underrate turnovers, pace, and offensive rebounds.

It's maddening. Players are rarely extolled for doing one of the things that should be so essential to their basketball identity: getting and maintaining possession of the ball. Players who do that really well, like Dennis Rodman, are almost always comically underrated.



But anyways, back to Davis v Kareem.


Try to think here for a sec (not trying to be demeaning). PER is measured on a per-possession or per-minute basis (pretty sure it is the former but I can't 100% remember). There were more possessions in an average game in '79-'80 than there were in '14-'15....SO.....what sense is there in posting their respective RAW stats, rather than their rate stats??


Here are there per 100 possession stats:

2014-2015: Davis: 35.9 PPG, 15.0 RPG, 2.4 AST, 2.5 STL, 4.3 BLKS, 1.9 TO

Kareem 1979-1980: 29.8, 13.0, 4.3, 1.2, 4.1, 4.4



See? Davis has a GARGANTUAN edge on a per-possession basis. He smokes him. So yeah. Still have difficulty wrapping your head around it? And that only took two seconds to look up. Try to critically examine this stuff before composing these uninformed posts.


Mind you, I hate PER too. But you're hating it for the wrong reasons LOL, and your methodology is perplexing, considering you're using per-game stats to criticize a per-possession statistic.




EDIT: nvm, I just remembered it is per-minute. BUT it is pace-adjusted so its essentially the same thing, lol. Still a rate stat and thus using per-game stats to contra it is ridiculous.

[PejaTheSerbSnip]

A peak KAJ in 70-71:

40.1 mpg, 31.7 ppg, 16.0 rpg, 3.3 apg, .577 FG%, .690 FT%, .577 eFG%, .606 TS%, 18.9 TRB%.

Keep in mind that the league eFG% was .449, and the league TS% was .500. So, Kareem just BLEW AWAY the league in terms of "efficiency."


PER... 29.0


Oh, and even his per/36...

28.4 ppg, 14.4 rpg, 3.0 apg.

Of course, per/36 is just another "mini" stat that basically punishes players who play more than that.


Oh, and KAJ (Alcindor) anchored the best defense in the league that year, as well.


Possessions, possessions, possessions.


per-100, Kareem and AD are very, very close.


Look, I get it. You're incredulous. You don't totally buy into all of these advanced stats. That's fine. I don't either, really. Those who think that we will be able to quantify a basketball players performance using statistics, to a T, is delusional. It's not a stationary sport like baseball.

But, the problem here is, you're going about critiquing these stats all wrong. You have absolutely no idea how they work, or how they are calculated. So that automatically disqualifies you from being a reasoned voice on the subject. You don't see a layperson going up to a physicist and saying "I don't understand time dilation. I can't wrap my head around it. How the hell does this make sense? How can time move at a different rate on Saturn than on Earth? You're wrong. It doesn't. It can't. Look at your watch. Now look at my watch. They're moving at the same rate. Come on. You're out to lunch. It's all BS because I don't know how it works."



Yes, I realize that was a bit of a caricature. :oldlol: but my point stands.

[ralph_i_el]

I'm glad we've established that Laz is dumb as ****

[Shih508]

PER makes player who plays fewer minute better than players who go all out and give everything they have night in and night out.

[La Frescobaldi]

LAZ - Have to strongly disagree with your inclusion of RAPM. Incredibly high predictive and explanatory value, derived using a peer-reviewed, universally accepted mathematical process, and some variant has been used by all the smarter front offices for over a decade.

In using it, one just has to be careful to keep in mind player role (play style and minutes played), and it's a terrific tool alongside watching games to evaluate and compare players IMO. Only problem is the digitalized play-by-plays used to derive it don't exist prior to 96-97 (though some of the other measures we have aren't bad...SRS with and without a guy, net on/off, etc.).

I don't bother with PER, win shares, individual ORtg/DRtg, and all the other tripe. All are based on limited information, and remove context.

yes RAPM has a lot of potential tho it is OT from Lazer's PER. Several teams have been using it with good results.

SRS is skewed for teams, not very effective for gauging players - for example if a fellow has a deep thigh bruise and misses 6 games, and the team loses all 6...... he seems great....... until you realize he is on a 40 win team playing against some high 50 or 60W teams who you weren't beating even if you had Bill Walton in the post.

[La Frescobaldi]

What's historic about that?

it's early days but that's a phenomenal stat line in today's league. He's out there on the ragged edge right now

[fpliii]

yes RAPM has a lot of potential tho it is OT from Lazer's PER. Several teams have been using it with good results.

SRS is skewed for teams, not very effective for gauging players - for example if a fellow has a deep thigh bruise and misses 6 games, and the team loses all 6...... he seems great....... until you realize he is on a 40 win team playing against some high 50 or 60W teams who you weren't beating even if you had Bill Walton in the post.
I agree about PER, very limited use. Hollinger runs the show for Memphis now, but he's not limited to that in terms of his knowledge/skillset.

Agree that SRS is limited in comparison to RAPM. You need a larger sample for it to become reliable (I think 12-15 games is the threshold), and it doesn't tell us a huge amount. I like net on/off, but unfortunately we don't have anything prior to 93-94 (which doesn't help much, since that was my second year watching, so I already have opinions formed on those players).

[Trollsmasher]

PER makes player who plays fewer minute better than players who go all out and give everything they have night in and night out.
it's a per minute stat...

[Asukal]

We don't need a "formula." The basic stats, per era, are all we need to know just how dominant a player is/was.

And that's why you are an ignorant biased POS. :hammerhead:

[LAZERUSS]

You underrate turnovers, pace, and offensive rebounds.

It's maddening. Players are rarely extolled for doing one of the things that should be so essential to their basketball identity: getting and maintaining possession of the ball. Players who do that really well, like Dennis Rodman, are almost always comically underrated.



But anyways, back to Davis v Kareem.


Try to think here for a sec (not trying to be demeaning). PER is measured on a per-possession or per-minute basis (pretty sure it is the former but I can't 100% remember). There were more possessions in an average game in '79-'80 than there were in '14-'15....SO.....what sense is there in posting their respective RAW stats, rather than their rate stats??


Here are there per 100 possession stats:

2014-2015: Davis: 35.9 PPG, 15.0 RPG, 2.4 AST, 2.5 STL, 4.3 BLKS, 1.9 TO

Kareem 1979-1980: 29.8, 13.0, 4.3, 1.2, 4.1, 4.4



See? Davis has a GARGANTUAN edge on a per-possession basis. He smokes him. So yeah. Still have difficulty wrapping your head around it? And that only took two seconds to look up. Try to critically examine this stuff before composing these uninformed posts.


Mind you, I hate PER too. But you're hating it for the wrong reasons LOL, and your methodology is perplexing, considering you're using per-game stats to criticize a per-possession statistic.




EDIT: nvm, I just remembered it is per-minute. BUT it is pace-adjusted so its essentially the same thing, lol. Still a rate stat and thus using per-game stats to contra it is ridiculous.

Nolw you are being ridiculous.

Per Possession stat?

In 79-80 KAJ averaged 16.9 FGA and 5.8 FTA per game.

In 14-15 AD is averaging 16.8 FGA and 6.9 FTA per game.



Their numbers, ACROSS the BOARD are nearly identical, except KAJ is considerably MORE EFFICIENT against his era than AD has been against his. TRB% was nearly identical, as well...even tough you argue that is a "per possession" stat. Please.

Please, you guys need to get a hold of REALITY.

[Shih508]

it's a per minute stat...

ya, but ppl tend to think if someone has 25 PER and playing 30min is better than someone with 21 PER and playing 38min a game that's the problem.

I'd take a player who can give me 21 PER 38min a game every day of week!

[LAZERUSS]

And that's why you are an ignorant biased POS. :hammerhead:

Ok idiot...

here is what we KNOW about Chamberlain...

In his ERA, he was winning ppg titles by 20 ppg, rpg titles by 5 rpg, and FG% titles by 17%. Furthermore, in his 50 ppg season, he BLEW AWAY the league in ppg/per 36, with 37 ppg,while his nearest scoring competitor was at 27 ppg per 36.

Post-season? Depends which Wilt. Up thru his scoring prime, he was scoring 33 ppg, grabbing 27 rpg, handing out 4 apg, and outshooting the post-season league eFG% mark by 10%. Furthermore, he was CRUSHING his HOF starting centers in those games by HUGE margins.

Regular season, post-season. Didn't matter. NO ONE was as dominant as Wilt. And before you bring up scoring...how about rebounding, FG% efficiency, passing, defense, and blocked shots?

The man absolutely OBLITERATED the league's POST-SEASON in '67, when he "only" scored 22 ppg, and "only" 18 ppg in the Finals. Furthermore, he proved he score anytime he wanted to in that post-season with games of 41 and 37 in his first two playoff games (on staggering FG%'s BTW.) Against Russell in the EDF clincher, he poured in 22 points in the first half, when the game was still close. He "only" finished with 29 because his scoring was not needed in the second half, when they blew open the game.

In his '72 post-season run, by virtually ALL accounts, he outplayed a PEAK Kareem, and in KAJ's greatest season, despite being outscored. In fact, KAJ, who had shot .574 against the league, was so horrid, that in the last four pivotal games of that series, he shot...get this... .414 from the field. In the Finals, he absolutely owned the Knicks with a 19-23 .600 7 bpg series, which included an overwhelming 24-29-8 10-14 FG/FGA clinching game. Hell, a 35 year old Wilt had nearly as many rebounds, by himself, as the entire NY team (39.)

The rest of his post-season career. Taking pure trash teams to within an eyelash of beating the greatest dynasty in the history of the NBA on several occasions. And he murdered Russell in the process. How about a seven game series, in which he took his 40-40 team to a game seven, one point loss, against a peak Celtic team that had gone 62-18. And in that series he carpet-bombed a helpless Russell with a 30 ppg, 31 rpg, .555 FG% (in a post-season NBA that shot an eFG% of .429...and also outblocked Russell, 29-8.

You can go right down the list. Every series, he was either, a force, the best player on his team, or the best player on the floor. And he was the best center on the floor in all of them (sorry, but Reed's FMVP's were a joke.)

You want more?


Aside from Chamberlain, there have been 36 30-30 games in NBA history, and Russell is the leader of that group, with 7 (Bellamy and Thurmond are next with 3 each.)

How about Wilt? 132.


40-30 (or 30-40) games: Other than Wilt, the NBA has had 9 40-30 games, with Baylor being the only player to have 2.

Chamberlain? 73


50-30 games: Pettit and Baylor each with 1

Wilt? 32


60-20 games: Aside from Wilt, there have been four (Baylor with 3 and Shaq with 1)

Chamberlain? 28


60-30 games: Baylor with 1

Wilt? 8


40-40 games: There have been 8 in the history of the NBA, and Chamberlain had all of them.


50-40 games: Obviously, Wilt would be the only player to have ever have accomplsihed that feat, which he did 5 times.


70-30 games: Chamberlain has the only 2, 78-43 and 73-36 (against Bellamy.)


And how about this...Wilt had 32 career 60+ point games...the entire rest of the NBA in it's long history... 32 COMBINED.




Take a look at this, and keep in mind that this is only a PARTIAL list of his all-time NBA achievements, and his ACTUAL RECORDS are easily the high HUNDREDS, and most likely in the THOUSANDS.

http://en.wikipedia.org/wiki/List_of_career_achievements_by_Wilt_Chamberlain


PURE and UTTER DOMINATION. No one else comes close.

PER my ass.

[LAZERUSS]

I'm glad we've established that Laz is dumb as ****

What we HAVE established, is that you have provided ZERO knowledge into this topic.

:facepalm

[PejaTheSerbSnip]

Nolw you are being ridiculous.

Per Possession stat?

In 79-80 KAJ averaged 16.9 FGA and 5.8 FTA per game.

In 14-15 AD is averaging 16.8 FGA and 6.9 FTA per game.

Their numbers, ACROSS the BOARD are nearly identical, except KAJ is considerably MORE EFFICIENT against his era than AD has been against his.

Please, you guys need to get a hold of REALITY.


Huh??? Dude....re-read my post



You listed their per-game stats. If there were way more possessions yesteryear per game than there are today, don't you think a per-possession comparison makes sense?? Wouldn't you say that the 10-15 extra possessions Kareem was a part of per game goes a long way in explaining the discrepancy in their per-game averages? What aren't you getting?


although again, I do think PER *is* an incredibly flawed and all but useless stat. But the fact that it is a per-minute pace-adjusted stat (not per-possession, I got it confused) is something that it has going for it.


I'm sleep-deprived, would somebody mind taking this one for me? I literally do not understand how I could have articulated myself any clearer.

[LAZERUSS]

Well, basically that's what the goal is. It's still an approximation, but it's a damn good one. From year-to-year it's consistent, and it does a terrific job of modeling scoring margin. Regressions (the main process here is one) are used in science, finance, computing, economics among other fields, and are universally accepted in the mathematical community, as well as the scholarly communities at large that use them.

Basically, one starts with the play-by-play, and produces a list of all players on the floor, calculates how many possessions each side has had, and notes the points scored by each team. Every time there is a lineup change, there's a new row in the matchup file. There are between 30,000-35,000 5 vs 5 matchups over the course of an 82-game season plus the playoffs.

What we're trying to do is produce an estimator that will come closest to modeling the observations. Ordinary Least Squares (OLS) does so in the L2 sense. The problem is, there are some 400 players in the league each season, and nearly 100 times as many matchups. So the system equations is *severely* overdetermined. Ridge regression comes in handy here, as instead of trying to minimize something like (Ax-b)^2, we seek to minimize its sum with a second term, which helps improve the problem.

From the matchup file, we produce a data matrix consisting of 0's (meaning a player wasn't on the floor), 1's (on the floor for the home team), -1's (on the floor for the away team). Our dependent variable is the scoring margin of the matchups per 100 possessions. We weigh each of the 30,000-35,000 observations by the number of possessions (so if a certain unit sees the floor more often), and can proceed with calculations.

First we use cross-validation (testing estimates from one part of the sample against other parts of the sample, to find an optimal value) to find a lambda, which is the scalar used here in the equation which helps us identify solutions that have smaller error terms. Quality lambdas can be derived analytically, or using different computational methods (using software such as R, matlab, Stata, etc.). Once one has the lambda (when lambda=0, it's just OLS), direct computation of the coefficients is possible:

(X^T *(w*X) + lambda * I) ^ (-1) * (X^T *(w*y) + lambda*p0)

x is the matrix described above, w is the weights, I is the identity matrix of the appropriate dimensions, y is margin. The other term p0 corresponds to an optional prior argument. The text you posted above notes how that group computes its priors (which are guesses as to where a player will land in this year, based on information from previous years). I use a similar method (with a different aging term, and a different value than .85 for regression to the mean), but theirs is pretty good (they're a gambling syndicate, which makes a ton of money betting based on their estimates). When p0=0, it means we have no guesses for players, so they are all guessed to be 0 (league average) from the start.

Once we have the coefficients, we center the player values to zero based on weights, and depending on one's preference, throw out values for guys below a certain threshold (I don't consider guys below 375 minutes in a season, or 750 over one season and the previous one combined; if I'm using RAPM for player comparisons, I'll throw out any players who have played fewer than the equivalent of 24mpg over an 82 game season).

The offensive/defensive splits are computed slightly differently. Instead of margin for the y variable, we use a term called diffod (100*home points / home possessions - 100*away points / away possessions). We replace all the 1's with -1's, since scoring environment is a product of both teams (so there will be high positive diffod for instance if both teams have good offensive lineups, which are bad defensively). We then run the same regression as above. After this second regression is completed, we use the results of the two regressions to compute ORAPM ([diffod+margin]/2) and DRAPM ([diffod-margin]/2).

Generally we don't want to compare guys from season to season directly (since league environments are different), so we can calculate z-scores to do so (after removing players below the threshold). Multiplying the values by ~2 will give you the best predictive value, and multiplying by ~3 will help you best analyze performances while the season is still going on.

tl;dr - RAPM isolates player impact, taking into account quality of teammates and opponents. It's fine, as long as you note that a player's impact is tied to his role (so only players in similar roles should compared), and throw out observations over a small sample. It's impossible to get perfect estimates, but again, this is a process universally accepted in economics, science, and the like based on how precise it is. The error is ridiculously small, with a RMSE (SD of population as a whole) of around +/- .5 SDs of the coefficients. It's pretty much the highest quality estimate of player impact out there, which is why high-leverage gamblers and front offices of successful/smart teams use it as faithfully and often as they do (and it's stood the test of time; the original APM was released in 02 by Jeff Sagarin and Wayne Winston, the latter of whom worked for the Mavs for around a decade).

I'm sorry Fpilli, but this is all NONSENSE.

This does NOT tell me an INDIVIDUAL players true IMPACT on the FLOOR. ALL of it is dependent upon the play of teammates, and opposing teammates. Anything that involves + / - and TEAMMATES is pure speculation, and ZERO fact.

Whoever spent the time coming up with all of this gibberish...wasted not only a ton of their own time, but anyone else's who tried to understand it.

Again...sorry to break it to you.

[PejaTheSerbSnip]

I'm sorry Fpilli, but this is all NONSENSE.

This does NOT tell me an INDIVIDUAL players true IMPACT on the FLOOR. ALL of it is dependent upon the play of teammates, and opposing teammates. Anything that involves + / - and TEAMMATES is pure speculation, and ZERO fact.

Whoever spent the time coming up with all of this gibberish...wasted not only a ton of their own time, but anyone else's who tried to understand it.

Again...sorry to break it to you.



Why should advanced metrics be discarded just because *YOU* don't understand them?

How about *TRYING* to understand them? How about reading into them if you're so hell-bent on proving that they're bunk?


And if you can't, no biggie. But at least acknowledge that they're out of your purview, instead of using statistics to refute.....



(wait for it)








statistics.

[fpliii]

I'm sorry Fpilli, but this is all NONSENSE.

This does NOT tell me an INDIVIDUAL players true IMPACT on the FLOOR. ALL of it is dependent upon the play of teammates, and opposing teammates. Anything that involves + / - and TEAMMATES is pure speculation, and ZERO fact.

Whoever spent the time coming up with all of this gibberish...wasted not only a ton of their own time, but anyone else's who tried to understand it.

Again...sorry to break it to you.
I don't work in basketball, but my job entails performing regression analysis on a daily basis. Ridge regression is a peer reviewed mathematical process that's stood the test of time (>50 years).

I guess we'll have to agree to disagree. I respect your point of view generally, but there's a reason that teams, and people wagering and winning money betting on the NBA subscribe to RAPM to the extent that they do.

[LAZERUSS]

Huh??? Dude....re-read my post



You listed their per-game stats. If there were way more possessions yesteryear per game than there are today, don't you think a per-possession comparison makes sense?? Wouldn't you say that (the 10-15 extra possessions Kareem was a part of per game) goes a long way in explaining the discrepancy in their per-game averages? What aren't you getting?


although again, I do think PER *is* an incredibly flawed and all but useless stat. But the fact that it is a per-minute pace-adjusted stat (not per-possession, I got it confused) is something that it has going for it.


I'm sleep-deprived, would somebody mind taking this one for me? I literally do not understand how I could have articulated myself any clearer.

Their "per game" averages, in nearly the same mpg, with the same FGAs, FTAs, TRB%'s, etc., right down the line...EXCEPT, KAJ was MORE EFFICIENT against his peers than AD has been against his. And yet, AD's PER was 32.5 and KAJ's was 25.3.

BTW, no one in their right mind would tell that this was KAJ's greatest season, either. In his two greatest seasons, 70-71 and 71-72, his PERs were 29.0 and 29.9...and he was FAR more dominant than AD has been THIS year.

Again...PER = JOKE STAT.

[3ball]

I don't work in basketball, but my job entails performing regression analysis on a daily basis. Ridge regression is a peer reviewed mathematical process that's stood the test of time (>50 years).

I guess we'll have to agree to disagree. I respect your point of view generally, but there's a reason that teams, and people wagering and winning money betting on the NBA subscribe to RAPM to the extent that they do.
hi flpiii - pretty nice explanation of RAPM... i have a couple questions..

what are the practical applications of RAPM?

where can we find RAPM data?

[LAZERUSS]

I don't work in basketball, but my job entails performing regression analysis on a daily basis. Ridge regression is a peer reviewed mathematical process that's stood the test of time (>50 years).

I guess we'll have to agree to disagree. I respect your point of view generally, but there's a reason that teams, and people wagering and winning money betting on the NBA subscribe to RAPM to the extent that they do.

I probably respect your opinions more than anyone else here. But there are just way too many variables in that "formula." WAY too many. Again, any time there are teammates, and opposing teammates involved, it is flawed.

I remember the argument on Russell's impact on the defensive end. Year-after-year his TEAM's had the best defense in the league. Of course, he was playing 45 mpg in the vast majority of them, and didn't have a legitimate backup center in the majority of them, either. So, the reality was, we simply didn't know his TRUE defensive impact, even if his team's defensive scoring and efficiency rose in his absence.

I do KNOW, however, that his TEAMs didn't miss a beat OFFENSIVELY withOUT Russell. Their scoring actually rose in the game's he missed. Hell, they set a one-time NBA record of 173 points in a game he missed.

In any case...if TEAMMATES and OPPOSING TEAMMATES are included in any + / - "formula" it is not accurate.

[PejaTheSerbSnip]

Their "per game" averages, in nearly the same mpg, with the same FGAs, FTAs, TRB%'s, etc., right down the line...EXCEPT, KAJ was MORE EFFICIENT against his peers than AD has been against his. And yet, AD's PER was 32.5 and KAJ's was 25.3.

BTW, no one in their right mind would tell that this was KAJ's greatest season, either. In his two greatest seasons, 70-71 and 71-72, his PERs were 29.0 and 29.9...and he was FAR more dominant than AD has been THIS year.

Again...PER = JOKE STAT.


JESUS CHRIST......dude....



1. i ALREADY explained that it is pace-adjusted (more possessions = more chances to pad your raw stats = accounted for by PER, which explains why Kareems per-game stats are better yet AD's PER is markedly superior).
2. i AGREE that it is a joke stat.

Jesus f'n Christ, you're not even reading my posts. What are you on about? its like we're literally talking about different things.

I'm sorry, but it's hard to be patient with the way you're arguing. I'm normally not this abrasive.

[j3lademaster]

hi flpiii - pretty nice explanation of RAPM... i have a couple questions..

what are the practical applications of RAPM?

where can we find RAPM data?http://stats-for-the-nba.appspot.com/

Believe it's on espn as well now.

[fpliii]

hi flpiii - pretty nice explanation of RAPM... i have a couple questions..

what are the practical applications of RAPM?

where can we find RAPM data?
1) Well, the dependent variable we're modeling is impact (or more precisely, ability of an individual to impact scoring margin, adjusted for teammates and opponents faced). If you have a team, and want to project how well they'll perform, you look at the expected minutes for the rotation, and calculate throughout (with a scaled RAPM, which again is 2x the z scores, though IN-SEASON 3x might be better). Example...

Say you have 12 players on a roster. Taking the sum of the products of their expected minutes, and regressed RAPM (you can even use the numbers from the previous year directly if it's pre-season, though most, including myself, apply an aging factor and a regression to the mean term) will give you a good estimator for SRS. Check out the APBRmetrics forum (apbr.org/metrics/), a few posters there have used SRS to predict the playoffs (http://apbr.org/metrics/viewtopic.php?f=2&t=8555).

2) Tougher question, but if you're interested I can PM you my personal excel file (comprised of those found online, and years I've computed to fill the gaps). Most of those publishing results online on their websites are hired by teams, or making money off of the estimates.

[La Frescobaldi]

And that's why you are an ignorant biased POS. :hammerhead:

it is still absolutely correct, in regards to an individual, far more than any other measurement.

Although rapm has a lot of potential.

per is a blind alley, for the most part.

[fpliii]

I probably respect your opinions more than anyone else here. But there are just way too many variables in that "formula." WAY too many. Again, any time there are teammates, and opposing teammates involved, it is flawed.

I remember the argument on Russell's impact on the defensive end. Year-after-year his TEAM's had the best defense in the league. Of course, he was playing 45 mpg in the vast majority of them, and didn't have a legitimate backup center in the majority of them, either. So, the reality was, we simply didn't know his TRUE defensive impact, even if his team's defensive scoring and efficiency rose in his absence.

I do KNOW, however, that his TEAMs didn't miss a beat OFFENSIVELY withOUT Russell. Their scoring actually rose in the game's he missed. Hell, they set a one-time NBA record of 173 points in a game he missed.

In any case...if TEAMMATES and OPPOSING TEAMMATES are included in any + / - "formula" it is not accurate.
It's not quite a formula though. The formula just models the solutions.

We don't have RAPM for Russell (only goes back to 96-97), so I can't comment on him. But if his backup center is a scrub, he wouldn't get a bonus. The regression adjusts for not only teammates and opponents, but THEIR teammates and opponents.

Maybe I shouldn't have used the word "impact" on its own. Rather, ability to impact scoring margin might be more apt.

[La Frescobaldi]

I'm sorry Fpilli, but this is all NONSENSE.

This does NOT tell me an INDIVIDUAL players true IMPACT on the FLOOR. ALL of it is dependent upon the play of teammates, and opposing teammates. Anything that involves + / - and TEAMMATES is pure speculation, and ZERO fact.

Whoever spent the time coming up with all of this gibberish...wasted not only a ton of their own time, but anyone else's who tried to understand it.

Again...sorry to break it to you.

you should study that a little more Laz, it's being used almost ubiquitously and does have merit.
You can't take a player out of the league; he's part of his surroundings. The League is like the Force :lol ..... it surrounds us

[Asukal]

it is still absolutely correct, in regards to an individual, far more than any other measurement.

Although rapm has a lot of potential.

per is a blind alley, for the most part.

There is no single stat which is perfect. Everything has to be taken in context of the what actually happened in the game. Here are a few examples:

- player A shoots bad through 1-3 quarters making 2-10 then at the beginning of the 4th his team is down by 30. He continues playing and ends up at 10-18 by the end of the game, they lose the game by 20. Stats will say he played well shooting 10-18 but the truth is he stat padded against scrubs in garbage time.

- star player gets guarded by one of the best defenders in the league in a game. Said star player is so good offensively that he can create shots against anyone but this game he misses a lot of open shots. People who only look at stats will say the defender did a great job but the truth is the star player missed shots he normally makes.

Stats never tell the whole story. :confusedshrug:

[LAZERUSS]

There is no single stat which is perfect. Everything has to be taken in context of the what actually happened in the game. Here are a few examples:

- player A shoots bad through 1-3 quarters making 2-10 then at the beginning of the 4th his team is down by 30. He continues playing and ends up at 10-18 by the end of the game, they lose the game by 20. Stats will say he played well shooting 10-18 but the truth is he stat padded against scrubs in garbage time.

- star player gets guarded by one of the best defenders in the league in a game. Said star player is so good offensively that he can create shots against anyone but this game he misses a lot of open shots. People who only look at stats will say the defender did a great job but the truth is the star player missed shots he normally makes.

Stats never tell the whole story. :confusedshrug:

Over the course of an 82 game season, stats DO tell a HUGE part of the story.

If they didn't, the league wouldn't bother with them.

[Miss Bella]

A peak KAJ in 70-71:

40.1 mpg, 31.7 ppg, 16.0 rpg, 3.3 apg, .577 FG%, .690 FT%, .577 eFG%, .606 TS%, 18.9 TRB%.

Keep in mind that the league eFG% was .449, and the league TS% was .500. So, Kareem just BLEW AWAY the league in terms of "efficiency."


There is no need to mention Kareems eFG% in 1970 considering the three point shot still had not been introduced to the NBA.

in 1970 efg% = fg%

[LAZERUSS]

There is no need to mention Kareems eFG% in 1970 considering the three point shot still had not been introduced to the NBA.

in 1970 efg% = fg%

It is important from the standpoint that, if you are going to hold "pace" against teams from the pre-shot clock era, you HAVE to adjust for eFG%'s, as well. For instance, if you don't adjust for eFG%, but just purely pace, the 1962 NBA would average about 87 ppg in today's NBA. If you do adjust for eFG%, then it comes out to nearly the 2014 ppg average (slightly less because the league FT% in '62 was a little less than in '14.)

BTW, the 3pt shot theoretically opens up post scoring. The floor is spread much more. Kareem likely would have been considerably more efficient in '14 than he was in '71.

In any case, KAJ's .606 TS% in '71 was almost 11% higher than the league average. AD's .618 this season is in a league that has a TS% of .538, or about 8% higher.

[Asukal]

Over the course of an 82 game season, stats DO tell a HUGE part of the story.

If they didn't, the league wouldn't bother with them.

I agree. That's why 30>22>18 is the only stat I need to trash you over and over and over. :rolleyes:

[Dr.J4ever]

It is important from the standpoint that, if you are going to hold "pace" against teams from the pre-shot clock era, you HAVE to adjust for eFG%'s, as well. For instance, if you don't adjust for eFG%, but just purely pace, the 1962 NBA would average about 87 ppg in today's NBA. If you do adjust for eFG%, then it comes out to nearly the 2014 ppg average (slightly less because the league FT% in '62 was a little less than in '14.)

BTW, the 3pt shot theoretically opens up post scoring. The floor is spread much more. Kareem likely would have been considerably more efficient in '14 than he was in '71.

In any case, KAJ's .606 TS% in '71 was almost 11% higher than the league average. AD's .618 this season is in a league that has a TS% of .538, or about 8% higher.

I guess TS% would be a better stat to compare eras, if you insist. I don't feel comfortable using EFG% since it skews it towards this era with the 3ball.

We all know good ole FG% was much higher in the 70s and 80s than it is today. As I said, more buckets were scored back then on a higher percentage. EFG% only skews and deceives people into thinking defenses were similar.

[Dr.J4ever]

Bobby Jones actually rates extremely well in +/- and net on/off. I got a hold of the Harvey Pollack guides from 76-77 to the present awhile back, and a poster on another board was willing to compute net on/off from the raw +/-:


YEAR MIN ORTG DRTG NET
1978-79 2304 0,7 -2,7 3,4
1979-80 2125 5,3 -3,1 8,4
1980-81 2046 3,8 -7,0 10,8
1981-82 2181 1,1 -3,1 4,2
1982-83 1749 9,7 -1,3 11,0
1983-84 1761 5,5 -2,1 7,6
1984-85 1633 8,4 -2,0 10,4
1985-86 1519 1,1 -2,7 3,8

The poster drew comparison to Manu Ginobili based on the data, but noted that Jones seems more balanced between offense/defense.

Wow, I didn't check the numbers but I suspected enough that Jones would probably have a good plus minus. I was agreeing with La Fresco about his impact to the game more than anything.

[tpols]

Flpii, could you think of any reason Pau Gasol has a negative offensive and defensive rapm this year when eye test tells us he's having very good impact? Also Dwayne Wade.. Good individual stats but small rapm. I especially can't figure Pau numbers out.

[fpliii]

Flpii, could you think of any reason Pau Gasol has a negative offensive and defensive rapm this year when eye test tells us he's having very good impact? Also Dwayne Wade.. Good individual stats but small rapm. I especially can't figure Pau numbers out.
I haven't seen any published RAPM for this season yet, and don't have the matchup files to compute it myself. ESPN's RPM includes box score stats to form a SPM as a prior, and could include last year as a prior (J.E. typically did so with his xRAPM, which was the precursor to RPM; a poster on this board noted that that might have changed this year, but I don't know one way or the other).

There's also likely a small sample issue. If published, pure RAPM (without box score elements) by the end of the year still underrates both Pau and DWade, then we can say it's an issue of priors. That's why it's important to look at both prior-informed and NPI RAPM alongside one another.

I think talkingpracticeblog noted they'll be publishing their first RAPM for the year on gotbuckets.com relatively soon, so it's probably worth waiting until then before identifying a difference between what our eyes are telling us and what the RAPM data is.

[SHAQisGOAT]

http://2.bp.blogspot.com/-9Q6sKxYXZw0/VJ7gxgMHvJI/AAAAAAAADnI/M4VJoYGpTiE/s1600/11.jpg


- PER summarizes a player's statistical accomplishments in a single number.

- PER represents statistical production per minute.

- PER is pace-adjusted -> players on a slow-paced team aren't penalized.

- Regular Season PER is useless.

- Playoff PER is strong indicator of dominance.

- It is really tough to compare PER across generations.

-The best players always put up dominant Playoff PER, relative to their era.


My advice :

- Combine performance in the NBA Finals with H2H, PER, USG% and ORtg/DRtg relative to their era.

:applause: Good stuff. PER has many "flaws" (when you can't use it in context and make it even close to something like the ultimate stat) and can't even be compared across eras.

[SHAQisGOAT]

Bobby Jones actually rates extremely well in +/- and net on/off. I got a hold of the Harvey Pollack guides from 76-77 to the present awhile back, and a poster on another board was willing to compute net on/off from the raw +/-:


YEAR MIN ORTG DRTG NET
1978-79 2304 0,7 -2,7 3,4
1979-80 2125 5,3 -3,1 8,4
1980-81 2046 3,8 -7,0 10,8
1981-82 2181 1,1 -3,1 4,2
1982-83 1749 9,7 -1,3 11,0
1983-84 1761 5,5 -2,1 7,6
1984-85 1633 8,4 -2,0 10,4
1985-86 1519 1,1 -2,7 3,8

The poster drew comparison to Manu Ginobili based on the data, but noted that Jones seems more balanced between offense/defense.

Damn, great stuff man. Can you send me those logs and also point me to that thread? Thanks

Bobby is pretty underrated/overlooked these days and it's just a shame that he's not in the HoF already (looking at some players already in).

He did everything for the sake of the team, even many "little thing" you won't find in box-scores; one of the best and most versatile defenders to ever play the game, got steals/blocks, drew charges, great at guarding many different types of players, good rebounder, decent passer, didn't score much but contributed some on very high FG%, hustled like crazy, always went about his business, clutch...

Even with occasional epileptic seizures, a chronic heart disorder and asthma (why he couldn't play major minutes on a regular basis) he managed to show great athleticism and became a big-time player, one of the best role players ever (when/if he wasn't more than that).

Plus, he even "gets his due" when looking at (different) advanced stats... He's also 2nd all-time in NBA all-defensive 1st teams (8 in 10 seasons), would've most likely been DPOY in 1977 if the award was already given, he's the 1st player to receive 6th-man, 4x all-star, champion as a solid contributor to one of the greatest teams ever...
Again, shameful that he's not a HoF'er by now.

[3ball]

-The best players always put up dominant Playoff PER, relative to their era.


Lebron's PER in the "Shaq/Duncan" era: 23.8 over 3 seasons plus 2 missed playoffs

[LAZERUSS]

JESUS CHRIST......dude....



1. i ALREADY explained that it is pace-adjusted (more possessions = more chances to pad your raw stats = accounted for by PER, which explains why Kareems per-game stats are better yet AD's PER is markedly superior).
2. i AGREE that it is a joke stat.

Jesus f'n Christ, you're not even reading my posts. What are you on about? its like we're literally talking about different things.

I'm sorry, but it's hard to be patient with the way you're arguing. I'm normally not this abrasive.

Look, you are way over-rating "pace" here.
I have never believed the "pace" numbers that BB-Reference gives teams from the '60's and 70's, so I prefer to use FGAs and FTAs.

Here are the comparisons from KAJ's 79-80 season, and AD's current season.

79-80: 90.6 FGAs, 27.8 FTAs
14-15: 83.4 FGAs, 23.5 FTAs

Think about those for a moment.

A TOTAL, per TEAM, of 7.2 more FGAs, and 4.3 more FTAs.

Then, AD has played a little less than 75% of his team's minutes.

So, reduce 7.2 down to 75%... 5.4 more FGAs.

Davis has not quite taken 20% of his team's FGAs. So, reduce 5.4 down to AD's prcentage of FGAs taken...and it is slightly over ONE more FGA per game, than what he has taken so far this season. Same with FTAs. By the time you break it down, AD takes ONE more FTA.

So, AD's FGAs and FTAs, "pace adjusted" now give him 17.8 FGAs per game, and 7.6 FTAs per game. Multiple by .566 and .795...and he now scores a little more than a ppg. 24.6 ppg to about 26.0 ppg.

So, in slightly less mpg (38.3 for KAJ, to 35.3 for AD)...

KD: 26.0 ppg to 24.8 ppg. Most of the rest of their numbers are too small for "pace" to affect them much at all. The only other relevant number is rpg. KAJ's '80 season averaged 44.9 rpg, KD's '15 season is currently averaging 42.9 rpg. Again, insignificant (both are 10.2 rpg.) In any case, KAJ's TRB% was 15.4, to AD's 16.4. Not much of a difference.

Again, all their other numbers are about even across the board (AD does average about 2 TOVs less per game)...EXCEPT, KAJ outshoots the league eFG%, and the league TS% by a somewhat larger margin.


With those tiny margins...and across the board... KD's PER is 32.5 to KAJ's 25.3. It STILL doesn't add up.

Furthermore, how do we know that an AD, playing slightly more minutes, and at a slightly faster pace (and slightly is the key word)...would maintain his "efficiency?"


And yes, I understand that you agree that PER is a joke stat.

[LAZERUSS]

I agree. That's why 30>22>18 is the only stat I need to trash you over and over and over. :rolleyes:

Talk about using "cherry-picked" stats.

Wilt played in 1045 regular season games, 160 playoff games, and 35 Finals games.

Only 52 of those 160 playoff games came in his "scoring seasons", and in those he averaged 33 ppg. Only SIX Finals games came in his "scoring seasons" and in those he averaged 29.2 ppg. Oh, and in those 52 playoff game... 30 against RUSSELL. Oh, and in those SIX Finals GAMES...ALL against Russell.

His ACTUAL "decline" in scoring in the years his team's made the Finals was from 23.0 ppg down to 18.6 ppg..and ALL were against HOF centers (with 19 of the 35 games against Russell and Thurmond.)

Of course, he ELEVATED his rebounding, and his FG% against post-season league average was a staggering 10% higher (and about 12% in his Finals.) And overall, he crushed his opposing centers in scoring, rebounding, and FG% in his post-season H2H's. (Just as an example...in his six Finals series, he outshot his opposing centers, ALL of whom are in the HOF, by a .559 to .439 margin from the field.)

One more time...a PRIME "scoring" Wilt, even into his '67 post-season...in 67 playoff games (41 of which were against Russell and Thurmond)...

30.4 ppg, 27.0 rpg, 4.5 apg, .515 (in post-season NBA's that shot about .430 on average), and most likely (more like most certainly) 8 bpg.

Now, I ask you...give me a list of GOAT players who averaged a 30-27-5-8 .515 (actually how about 9% above league eFG% average), over the course of even ONE SERIES? Hell, give me the list of GOATs who had ONE PLAYOFF GAME with a 30-27-5-8 .515 stat-line. Oh, and holding their opposing center to probably about 10% below their normal regular season FG%, too.

And how about CLUTCH?

You love scoring...guess who has the THIRD highest PPG in "must-win" post-season games? Of course you know...WILT (31.1 ppg, and just behind Lebron's 31.9 and MJ's 31.3.) Incidently, Chamberlain shot a much higher FG%, than either MJ or Lebron, too. And not only that, against league average eFG%...it was LIGHT YEARS higher.

50+ point games? MJ had eight...and guess who is next? Of course, you already know it was WILT, with FOUR. BTW, who has the ONLY THREE 50+ point games in "must-win" games? Yep...WILT.

Rebounding? Wilt played in 29 post-season series. NEVER outrebounded by his opposing starting center in ANY of them. And was only outrebounded in ONE series (and that was by a 21-20 rpg margin.) Now I ask you...how many other GOATs can claim that they outrbounded everyone they faced?

Defense? Again, holding his opposing centers to WAY below their normal FG%'s. Just ask a PEAK Kareem. In his '71 regular season, Kareem shot .577 against the league. Against Wilt in the '71 WCF's... .481. In his '72 regular season, KAJ shot .574 against the league. Against Wilt in the '72 WCF's... .457 (and only .414 in the last four pivotal games of that series.) So, overall, in his playoff H2H's with an aging Chamberlain, a PEAK Kareem shot about an 11% lower FG% than what he had shot against the league in his regular seasons.

The MOST DOMINANT player of all-time.

:bowdown:

[Asukal]

Talk about using "cherry-picked" stats.

Cherry picked? That's his career total. :oldlol: :roll: :facepalm

[LAZERUSS]

I guess TS% would be a better stat to compare eras, if you insist. I don't feel comfortable using EFG% since it skews it towards this era with the 3ball.

We all know good ole FG% was much higher in the 70s and 80s than it is today. As I said, more buckets were scored back then on a higher percentage. EFG% only skews and deceives people into thinking defenses were similar.

They weren't higher in the 70's.

Which is interesting. A PRIME KAJ had seasons of .539, .529, .518, and .513 in the decade of the 70's (with the .513 smack in the middle of them.) His lowest FG% in the 80's (before his last two seasons at ages 40 and 41) was .564. And his highest was .604 (as well as a .599 season at ag 37.)

Then, think about this. A PEAK Kareem, in 70-71 and 71-72, averaged a combined 33 ppg on a .563 FG% against the league. In his five post-season series against an aging Thurmond and an aging Wilt... a combined 26 ppg on a .456 FG%. A HUGE decline.

Now, how about this. A 38-39 year old KAJ, in his 10 straight H2H games with a 23-24 year old Hakeem... 33 ppg on a .621 FG%. Included were THREE games of 40+, with a high of 46 (and in only 37 minutes, and on a 21-30 FG/FGA.) Even at age 40, and without Sampson in the lineup, KAJ outscored a 25 year old Hakeem in their four H2H's, and outshot him by a .563 to .403 margin. Overall, in his 23 career H2H's with Hakeem, he outscored Olajuwon, and outshot from the field by a .607 to .512 margin. And, he outscored Ewing in his carer H2H's, while outshooting Patrick by a .567 to .446 margin.

I am not pointing this at you...but it certainly is food for thought for those that somehow claim that there was no defense being played in the 60's and 70's. An ancient Kareem was carpet-bombing two of the best defensive centers of the 90's.

[LAZERUSS]

Cherry picked? That's his career total. :oldlol: :roll: :facepalm

Yes, "cherry-picked".

You are using 10% of his games, most all of which came against HOF centers, and a huge amount of those, against Russell.

CHERRY PICKED, and more importantly, NO CONTEXT.

A much more accurate measurement...COMBINE regular season games, which were usually about 80, with post-season games, which, for almost every player was between 3 to 20 games, and most often around 10.

How do you think Wilt's numbers would look then? Especially a PRIME "scoring" Wilt?

Of course, even using your "cherry-picked" 30-22-18, which BTW, was actually 30-23-19...Wilt missed the playoffs in his 45 ppg season in '63...and wasn't in the Finals in '60, 61, '62, '65, '66, '68, and '71. Furthermore, he was "post-injury" in '70, '71, '72, and '73.

Let's use his 50 ppg regular season, as an example. Combine his 80 regular season games, with his 12 post-season games...and guess what, his ppg average falls all the way down to... 48.4 ppg.

You can do that with any of his seasons. His "decline" will be minimal.

Now, go ahead and give me your take on his FULL SEASONS, INCLUDING the post-season.

Furthermore, why do you ignore his ELEVATING his rebounding? Or the fact that he ousthot the post-season league eFG%'s by huge margins? Or the fact that he was slaughtering his HOF peers by staggering margins in ppg, rpg, apg, and FG%?

CHERRY PICKED.

[Asukal]

Yes, "cherry-picked".

You are using 10% of his games, most all of which came against HOF centers, and a huge amount of those, against Russell.

CHERRY PICKED, and more importantly, NO CONTEXT.

A much more accurate measurement...COMBINE regular season games, which were usually about 80, with post-season games, which, for almost every player was between 3 to 20 games, and most often around 10.

How do you think Wilt's numbers would look then? Especially a PRIME "scoring" Wilt?

Of course, even using your "cherry-picked" 30-22-18, which BTW, was actually 30-23-19...Wilt missed the playoffs in his 45 ppg season in '63...and wasn't in the Finals in '60, 61, '62, '65, '66, '68, and '71. Furthermore, he was "post-injury" in '70, '71, '72, and '73.

Let's use his 50 ppg regular season, as an example. Combine his 80 regular season games, with his 12 post-season games...and guess what, his ppg average falls all the way down to... 48.4 ppg.

You can do that with any of his seasons. His "decline" will be minimal.

Now, go ahead and give me your take on his FULL SEASONS, INCLUDING the post-season.

Furthermore, why do you ignore his ELEVATING his rebounding? Or the fact that he ousthot the post-season league eFG%'s by huge margins? Or the fact that he was slaughtering his HOF peers by staggering margins in ppg, rpg, apg, and FG%?

CHERRY PICKED.

Says the guy who loves to cherry pick single games and stats to prove his point. :rolleyes:

30 is his career PPG in the regular season you dumb fvck. 22 is his career playoff PPG. And 18 is his career finals PPG. 10% my ass.

[houston]

stat geeking does get out hand alot of times

[LAZERUSS]

Says the guy who loves to cherry pick single games and stats to prove his point. :rolleyes:

30 is his career PPG in the regular season you dumb fvck. 22 is his career playoff PPG. And 18 is his career finals PPG. 10% my ass.

10%. Idiot. And again, CONTEXT.

But here is an example of "22-18"...


The Wilt-bashers always use the "30-22-18" argument (which, BTW, would round to 30-23-19) as some kind of example of Chamberlain's "decline" in his post-season play.

How about his 66-67 post-season, in which he averaged 21.7 ppg, 29.1 rpg, and 9.0 apg, then? Obviously he must REALLY have been a flop that year, right? At that point in his NBA career, his regular season averages were 37.6 ppg, 24.0 rpg, 4.0 apg, and on a .522 FG%. Of course, he was always a great defensive center, and most likely the greatest shot-blocker of his generation (sorry Russell,, but in the known research we have, Wilt was easily the GOAT.)

So, "Mr. 30-22-18", which only refers to his post-season scoring, and ignores everything else he did...could "only" put up a 22 ppg post-season run, and an even worse 18 ppg Finals, in arguably, his peak season.

Oh wait..he didn't average 30 ppg in 66-67. In fact, he "only" averaged 24.1 ppg. So, his "decline" was really from 24 down to 22 in the post-season, and down to 18 in the Finals.

Now, how did he fare against the same teams in the post-season, that he faced in his regular season? Against the Royals, his regular season averages were 28.6 ppg, 23.7 rpg, 9.4 apg, and on a .695 FG%. Against Russell's Celtics, his regular season averages were 20.3 ppg, 26.7 rpg, 6.6 apg, and a .549 FG%. Now, in his nineregular season H2H's with the Warriors, he faced Nate Thurmond in six of them. He actually fared worse in the games that Nate missed, so here were his cumulative averages... 20.7 ppg, 25.9 rpg, 8.8 apg, and on a .562 FG%. In his six H2H's with Thurmond, he averaged 20.8 ppg, 25.0 rpg, 8.5 apg, and on a .633 FG% (yes, a .633 FG% against THURMOND.)

BTW, here were his regular season H2H's with Russell and Thurmond:

Russell: 12.2 ppg, 21.1 rpg, 4.1 apg, .425 FG%
Wilt: 20.3 ppg, 26.7 rpg, 6.6 apg, .549 FG%

Thurmond: 13.2 ppg, 23.8 rpg, .320 FG%
Wilt: 20.8 ppg, 25.0 rpg, 8.5 apg, and on a .633 FG%


Ok, round-by-round in the playoffs:

For those that claim that Wilt wasn't even Philly's scoring leader in the post-season (he finished 2nd at 21.7 ppg behind Hal Greer's 27.7 ppg), how about the Sixers very first playoff game? Chamberlain scored what would be a team playoff high in that game, with 41 points, on 19-30 shooting. In his very next game, he hung 37 points, on 16-24 shooting. After that, his shot attempts dropped considerably. Why? Was it because he couldn't score? Well, in his third playoff game, he could only score 16 points, on 8-13 shooting, BUT in addition to pulling down 30 rebounds, he also handed out...get this... 19 assists! Furthermore, while we don't have his exact block totals, there were accounts of as high as 20 blocked shots. And in the series finale (the first round was only a best-of-five back then), he put up an "ordinary" 18 points, on 7-14 shooting, with 27 rebounds, and 9 assists, in a blowout win.

First round numbers:

28.0 ppg, 26.5 rpg, 11.0 apg, and on a .617 FG%. The "bashers" would tell you that he did decline, dropping from 28.6 ppg and on a .695 FG% in his regular season H2H's, all the way down to a 28.0 ppg on a .617 FG% (oh, and as he nearly always did, he ELEVATED his rebounding), from 23.7 ppg to 26.5 rpg, as well as raising his apg from 9.4 apg to 11.0 apg.

And, for those that did not catch that...yes, he AVERAGED a TRIPLE DOUBLE. In fact it was also a double-double in scoring and rebounding, 28-27, and in fact, a near TRIPLE-TRIPLE!


In the 66-67 EDF's, the "declining" Wilt then carried his 68-13 Sixers to a 4-1 series blowout over the eight-time defending 60-21 Celtics. In fact, they nearly swept the Celtics, narrowly losing game four in Boston by a 121-117 margin (and Wilt was hobbled with an assortment of nagging injuries in that game, BTW.) Furthermore, in the clinching game five win, the Celtics had jumped out to a 17 point first quarter lead, before the Sixers mounted a comeback. By mid-way thru the 4th period, the Sixers had blown the game wide-open, and were leading by a 131-104 margin, en route to a 140-116 win.

As usual, Chamberlain just demolished Russell. However, for the first time in his career H2H's with Russell,he had an equal supporting cast...that was healthy, and that played to their normal capabilities in the playoffs.

In game one, Russell played his best game of the series. He scored 20 points, on 7-14 shooting, but could only snare 15 rebounds. Meanwhile, Chamberlain hung a KNOWN quad-double on Russell, with a 24 point game, on 9-13 shooting, 32 rebounds, 13 asisists, and 12 blocks. BTW, not only did Chamberlain crush Russell on the glass, his 32 rebounds came in a game with a total of 128. Oh, and Wilt not only blocked 12 shots, he out-blocked Russell by a 12-1 margin.

In game two, Chamberlain easily outplayed Russell, outscoring him by a 15-14 margin; outshooting him by a 5-11 to 5-14 margin; outrebounding him by a 29-24 margin; he equalled Russell in assists at 5; and he outblocked him by a 5-3 margin.

The Sixers moved out to an insurmountable 3-0 series lead in game three. Russell had a huge rebounding game, with 29. HOWEVER, Chamberlain STILL slaughtered him on the glass, with a playoff record 41 rebounds. BTW, those 41 rebounds came in a game in which there were a total of 134 (or 30% of the available rebounds.) Both Wilt and Russell handed out nine assists, but Wilt outscored Russell, 20-10, while outshooting him by an 8-14 to 3-13 margin.

Again, in game four, Chamberlain's knees were bothering him, and his mobility was really hindered. This was Russell's last hurrah. He was still easily outplayed by Chamberlain, but he did manage to outrebound Wilt for the only time in the series, 28-22. Wilt outscored him, 20-9; outshot him, 8-18 to 2-7, and outassisted him, 10-5.

I already provided the score in the clinching game five blowout win, but Wilt's completely overwhelmed a helpless Russell in the game. In the first half, the game was still close, and by half-time, Chamberlain had hung 22 points on Russell. So, it was pretty clear, that had he needed to score, he was well on his way to yet another 40 point pasting of Russell. However, he didn't need to. His teammates finally outplayed Russell's, and the rout was on. Hal Greer led Philly in scoring, with 32 points, but on 12-28 shooting. Chamberlain finished with 29 points, but on 10-16 from the field. Overall, Wilt crushed Russell in every facet of the game. He outscored Russell, 29-4; out-shot Russell, 10-16 to 2-5; outassisted Russell, 13-7; and outrebounded Russell by a staggering 36-21 margin (in a game with 134 available rebounds.) We don't have Russell's blcoked shots, but Wilt added 7 of those.

BTW, the Russell-supporters have never been able to answer this question:

Just the season before, Wilt and his Sixers were down in the EDF series to Boston by a 3-1 margin. So, with his back to the wall, Chamberlain erupted for a 46 point game, on 19-34 shooting, and with 34 rebounds...in a close loss.

Ok, now in the '67 EDF's, it was Russell who was faced with that exact same scenario. His team was also down 3-1, and facing elimination in game five. Did Russell explode for a 46-34 game against Wilt? Hell no, he meekly led his team to slaughter with a meager 4 point, 2-5 shooting, 21 rebound game (while Wilt was just murdering him.)

How come? If Russell truly "owned" Wilt, as the "bashers" will tell you (the famous "11-2" crap), what happened in '67? BTW, Chamberlain essentially outplayed Russell by those margins in nearly all of their H2H series.

For the series:

Russell: 10.2 ppg, 23.4 rpg, 6.0 apg, and on a .358 FG%.
Wilt: 21.6 ppg, 32.0 rpg, 10.0 apg, and on a .556 FG%.

And again, a TRIPLE-DOUBLE series. As a sidenote, in their known blocks, Wilt held a 29-8 margin.


In the Finals, Wilt's Sixers stumbled some, but still won by a convincing 4-2 margin. And, at their best in the series, they obliterated the Warriors in game two, 126-95.

Keep in mind that this was also a PEAK Thurmond. Nate would finish second to Wilt in the MVP balloting, and for the season against the NBA, he averaged 18.7 ppg, 20.5 rpg, and shot .437 from the field.

And yet Chamberlain just mopped the floor with Nate in the Finals. True, his scoring declined a little, but the domination was evident in the entire series. Wilt outscored Nate in five of the six H2H's; outrebounded Nate in five of the six H2H's; outassisted Nate in five of the six H2H's; and outshot Thurmond from the field in all six. We don't have much block info, but Wilt had a known 10 blocked shots in game two, and an estimated 15 in game four. BTW, Wilt's game two was yet a another QUAD DOUBLE (10-38-10-10.)

And again, when it mattered most, it was WILT who was scoring. In the clinching game six win, Greer could only score 15 points, on 5-16 shooting. Meanwhile, Chamberlain oustcored Nate, 24-12, while outshooting Thurmond from the field by an 8-13 to 4-13 margin.

Overall:

Nate: 14.3 ppg, 26.7 rpg, 3.3 apg, and on...get this... a .343 FG%.
Wilt: 17.5 ppg, 28.5 rpg, 6.8 apg, and on a .560 FG%.


So, here was an example of a PEAK Chamberlain, who could "only" put up a post-season of 22 ppg, and a Finals of 18 ppg. Way down from his career high of 50 ppg. Yep...a great example of the "30-22-18" argument used against Wilt. THAT was Wilt "the choker."

THE most dominant "22-18" in NBA history.

[Asukal]

10%. Idiot. And again, CONTEXT.

But here is an example of "22-18"...



THE most dominant "22-18" in NBA history.

The most dominant choker! :applause:

[La Frescobaldi]

Wow, I didn't check the numbers but I suspected enough that Jones would probably have a good plus minus. I was agreeing with La Fresco about his impact to the game more than anything.
dang was i ever wrong lol

i figured it HAD to be bad.

[La Frescobaldi]

Bobby Jones actually rates extremely well in +/- and net on/off. I got a hold of the Harvey Pollack guides from 76-77 to the present awhile back, and a poster on another board was willing to compute net on/off from the raw +/-:


YEAR MIN ORTG DRTG NET
1978-79 2304 0,7 -2,7 3,4
1979-80 2125 5,3 -3,1 8,4
1980-81 2046 3,8 -7,0 10,8
1981-82 2181 1,1 -3,1 4,2
1982-83 1749 9,7 -1,3 11,0
1983-84 1761 5,5 -2,1 7,6
1984-85 1633 8,4 -2,0 10,4
1985-86 1519 1,1 -2,7 3,8

The poster drew comparison to Manu Ginobili based on the data, but noted that Jones seems more balanced between offense/defense.

I had to go back and find this because wow that's breathtaking what Bobby was doing and he STILL shows up like this??? :bowdown: :bowdown:

boy was i wrong on that one :wtf: :wtf: :lol

[LAZERUSS]

Fpliii,

Just curious...

what did RAPM tell us about how the Cavs and Heat would perform this year?

[comerb]

Westbrook leading the NBA at 33.3, and AD right behind at 32.5.

Neither anywhere close to historic seasons, and yet both are well ahead of the all-time record.

Just another of the MANY "mini stats" that can be thrown out today. RAPM, ORtG, DRtG, WS, WS/48, OBPM, DBPM...all WORTHLESS stats.

Reminds me of some baseball's ridiculous statistics, like how a player has batted against a certain left-handed pitcher, in a day game, after the seventh inning, with runners in scoring position, and with a two-strike count.


You just posted two of the most dominant players in the league and used that as a reason for why the stat that shows how dominant they are is a "joke".

Maybe you should pull your head out of your ass before you post next time.

[LAZERUSS]

You just posted two of the most dominant players in the league and used that as a reason for why the stat that shows how dominant they are is a "joke".

Maybe you should pull your head out of your ass before you post next time.

So, what you are saying is this, then...

either the rest of the current NBA is pure shit, or, PER is a useless stat.

Which is it?

[fpliii]

Fpliii,

Just curious...

what did RAPM tell us about how the Cavs and Heat would perform this year?
I projected 43 wins for Miami and 60 for Cleveland using an RAPM model. I believe fivethirtheight.com (using RPM) had the Heat a little higher (mid 40s) and the Cavs a bit lower (mid-high 50s).

[LAZERUSS]

I projected 43 wins for Miami and 60 for Cleveland using an RAPM model. I believe fivethirtheight.com (using RPM) had the Heat a little higher (mid 40s) and the Cavs a bit lower (mid-high 50s).

I don't have anything in print, nor would anyone else care, but your projections are about the same as mine. Cleveland with 60 wins, and Miami with 45.

Of course, injuries can and already have changed things.

And while it doesn't look like the Cavs will hit 60 wins, I suspect that they will get it together before long, and likely make it to the Finals...albeit, Chicago is looking like the favorite from the East right now.

[kshutts1]

Westbrook leading the NBA at 33.3, and AD right behind at 32.5.

Neither anywhere close to historic seasons, and yet both are well ahead of the all-time record.

Just another of the MANY "mini stats" that can be thrown out today. RAPM, ORtG, DRtG, WS, WS/48, OBPM, DBPM...all WORTHLESS stats.

Reminds me of some baseball's ridiculous statistics, like how a player has batted against a certain left-handed pitcher, in a day game, after the seventh inning, with runners in scoring position, and with a two-strike count.
Coming way late to the party, so I'm sure my argument has been said already.

I hate all advanced stats as much as anyone, yet I recognize that they, much like any traditional stat, does not tell the full picture. PER is a tool, just like APG or FG% is a tool. None is perfect.

[LAZERUSS]

Coming way late to the party, so I'm sure my argument has been said already.

I hate all advanced stats as much as anyone, yet I recognize that they, much like any traditional stat, does not tell the full picture. PER is a tool, just like APG or FG% is a tool. None is perfect.

100% agreed.

The bottom line is this...

you can choose advanced, or even any stats, to validate what we already know.

[3ball]

tl;dr - RAPM isolates player impact, taking into account quality of teammates and opponents. It's fine, as long as you note that a player's impact is tied to his role (so only players in similar roles should compared), and throw out observations over a small sample. It's impossible to get perfect estimates, but again, this is a process universally accepted in economics, science, and the like based on how precise it is. The error is ridiculously small, with a RMSE (SD of population as a whole) of around +/- .5 SDs of the coefficients. It's pretty much the highest quality estimate of player impact out there, which is why high-leverage gamblers and front offices of successful/smart teams use it as faithfully and often as they do (and it's stood the test of time; the original APM was released in 02 by Jeff Sagarin and Wayne Winston, the latter of whom worked for the Mavs for around a decade).


when you say RAPM isolates a player's impact on scoring margin - it only truly measures a player's impact on scoring margin alongside the various lineups on his own team - this value can subsequently be used as an approximation for what the player's impact on scoring margin would be on another team in a similar role, but never the same role (same teammates, same system), so it's far from exact.

RAPM couldn't predict carmelo's performance this season under the new management and coaching.. and previous RAPM data is not reliable in predicting love's current and future RAPM with the Cavs... coaching/system issues and player chemistry issues hurt RAPM's ability to predict how a player would play on another team, so i'm not sure you can say "RAPM isolates a player's impact" and just stop right there - i think you need to add "within the role a player plays on his own team".

since RAPM only measures the impact a player has alongside his own teammates, and not the impact a player would have on ANY team, RAPM is the no different from PER or most other stats - it measures what the player does on his team, against other teams.

also, one thing i think that someone with RAPM knowledge might miss, is that a cohesive team using superior strategy will perform at a higher level without it's star, than a less cohesive team with lesser strategy would without THEIR star.. in jordan's case, his teams operated pretty well without him (or an inside presence or shot-blocker for that matter) because of superior strategy and a championship culture - so the fact that MJ was still able to add a top 5 RAPM on top of that, is more remarkable than the RAPM of a guy playing with a bunch of alleged scrubs who are "D-League" without him.

so Jordan's very high RAPM on teams using optimal strategy that could cope without him, is more impressive than Lebron's RAPM on crappy teams that sucked without him.

Also, the Bulls ability without Jordan or a shot blocker shows how much better their superior their strategy was, how optimal the triangle offense was for the playing environment in that era, and how good the TEAM had gotten at the system... it's a testament to Jordan's skills, that he could achieve GOAT production (scoring and assists) within such an optimal system, without taking anything away from his teammates, as evidenced by his very high RAPM.

Btw, as other posters have noted, by 1993 after two championships, Jordan wasn't as motivated or trying to beat guys as badly as previous seasons due to mental fatigue... here's the bulls team talking about MJ's lost "zeal" for the game in 1993... https://www.youtube.com/watch?v=NZ_jNwyNu0Q&t=4m20s

[HurricaneKid]

when you say RAPM isolates a player's impact on scoring margin - it only truly measures a player's impact on scoring margin alongside the various lineups on his own team - this value can subsequently be used as an approximation for what the player's impact on scoring margin would be on another team in a similar role, but never the same role (same teammates, same system), so it's far from exact.

RAPM couldn't predict carmelo's performance this season under the new management and coaching.. and previous RAPM data is not reliable in predicting love's current and future RAPM with the Cavs... coaching/system issues and player chemistry issues hurt RAPM's ability to predict how a player would play on another team, so i'm not sure you can say "RAPM isolates a player's impact" and just stop right there - i think you need to add "within the role a player plays on his own team".

since RAPM only measures the impact a player has alongside his own teammates, and not the impact a player would have on ANY team, RAPM is the no different from PER or most other stats - it measures what the player does on his team, against other teams.

also, one thing i think that someone with RAPM knowledge might miss, is that a cohesive team using superior strategy will perform at a higher level without it's star, than a less cohesive team with lesser strategy would without THEIR star.. in jordan's case, his teams operated pretty well without him (or an inside presence or shot-blocker for that matter) because of superior strategy and a championship culture - so the fact that MJ was still able to add a top 5 RAPM on top of that, is more remarkable than the RAPM of a guy playing with a bunch of alleged scrubs who are "D-League" without him.

so Jordan's very high RAPM on teams using optimal strategy that could cope without him, is more impressive than Lebron's RAPM on crappy teams that sucked without him.

Also, the Bulls ability without Jordan or a shot blocker shows how much better their superior their strategy was, how optimal the triangle offense was for the playing environment in that era, and how good the TEAM had gotten at the system... it's a testament to Jordan's skills, that he could achieve GOAT production (scoring and assists) within such an optimal system, without taking anything away from his teammates, as evidenced by his very high RAPM.

Btw, as other posters have noted, by 1993 after two championships, Jordan wasn't as motivated or trying to beat guys as badly as previous seasons due to mental fatigue... here's the bulls team talking about MJ's lost "zeal" for the game in 1993... https://www.youtube.com/watch?v=NZ_jNwyNu0Q&t=4m20s

So 3Ball wants a stat that will predict back injuries now.

And LOL at saying MJs RAPM (which isn't even available) is more accurate than LBJs.

[j3lademaster]

So 3Ball wants a stat that will predict back injuries now.

And LOL at saying MJs RAPM (which isn't even available) is more accurate than LBJs.It is actually http://stats-for-the-nba.appspot.com/ratings/90s.html.

It's not as complete of a data bank to draw off of as more recent years, but it gives you a rough idea.

[ralph_i_el]

http://i62.tinypic.com/2hd7kex.jpg


everybody run this thread is no longer safe

[Dr.J4ever]

http://i62.tinypic.com/2hd7kex.jpg


everybody run this thread is no longer safe

We're safe in our anonymity. :lol

[ralph_i_el]

We're safe in our anonymity. :lol

Jeff is a radical imam. All our emails are being forwarded to ISIS as we speak

[Dr.J4ever]

Jeff is a radical imam. All our emails are being forwarded to ISIS as we speak

Then he better watch out for drones outside his office/house:lol

[ralph_i_el]

Then he better watch out for drones outside his office/house:lol

I for one welcome our new muslim overlords :roll: