PDA

View Full Version : What are your thoughts on PER?



JasonJohnHorn
03-26-2014, 01:45 PM
I noticed that a lot of guys are big on PER, however, in doing the Line of the Night threads, and keeping track of them over the last two seasons, I've noticed that in many instances the guys with the best PER don't get voted for LOTN.

Last year, for example, Noah got Line of the Year with 23 points, 21 rebounds 11 blocks, 3 steals and 1 assist, but his PER for that line wasn't even in the top ten for the season, so obviously the majority of people don't agree with PER when casting their votes, but the majority seem to think that it is a strong indicator of which players are better.

Bruno
03-26-2014, 02:05 PM
its okay. i tend to use it a lot as a point of reference. it's nice when you don't feel like thinking and want a number to justify your feelings/eye test about a given player. i think its useful but player A having a higher PER than player B doesn't mean player A is a better player in every comparison. sometimes it does.

flea
03-26-2014, 02:07 PM
Nobody should ever use it, but for some reason they keep doing it. It's about like using RBIs in baseball, except you think you're sounding smart.

TheMightyHumph
03-26-2014, 02:07 PM
What's PER?

valade16
03-26-2014, 02:17 PM
I noticed that a lot of guys are big on PER, however, in doing the Line of the Night threads, and keeping track of them over the last two seasons, I've noticed that in many instances the guys with the best PER don't get voted for LOTN.

Last year, for example, Noah got Line of the Year with 23 points, 21 rebounds 11 blocks, 3 steals and 1 assist, but his PER for that line wasn't even in the top ten for the season, so obviously the majority of people don't agree with PER when casting their votes, but the majority seem to think that it is a strong indicator of which players are better.

That's because LOTN is for 1 game. PER gets better the larger the sample size.

It is a good stat, however it is not an end all be all as it can't really measure defense outside blocks and steals.

abe_froman
03-26-2014, 02:22 PM
overall its a useful quick gauge on a players offense,but not end all/be all

as for the example given,you never mention what noah shot as PER takes into account efficiency and since is an offense oriented stat will favor high scoring games,something like a 20/20 game wont be as high as a 40 or 50 point game

mightybosstone
03-26-2014, 02:23 PM
I noticed that a lot of guys are big on PER, however, in doing the Line of the Night threads, and keeping track of them over the last two seasons, I've noticed that in many instances the guys with the best PER don't get voted for LOTN.

Last year, for example, Noah got Line of the Year with 23 points, 21 rebounds 11 blocks, 3 steals and 1 assist, but his PER for that line wasn't even in the top ten for the season, so obviously the majority of people don't agree with PER when casting their votes, but the majority seem to think that it is a strong indicator of which players are better.

I wouldn't use PER in that way. You're looking at a single game sample size rather than the big picture of an entire season. I don't use it to judge a game, I use it to judge a player's production over long periods of time. Also, there's a big difference between what stats we find interesting on a box score and what stats these formulas believe to be the best. A guy with 20 points, 20 rebounds and 10 blocks has put up an insanely rare stat line, but advanced formulas would probably prefer a guy who put up something like 40 points, 10 rebounds, 6 assists and 4 steals.

But PER is like anything else. It's a flawed barometer that can't just be used by itself. Personally, I prefer WS to PER and find it to be a better indicator of a player's worth, but it's nice to have a variety of stats to look at when you're trying to judge a player's value.

torocan
03-26-2014, 02:43 PM
I'm personally not a fan of PER. The over emphasis on shot attempts really skews it towards chuckers even if they're crappy shooters.


The same critique offered for NBA Efficiency also applies to Hollinger’s PERs, except the problem is even worse. Hollinger argues that each two point field goal made is worth about 1.65 points. A three point field goal made is worth 2.65 points. A missed field goal, though, costs a team 0.72 points.

Given these values, with a bit of math we can show that a player will break even on his two point field goal attempts if he hits on 30.4% of these shots. On three pointers the break-even point is 21.4%. If a player exceeds these thresholds, and virtually every NBA played does so with respect to two-point shots, the more he shoots the higher his value in PERs. So a player can be an inefficient scorer and simply inflate his value by taking a large number of shots.

http://wagesofwins.com/2006/11/17/a-comment-on-the-player-efficiency-rating/

Essentially, as long as I shoot over 30%, I can chuck all game and my PER will look great.

If I'm going to look at a player, I prefer stats compiled over longer samples (seasons vs games), and ones that are somewhat relative. So stuff like RAPM, WS if you're talking inclusive stats. Or stuff like TS%, ORR, DRR, PPS, or PPP if you're talking granular stats.

PER is just WAY too subject to abuse by the like of players like Rudy Gay for me to feel comfortable with forming any conclusions.

bleedprple&gold
03-26-2014, 02:48 PM
its okay. i tend to use it a lot as a point of reference. it's nice when you don't feel like thinking and want a number to justify your feelings/eye test about a given player. i think its useful but player A having a higher PER than player B doesn't mean player A is a better player in every comparison. sometimes it does.

This. It's useful as a quick summary since you only have to look at one number. But you need to dig deeper if you want to really analyze a game or a player.

ManRam
03-26-2014, 02:55 PM
A decent foot in the door when trying to determine a player's overall value offensively. Nothing more, nothing less. Good for finding a starting point when trying to compare players.

tredigs
03-26-2014, 02:59 PM
I'm personally not a fan of PER. The over emphasis on shot attempts really skews it towards chuckers even if they're crappy shooters.



http://wagesofwins.com/2006/11/17/a-comment-on-the-player-efficiency-rating/

Essentially, as long as I shoot over 30%, I can chuck all game and my PER will look great.

If I'm going to look at a player, I prefer stats compiled over longer samples (seasons vs games), and ones that are somewhat relative. So stuff like RAPM, WS if you're talking inclusive stats. Or stuff like TS%, ORR, DRR, PPS, or PPP if you're talking granular stats.

PER is just WAY too subject to abuse by the like of players like Rudy Gay for me to feel comfortable with forming any conclusions.

Though I agree with the premise, it's worth mentioning that this is also the site whose metric (WinsProduced) will tell you that Dandre Jordan is twice the producer as Blake Griffin (and that Lebron + KD don't quite match his impact), and that all fall short of Andre Drummond. Basically, they value rebounds... followed by rebounds, then assists and scoring volume+efficiency.

If you just want to look at offense, I think PER gets much more right then wrong, but it can make you look very stupid if you're just pointing at that stat alone in an attempt to make a point. And it isn't the best single stat, it's simply one of the first and by far most popular (the one closest linked to the ESPN hype machine).

slashsnake
03-26-2014, 03:29 PM
Though I agree with the premise, it's worth mentioning that this is also the site whose metric (WinsProduced) will tell you that Dandre Jordan is twice the producer as Blake Griffin (and that Lebron + KD don't quite match his impact), and that all fall short of Andre Drummond. Basically, they value rebounds... followed by rebounds, then assists and scoring volume+efficiency.

If you just want to look at offense, I think PER gets much more right then wrong, but it can make you look very stupid if you're just pointing at that stat alone in an attempt to make a point. And it isn't the best single stat, it's simply one of the first and by far most popular (the one closest linked to the ESPN hype machine).

Well there is no one number that figures everything in there. Players still need to pass the eye test.

But with PER over an extended period of time, it doesn't look bad. The all time PER leaders are pretty much all time greats. There's a few outliers. Stoudemire would be an example, but I have a feeling as his career progresses he will fall down that list.

flea
03-26-2014, 03:32 PM
Well there is no one number that figures everything in there. Players still need to pass the eye test.

But with PER over an extended period of time, it doesn't look bad. The all time PER leaders are pretty much all time greats. There's a few outliers. Stoudemire would be an example, but I have a feeling as his career progresses he will fall down that list.

A stat that can tell me that Lebron is better than Marcus Thornton is not needed. PER is epitome of the axiom "junk in, junk out." Even the guy that created it thinks it sucks and says we shouldn't use it.

dhopisthename
03-26-2014, 03:33 PM
I'm personally not a fan of PER. The over emphasis on shot attempts really skews it towards chuckers even if they're crappy shooters.



http://wagesofwins.com/2006/11/17/a-comment-on-the-player-efficiency-rating/

Essentially, as long as I shoot over 30%, I can chuck all game and my PER will look great.

If I'm going to look at a player, I prefer stats compiled over longer samples (seasons vs games), and ones that are somewhat relative. So stuff like RAPM, WS if you're talking inclusive stats. Or stuff like TS%, ORR, DRR, PPS, or PPP if you're talking granular stats.

PER is just WAY too subject to abuse by the like of players like Rudy Gay for me to feel comfortable with forming any conclusions.

here is hollingers response to this statement



Berri leads off with a huge misunderstanding of PER — that the credits and debits it gives for making and missing shots equate to a “break-even” shooting mark of 30.4% on 2-point shots. He made this assumption because he forgot that PER is calibrated against the rest of the league at the end of the formula.

Actually, if we took a player was completely average in every other respect for the 2006-07 season — rebounds, free throws, assists, turnovers, etc. — and gave him a league-average rate of shots, and all of them were 2-pointers, and he shot 30.4%, he’d end up with a PER of 7.18. As long-time PER fans know, that would make him considerably worse than nearly every player in the league.

To end up with a league-aveage PER of 15,00, the actual break-even mark in this case is 48.5%, which is exactly what the league average is on 2-point shots this season.

D-Leethal
03-26-2014, 03:43 PM
Impact level will never be properly weighed by an "all-inclusive" stat.

torocan
03-26-2014, 03:49 PM
Though I agree with the premise, it's worth mentioning that this is also the site whose metric (WinsProduced) will tell you that Dandre Jordan is twice the producer as Blake Griffin (and that Lebron + KD don't quite match his impact), and that all fall short of Andre Drummond. Basically, they value rebounds... followed by rebounds, then assists and scoring volume+efficiency.

If you just want to look at offense, I think PER gets much more right then wrong, but it can make you look very stupid if you're just pointing at that stat alone in an attempt to make a point. And it isn't the best single stat, it's simply one of the first and by far most popular (the one closest linked to the ESPN hype machine).

Um, don't know where you get that unless you were talking mid-season. Last I checked Lebron and KD were ahead of Deandre Jordan.

And to be fair, Centers have a disproportionate impact on the game due to rim protection. Griffin may be the better player than DJ, but having NO center is much worse than having NO Griffin if you're talking winning games. Other players can make up for scoring, not so much when you're talking rim protection. Especially if you have no credible back ups.

Impact is not the same as comparing how good one player is versus another, especially when there's a credible back up.

That said, Wins Produced is far from perfect, but I'd take it with a smaller grain of salt than PER as a starting point. PER is okay if all you want to look at is scoring, assuming you look at context and efficiency. As a standalone stat, it leaves FAR more to be desired that WP48 if you ask me. :)

torocan
03-26-2014, 03:57 PM
here is hollingers response to this statement

And this is Berri's response to Hollinger...


My response to his response

Let me summarize. I am arguing that if a player shoots inefficiently he can increase his PERs ranking by simply taking more shots. Hollinger responds by noting that a player can increase his PERs value by shooting more efficiently. You will note that this is not actually a response to what I am saying.

Let me try and illustrate with an example. Say a player shoots 40% from two point range. If he takes 10 shots his PERs ranking will rise by 6.6 (4*1.65) and decline by 4.32 (6*0.72). So the net gain is 2.28. If he doubles his shot attempts to 20 he will see his PERs ranking rise by 13.2 (8*1.65) and decline by 8.64 (12*0.72). Now his net gain is 4.56. The player is still shooting 40%, which is below average. But his PERs ranking increases the more shots he takes.

PERs, and other metrics like NBA Efficiency and Points Created, tell us that an inefficient shooter becomes more valuable the more shots he wastes. But that is not consistent with winning basketball games. Inefficient shooting does not simply become a good thing because you do more of it.

http://wagesofwins.com/2006/11/24/john-hollinger-responds/

That said, it's a p*ssing match between two stats heads with things to gain and lose. Interpret them how you will. :)

ewing
03-26-2014, 03:58 PM
I like it when my cats per

Sadds The Gr8
03-26-2014, 04:03 PM
Another piece to the puzzle, like every stat...

still1ballin
03-26-2014, 04:08 PM
wack

ManRam
03-26-2014, 04:12 PM
Impact level will never be properly weighed by an "all-inclusive" stat.

And no creator of any of these stats will suggest that the particular stat in question does just that...

tredigs
03-26-2014, 04:13 PM
Um, don't know where you get that unless you were talking mid-season. Last I checked Lebron and KD were ahead of Deandre Jordan.

And to be fair, Centers have a disproportionate impact on the game due to rim protection. Griffin may be the better player than DJ, but having NO center is much worse than having NO Griffin if you're talking winning games. Other players can make up for scoring, not so much when you're talking rim protection. Especially if you have no credible back ups.

Impact is not the same as comparing how good one player is versus another, especially when there's a credible back up.

That said, Wins Produced is far from perfect, but I'd take it with a smaller grain of salt than PER as a starting point. PER is okay if all you want to look at is scoring, assuming you look at context and efficiency. As a standalone stat, it leaves FAR more to be desired that WP48 if you ask me. :)

Ah sorry, Lebron + KD do show as 2 and 3 thousandths of a percent better per WP/48 (So, identical). And all fall slightly short of the mighty Drummond.

And I heavily disagree with your reasoning. Obviously you are going to have a backup for any position; 10 times out of 10 you're going to want a D league backup in place of Deandre Jordan instead of Blake Griffin. What WP doesn't realize is that a drop in rebounding rate is often a positive as it relates to defense and closing out on defenders. There's also the inherent rebounding advantage for bigs in simply being the guys closest to the basket for a gimme bounce to them (especially noticed on FT's). If they simply accounted for contested rebounds and threw out the junk, they'd instantly improve what they're ultimately going for with the stat. But, they don't. And we see Bogut/Jordan/Drummond as top 10 NBA producers per WinsProduced as a result. Chris Anderson and Brandan Wright being other notables in this range.

Personally, although I understand their pluses, I can't take either stat entirely seriously.

ztilzer31
03-26-2014, 04:26 PM
Unusable stat with a bogus name.

JasonJohnHorn
03-26-2014, 04:33 PM
What's PER?

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

In a nut shell, it just takes all of a players stats and boils them down into one number. Some people have different formulas. I think ESPN, BBR and NBA.com all use slightly different formulas, but they are all very simialr.



The thing is, a given formula might give more or less weight to the importance of rebounding or assists or turnovers than somebody else might, so some people disagree with it.

That said, as most posters are saying, it is a good tool, but should not be used exclusively.

JasonJohnHorn
03-26-2014, 04:35 PM
I like it when my cats per

lol

bagwell368
03-26-2014, 04:36 PM
It's a waste of time.

nickdymez
03-26-2014, 04:58 PM
I just like how people who use these stats act like they are basketball guru's.

jerellh528
03-26-2014, 05:06 PM
Stats alone can never quantify a players worth. This isn't chess where a supercomputer can calculate moves and beat a human. The human aspect effects basketball perhaps more than any other team sport. Using per to say one player is better than another is plain ludicrous to me. It is though, a decent formula to use as a jump off point in the start of talking about a player's production.

KnicksorBust
03-26-2014, 05:14 PM
I love PER and +/- and WS/48 and TS%. These are all tools that allow me to quantify things that I can't necessarily describe. It allows me to compare players. They give me a clearer picture of the game that I love to watch and allow me to discuss things on a deeper level.

Who are the career leaders in PER?

#1. Michael Jordan
#2. LeBron James
#3. Shaquille O'Neal
#4. David Robinson
#5. Wilt Chamberlain

How flawed can this stat really be when it clearly brings the cream to the top every time?

KnicksorBust
03-26-2014, 05:17 PM
http://www.basketball-reference.com/leaders/per_career.html

I challenge you to find me truly elite franchise changing player, title clinching superstar that isn't even in the top 30. There is no other stat that is that damn good.

jerellh528
03-26-2014, 05:20 PM
http://www.basketball-reference.com/leaders/per_career.html

I challenge you to find me truly elite franchise changing player, title clinching superstar that isn't even in the top 30. There is no other stat that is that damn good.

Haha yea almost any stat. Look at the top 10 all time leaders in points.
1 Kareem Abdul-Jabbar 38,387
2 Karl Malone 36,928
3 Michael Jordan 32,292
4 KOBE BRYANT 31,700
5 Wilt Chamberlain 31,419
6 Shaquille O'Neal 28,596
7 Moses Malone 27,409
8 Elvin Hayes 27,313
9 Hakeem Olajuwon 26,946
10 Oscar Robertson

A bunch of franchise changers there too lol.

tredigs
03-26-2014, 05:21 PM
http://www.basketball-reference.com/leaders/per_career.html

I challenge you to find me truly elite franchise changing player, title clinching superstar that isn't even in the top 30. There is no other stat that is that damn good.

Rick Barry. Done.

flea
03-26-2014, 05:27 PM
I love PER and +/- and WS/48 and TS%. These are all tools that allow me to quantify things that I can't necessarily describe. It allows me to compare players. They give me a clearer picture of the game that I love to watch and allow me to discuss things on a deeper level.

Who are the career leaders in PER?

#1. Michael Jordan
#2. LeBron James
#3. Shaquille O'Neal
#4. David Robinson
#5. Wilt Chamberlain

How flawed can this stat really be when it clearly brings the cream to the top every time?

Is this a serious post?

Chronz
03-26-2014, 05:27 PM
Useful in that its extremely easy to disect and makes pace/minutes adjustments for you.

tredigs
03-26-2014, 05:40 PM
Then there's the Anti-PER contributors like Ben Wallace and Dennis Rodman - who according to the stat we'd have never known were even top 20 players in the league despite having impact for many seasons that were both MASSIVE and irreplaceable. Pistons don't sniff the Finals let alone win without Big Ben.

WadeKobe
03-26-2014, 05:42 PM
Though I agree with the premise, it's worth mentioning that this is also the site whose metric (WinsProduced) will tell you that Dandre Jordan is twice the producer as Blake Griffin (and that Lebron + KD don't quite match his impact), and that all fall short of Andre Drummond. Basically, they value rebounds... followed by rebounds, then assists and scoring volume+efficiency.

If you just want to look at offense, I think PER gets much more right then wrong, but it can make you look very stupid if you're just pointing at that stat alone in an attempt to make a point. And it isn't the best single stat, it's simply one of the first and by far most popular (the one closest linked to the ESPN hype machine).

Odd, because the WoW crow don't say that at all. They have stated Jordan is the best center, better than Drummond, and that KD and LeBron are the teo best players in the game.

The biggest problem with PER is that it is arbitrary, not empirical.

Chronz
03-26-2014, 05:43 PM
Then there's the Anti-PER contributors like Ben Wallace and Dennis Rodman - who according to the stat we'd have never known were even top 20 players in the league despite having impact for many seasons that were both MASSIVE and irreplaceable. Pistons don't sniff the Finals let alone win without Big Ben.

Pistons dont win without Rodman either IMO, but yea, PER does not pretend to measure positional defense. Part of the problem with WP, and why we both think they overrate rebounders, is because they credit those players for defensive stops as well.

tredigs
03-26-2014, 05:48 PM
Odd, because the WoW crow don't say that at all. They have stated Jordan is the best center, better than Drummond, and that KD and LeBron are the teo best players in the game.

The biggest problem with PER is that it is arbitrary, not empirical.

Wow "crow"? Crowd? I'm not sure what you're getting at. I'm simply looking at the #'s shown by the Wins Produced formula, which indicate that Cp3 is currently the top producer in the game, followed by Drummond and then Lebron/KD/Deandre Jordan. It will also highlight that Blake Griffin is not a top ten power forward; ranking lower than the likes of Amir Johnson, Jordan Hill, Kris Humphries and Terrence Jones. All of these being far ahead of Dirk, who doesn't crack the top 25 power forwards. You can only have so many anomalies in your stats findings before the stat as a whole needs revamping. And that stat needs massive revamping.

Method28
03-26-2014, 05:55 PM
http://www.basketball-reference.com/leaders/per_career.html

I challenge you to find me truly elite franchise changing player, title clinching superstar that isn't even in the top 30. There is no other stat that is that damn good.

Franchise changing player you have Iverson below 30.

Also this stat claims ak47 is better than pippen, Mullin and payton.

Jamiecballer
03-26-2014, 06:05 PM
PER isn't great but if that's your only advanced stat you are still far ahead those who only use their eyes.

abe_froman
03-26-2014, 06:14 PM
http://www.basketball-reference.com/leaders/per_career.html

I challenge you to find me truly elite franchise changing player, title clinching superstar that isn't even in the top 30. There is no other stat that is that damn good.
ok...
rick barry,john stockton,pat ewing,walt fraizer,bull russell...shall i go on?

look its a nice quick reference point that can be used in conjunction with others in an overall argument,but you are trying to use it like wins above replacement in baseball,and you cant.its just not as great and all inclusive as you think.

b@llhog24
03-26-2014, 06:25 PM
It's a nice reference point and makes comparing across positions and eras a lil easier. It's far from perfect though and as a whole shouldn't be used by itself.


Sent via Tapatalk

AddiX
03-26-2014, 07:26 PM
PER gives some posters here boners. True story...

Jamiecballer
03-26-2014, 07:28 PM
it's good for measuring offensive contribution, really good actually but doesn't help evaluate defense at all. then again not much does.

ManRam
03-26-2014, 07:33 PM
Franchise changing player you have Iverson below 30.

Also this stat claims ak47 is better than pippen, Mullin and payton.

PER isn't "claiming" that at all. It's a stat; it's goal is not to rank every player all time. Nothing says "if a players' PER is higher than someone else's, he's better than that player". And if you think that's what it is doing, well, the problem is you...not the stat. And if someone is using it on its lonesome to prove or "claim" something like that, than they're using the stat wrong too.


A stat isn't crap merely because people use it incorrectly. People often use PER to sum up an entire player's worth, and that's foolish. But if you know what it is, what it's telling you, and realize that it's just one of many tools to evaluate players, well...then it has a solid purpose :shrug: There is value to it. It isn't meaningless...but it most certainly isn't the one stat to rule them all. Far from it.

jerellh528
03-26-2014, 07:38 PM
PER isn't "claiming" that at all. It's a stat; it's goal is not to rank every player all time. Nothing says "if a players' PER is higher than someone else's, he's better than that player". And if you think that's what it is doing, well, the problem is you...not the stat. And if someone is using it on its lonesome to prove or "claim" something like that, than they're using the stat wrong too.


A stat isn't crap merely because people use it incorrectly. People often use PER to sum up an entire player's worth, and that's foolish. But if you know what it is, what it's telling you, and realize that it's just one of many tools to evaluate players, well...then it has a solid purpose :shrug: There is value to it. It isn't meaningless...but it most certainly isn't the one stat to rule them all. Far from it.

This. The only bad part is many on psd don't seem to agree. Often per is used as an end all by the uninformed fan.

JJ_JKidd
03-26-2014, 09:03 PM
Ahhh the love for PER and any other advanced (individual players) stat. And I thought basketball is a team game?

IndyRealist
03-26-2014, 09:29 PM
here is hollingers response to this statement

Hollingers response is a classic STRAW MAN ARGUMENT. He never actually addresses the issue, that shooting poorly will cause your PER to go up the more you shoot.

His response is instead "with PER if you are average at everything, you will have an average PER." How does that respond to the point at all?

IndyRealist
03-26-2014, 09:44 PM
Wow "crow"? Crowd? I'm not sure what you're getting at. I'm simply looking at the #'s shown by the Wins Produced formula, which indicate that Cp3 is currently the top producer in the game, followed by Drummond and then Lebron/KD/Deandre Jordan. It will also highlight that Blake Griffin is not a top ten power forward; ranking lower than the likes of Amir Johnson, Jordan Hill, Kris Humphries and Terrence Jones. All of these being far ahead of Dirk, who doesn't crack the top 25 power forwards. You can only have so many anomalies in your stats findings before the stat as a whole needs revamping. And that stat needs massive revamping.

That's actually the exact opposite of what you should do. The fact that it doesn't line up with popular perception is not a detriment, because popular perception is WRONG so often.

That's what PER is designed to do, line up with what people already believe. But looking at the forumla you can see that PER thinks poor below average shooters are good players, as long as they shoot a lot. Coincidentially, that's what people tend to think as well. If you look at every voted on award, MVP, ROY, all-star voting, etc. they are biased toward players who score a large number of points per game. Doesn't matter how well they score, just the volume. People even take PPG into considering for DPOY, for crying out loud.

On WP specifically, you're looking at WP48, which is controlled for time played. Per minute, yes Chris Paul is the highest producer. Overall, though, WP has Kevin Durant as the MVP.

GunFactor187
03-26-2014, 10:14 PM
To me, it's just a small part of the big picture. Eye view, basic stats, and advanced stats - all important aspects leading up to a complete philosophy of the game. Gestalt Principle, the whole is greater than the sum of its parts.

Tony_Starks
03-26-2014, 11:24 PM
PER is irrelevant outside of the internet. I've yet to watch a game and think "wow that guy is good, I wonder what his PER is?"

Trwood12
03-26-2014, 11:44 PM
I don't like it. It doesn't really measure a player's overall impact on the game.

Fnom11
03-27-2014, 12:07 AM
No one stat should define a player or an opinion about one

tredigs
03-27-2014, 02:51 AM
That's actually the exact opposite of what you should do. The fact that it doesn't line up with popular perception is not a detriment, because popular perception is WRONG so often.

That's what PER is designed to do, line up with what people already believe. But looking at the forumla you can see that PER thinks poor below average shooters are good players, as long as they shoot a lot. Coincidentially, that's what people tend to think as well. If you look at every voted on award, MVP, ROY, all-star voting, etc. they are biased toward players who score a large number of points per game. Doesn't matter how well they score, just the volume. People even take PPG into considering for DPOY, for crying out loud.

On WP specifically, you're looking at WP48, which is controlled for time played. Per minute, yes Chris Paul is the highest producer. Overall, though, WP has Kevin Durant as the MVP.

You're ignoring the major point of the over-valuing of rebounding and exactly why that's created such a false dominance of certain players in WP, namely many I've mentioned.

amos1er
03-27-2014, 03:11 AM
This. The only bad part is many on psd don't seem to agree. Often per is used as an end all by the uninformed fan.

Couldn't agree more. According to PER, David Robinson is a far greater player than Magic Johnson. That alone should tell you that the stat has significant flaws.

Kashmir13579
03-27-2014, 03:51 AM
PER tells me what I already know. Useless stat.

Crackadalic
03-27-2014, 03:54 AM
A basic advance stat for dummies that has been outclass by better metric advance stats that is out there

Kashmir13579
03-27-2014, 03:55 AM
Then there's the Anti-PER contributors like Ben Wallace and Dennis Rodman - who according to the stat we'd have never known were even top 20 players in the league despite having impact for many seasons that were both MASSIVE and irreplaceable. Pistons don't sniff the Finals let alone win without Big Ben. Bruce Bowen

PatsSoxKnicks
03-27-2014, 04:16 AM
Hollingers response is a classic STRAW MAN ARGUMENT. He never actually addresses the issue, that shooting poorly will cause your PER to go up the more you shoot.

His response is instead "with PER if you are average at everything, you will have an average PER." How does that respond to the point at all?

I don't know if Hollinger factored this in but what about the fact that contested FG% correlates higher with offensive efficiency? That's mainly because of the fact that most shots in the NBA are contested. Those volume shooters also tend to be the shooters who have the highest ability to hit contested shots, especially near the basket where there's a huge difference.

Wins Produced isn't really better than PER, it's based on box score stats which are flawed. It doesn't accurately capture defense either, which at least RAPM does- albeit even if it misses the why. But there'll be better defensive metrics.

PatsSoxKnicks
03-27-2014, 04:28 AM
I'm personally not a fan of PER. The over emphasis on shot attempts really skews it towards chuckers even if they're crappy shooters.



http://wagesofwins.com/2006/11/17/a-comment-on-the-player-efficiency-rating/

Essentially, as long as I shoot over 30%, I can chuck all game and my PER will look great.

If I'm going to look at a player, I prefer stats compiled over longer samples (seasons vs games), and ones that are somewhat relative. So stuff like RAPM, WS if you're talking inclusive stats. Or stuff like TS%, ORR, DRR, PPS, or PPP if you're talking granular stats.

PER is just WAY too subject to abuse by the like of players like Rudy Gay for me to feel comfortable with forming any conclusions.

I like RAPM but it doesn't tell you why. And thats important. But the way of analytics is using the new data. Not gonna say who or where I heard this but one of the teams' basketball analytics director mentioned he got 32 gigs of unique data from SportVu. Amazing the amount of data available now.

torocan
03-27-2014, 09:48 AM
I like RAPM but it doesn't tell you why. And thats important. But the way of analytics is using the new data. Not gonna say who or where I heard this but one of the teams' basketball analytics director mentioned he got 32 gigs of unique data from SportVu. Amazing the amount of data available now.

To be fair, RAPM and WS were formulated BEFORE the advent of SportsVu. And they're operating off the extremely limited data set that was available prior to SportsVu and Vantage.

Now that we're seeing more people with access to more granular and reliable statistics (versus the brute force of the Box Score and manual counting off grainy game tape), I expect both RAPM and WS to obsolesce and eventually fall to the way side much as PPG has drifted off the stats map.

Yes, there are a ton of better advanced stats out there. Most of it is proprietary and not accessible to punters like us. The amount of data that teams now have access to is stunning, and that gives them the ability to craft the data in ways that we couldn't even have imagined 5 years ago. And a lot of it is hard to verify unless you've ponied up the big bugs to access the SportsVu product or Vantage.

That said, until we see a wider dissemination of that data (if you're reading this NBA, make MORE SportsVu data available please!), we got what we got.

In 10 years, I suspect RAPM and WS (or their future equivalents) will look VERY different than they do today.

As for PER, PER is a stat at the tail end of it's shelf life. Too fundamentally flawed for my taste, and too many good minds working on better formulations based on better data sets.

IndyRealist
03-27-2014, 12:00 PM
You're ignoring the major point of the over-valuing of rebounding and exactly why that's created such a false dominance of certain players in WP, namely many I've mentioned.
But you are only considering it overvalued because popular perception says so. Offensive rebounds are the easiest way to generate extra possessions, and controlling possessions is one of the primary drivers of wins.

Again, what you're doing is the exact opposite of the scientific method. "The research doesn't agree with what I think to be true, so the research must be wrong." I'm not saying WP has it right, but using popular perception as the barometer for any metric is simply unscientific.

Chronz
03-27-2014, 12:37 PM
PER is irrelevant outside of the internet. I've yet to watch a game and think "wow that guy is good, I wonder what his PER is?"
Coaches/Analyst within the NBA have looked at PER and/or other forms of linear weights. Just because you dont think the way some people who actually work in the NBA , doesn't mean its limited to where you've experienced it. The hubris of some people. Plz dont talk on things you know very little about.

tredigs
03-27-2014, 12:57 PM
But you are only considering it overvalued because popular perception says so. Offensive rebounds are the easiest way to generate extra possessions, and controlling possessions is one of the primary drivers of wins.

Again, what you're doing is the exact opposite of the scientific method. "The research doesn't agree with what I think to be true, so the research must be wrong." I'm not saying WP has it right, but using popular perception as the barometer for any metric is simply unscientific.
#1, you're drastically jumping to conclusions as to what I value and why I value it.
#2, you're completely sidestepping my underlying point of why certain players receive those rebounds and why it should be weighted in accordance with whether or not they were contested or not. To put it (very) simply, for some players it's their sole purpose to attempt to rebound - and while that is one valuable commodity - I feel that they have a MUCH easier replacement value and that the far more advanced skill-sets of players who can contribute in a myriad of ways and are relied upon for such are being undermined by the stat.

flea
03-27-2014, 01:03 PM
Coaches/Analyst within the NBA have looked at PER and/or other forms of linear weights. Just because you dont think the way some people who actually work in the NBA , doesn't mean its limited to where you've experienced it. The hubris of some people. Plz dont talk on things you know very little about.

Are Hollinger's constants in PER actually based on historical linear weights? Because I thought he just kind of made up some constants based on what he thought, i.e., what a rebound is worth compared to a 3 point field goal.

IndyRealist
03-27-2014, 01:30 PM
#1, you're drastically jumping to conclusions as to what I value and why I value it.
#2, you're completely sidestepping my underlying point of why certain players receive those rebounds and why it should be weighted in accordance with whether or not they were contested or not. To put it (very) simply, for some players it's their sole purpose to attempt to rebound - and while that is one valuable commodity - I feel that they have a MUCH easier replacement value and that the far more advanced skill-sets of players who can contribute in a myriad of ways and are relied upon for such are being undermined by the stat.

It's hard not to "jump to conclusions" when you simply say it overvalues rebounds as if it were a truism without any explanation.

Proper rebounding is an extremely advanced skill set, especially when you have to incorporate it into screen setting/defensive positioning which puts the rebounder out of position to block out. Simply put, if it were easy everyone would do it. And without further clarification from you it again seems like you're relying on conventional wisdom as opposed to scientific method by stating that rebounding is comparatively easy to replace over more "advanced skill-sets."

I would argue that rebounding is harder to replace than most other skill sets because size is a large contributing factor and as you look for disproportionately large people your pool of skilled players diminish drastically. 6'6" guys who are athletic are a dime a dozen, every Dleague team has 4 or 5. However 6'11" who is athletic, coordinated, AND understands basketball on an NBA level? Much more difficult.

Chronz
03-27-2014, 01:42 PM
Are Hollinger's constants in PER actually based on historical linear weights? Because I thought he just kind of made up some constants based on what he thought, i.e., what a rebound is worth compared to a 3 point field goal.
Its based on "VOP" (Value of possession). Pretty much all of these rating systems come to different conclusions because of how they identify those weights. WP for example, comes about it by translating team importance to team wins, down to the individual player.

All-In
03-27-2014, 01:49 PM
I like RAPM but it doesn't tell you why. And thats important. But the way of analytics is using the new data. Not gonna say who or where I heard this but one of the teams' basketball analytics director mentioned he got 32 gigs of unique data from SportVu. Amazing the amount of data available now.
I’m pretty sure RAPM takes in account how good the players your playing with, the players your playing against, if its home or away, if its back to back or one day of rest…theres probably more variables but those are the ones I know….RAPM is my personally my favorite stat….along with Defensive/Offensive rating over the team mean, and on-off court net efficiency….but there is no one go to stat…I don’t like PER or WS…VORP is ok

tredigs
03-27-2014, 01:50 PM
It's hard not to "jump to conclusions" when you simply say it overvalues rebounds as if it were a truism without any explanation.

Proper rebounding is an extremely advanced skill set, especially when you have to incorporate it into screen setting/defensive positioning which puts the rebounder out of position to block out. Simply put, if it were easy everyone would do it. And without further clarification from you it again seems like you're relying on conventional wisdom as opposed to scientific method by stating that rebounding is comparatively easy to replace over more "advanced skill-sets."

I would argue that rebounding is harder to replace than most other skill sets because size is a large contributing factor and as you look for disproportionately large people your pool of skilled players diminish drastically. 6'6" guys who are athletic are a dime a dozen, every Dleague team has 4 or 5. However 6'11" who is athletic, coordinated, AND understands basketball on an NBA level? Much more difficult.

It's one skill set, and it's also a hustle-based stat and a major reason why these "energy" guys are often the ones who are so good at it (bottom line being that if they didn't contribute here, many would not have a place in the NBA). I think (know) you're underestimating how important positioning is to rebounding and that most players/teams will simply defer rebounding to certain players (be it by design of the offense/defense or by guards just allowing the ball to fall into the bigs hands before they take it up the court so they can pad their stats... which happens countless times a game). Offensive rebounding does indeed hold higher value, but part of that is also due to just being in the right place at the right time while others create for the offense. And that's what I don't want to see showing up in a stat like WP as if it is this massively powerful skillset that should garner extra weight.

If they simply accounted for contested rebounds as I'm saying, the stat would already be 10 fold better and weighted more appropriately where we wouldn't be seeing guys like Jordan Hill and Kris Humphries show as more productive assets than Blake Griffin and Dirk Nowitzki. Don't try and tell me this is me doing the scientific method wrong and falling into common perception, they simply hold nowhere CLOSE to the overall value of the latter, and any stat that alludes otherwise needs work.

Jamiecballer
03-27-2014, 01:50 PM
PER tells me what I already know. Useless stat.

any stat that indicates that guys like Iverson were massively overrated has at least some merit. believe it or not, not everyone can distinguish between bulk accomplishments and quality performance.

tredigs
03-27-2014, 01:57 PM
I do have a question Indy: Who do you think Morey would choose first in an open-draft scenario where cap space was a non issue. Kenneth Faried or Blake Griffin?

PatsSoxKnicks
03-27-2014, 02:57 PM
Yes, there are a ton of better advanced stats out there. Most of it is proprietary and not accessible to punters like us.

Speak for yourself ;)



In 10 years, I suspect RAPM and WS (or their future equivalents) will look VERY different than they do today.


Or this summer. Maybe even a few weeks from now. "New data" prior informed RAPM would be interesting.

PatsSoxKnicks
03-27-2014, 03:06 PM
Still waiting for IndyRealist to address the fact that contested eFG% correlates better with offensive efficiency than open FG%. Again, obviously based on the fact most shots are contested but don't you think most of those players who shoot a lot have that ability to hit contested shots at the highest rate? Probably why they are taking those shots. While their %'s might not be good, it's probably better for them to take that contested shot than another player who maybe doesn't have as good skill at making contested shots.

I doubt PER is attempting to factor that in but it inadvertently has I think. Of course probably best to look at the relationship between TS% and contested eFG% but I think there's something to that.

PatsSoxKnicks
03-27-2014, 03:21 PM
I’m pretty sure RAPM takes in account how good the players your playing with, the players your playing against, if its home or away, if its back to back or one day of rest…theres probably more variables but those are the ones I know….RAPM is my personally my favorite stat….along with Defensive/Offensive rating over the team mean, and on-off court net efficiency….but there is no one go to stat…I don’t like PER or WS…VORP is ok

Yeah it does but that's not the why I'm talking about. I'm talking about the reasons a player is good defensively. RAPM doesn't tell you if the player effectively defends screens or plays good help defense or contests most of his shots etc. There's no components to it that can be broken down. That's the big issue with the stat (Dean Oliver will tell you the same thing among others).

VORP I think was found to be the most accurate stat based on a retrodiction contest. I think it'll get passed though by a new RAPM metric that'll come out (maybe not tho, we'll see). Personally I like Pelton's WARP, for whatever reason it wasn't included in the retrodiction contest, so not sure how it performs relative to others.

Chronz
03-27-2014, 03:33 PM
I do have a question Indy: Who do you think Morey would choose first in an open-draft scenario where cap space was a non issue. Kenneth Faried or Blake Griffin?

Blake, but in some part due to the defensive liability that Faried rates as in metrics that these all inclusive stats dont account for. Hes big on +/-, Blake is a monster there. When the 2 agree, its usually a no brainer I think.


As for WP, the biggest turn off for me is in its valuation of efficiency. They've toned down the importance in rebounding some years ago, but O/D Efficiency always being the number 1 factor in team wins means its the most influential regardless. Its like the anti-PER in some cases where it overrates limited role players who play within the confine of their "skill curve", who could otherwise suffer catastrophic declines in efficiency if more offensive responsibility was tasked on them. But still, you could argue that being the case in most metrics depending on the dropoff. I just rather believe the answer lies somewhere in the middle, while keeping an eye out for the impact any subtle changes in role/teammates may bring, by all measures.

Chronz
03-27-2014, 03:40 PM
Yeah it does but that's not the why I'm talking about. I'm talking about the reasons a player is good defensively. RAPM doesn't tell you if the player effectively defends screens or plays good help defense or contests most of his shots etc. There's no components to it that can be broken down. That's the big issue with the stat (Dean Oliver will tell you the same thing among others).

VORP I think was found to be the most accurate stat based on a retrodiction contest. I think it'll get passed though by a new RAPM metric that'll come out (maybe not tho, we'll see). Personally I like Pelton's WARP, for whatever reason it wasn't included in the retrodiction contest, so not sure how it performs relative to others.
http://www.basketballprospectus.com/article.php?articleid=1985

Not sure how it stacks up today but he ran his own tests in 2009. Its interesting that he mentions a "tom thibs" metric, any idea if arguably the best coach in the game has his own pet stat?

flea
03-27-2014, 04:10 PM
Its based on "VOP" (Value of possession). Pretty much all of these rating systems come to different conclusions because of how they identify those weights. WP for example, comes about it by translating team importance to team wins, down to the individual player.

Yeah VOP is his constant but isn't it just a constant he's chosen, for reasons he hasn't said and with a value he hasn't stated? Not sure why I should take a stat like that seriously, especially when the output correlates with nothing in particular - it's just a positive number somewhere between 1 and ~35-40.

Even if he did have a pretty good way of determining his VOP constant, his adjustments for pace do not fix the issues inherent in having a constant for VOP. A possession for the Memphis Grizzlies is a lot more valuable than one for the Rockets - and pace of play does not factor that in.

All-In
03-27-2014, 04:17 PM
Yeah it does but that's not the why I'm talking about. I'm talking about the reasons a player is good defensively. RAPM doesn't tell you if the player effectively defends screens or plays good help defense or contests most of his shots etc. There's no components to it that can be broken down. That's the big issue with the stat (Dean Oliver will tell you the same thing among others).

VORP I think was found to be the most accurate stat based on a retrodiction contest. I think it'll get passed though by a new RAPM metric that'll come out (maybe not tho, we'll see). Personally I like Pelton's WARP, for whatever reason it wasn't included in the retrodiction contest, so not sure how it performs relative to others.

Yea, I’m with you, I’m hoping that one day things like rotations/help-defense, post defense, P&R defense, will be available for the public domain because I think they exist with statVU/player-tracking data but are withheld from teams for whatever reason….but then again breaking down stuff P&R defense and rotations/help–defense is relative to what offensive set the team is running/offensive philosophy, what type of defense does the team run/how they play the P&R…theres so many variables I don’t know if there will ever be a true quantitative stat on defense

PatsSoxKnicks
03-27-2014, 04:17 PM
I do have a question Indy: Who do you think Morey would choose first in an open-draft scenario where cap space was a non issue. Kenneth Faried or Blake Griffin?

If I had some balls, I'd ask Eli (VP for the Rockets, used to be at the Count the Basket blog) myself but doubt I'd get a response since those analytics guys are so busy. Up till like 2-3am watching tape, though they get into work later I think. Pretty much any basketball job has late hours tho, be it an ESPN basketball writer or an exec in the front office.

flea
03-27-2014, 04:21 PM
Yeah VOP is his constant but isn't it just a constant he's chosen, for reasons he hasn't said and with a value he hasn't stated? Not sure why I should take a stat like that seriously, especially when the output correlates with nothing in particular - it's just a positive number somewhere between 1 and ~35-40.

Even if he did have a pretty good way of determining his VOP constant, his adjustments for pace do not fix the issues inherent in having a constant for VOP. A possession for the Memphis Grizzlies is a lot more valuable than one for the Rockets - and pace of play does not factor that in.

I'll answer my own question, here's the formula for VOP:

VOP = lg_PTS / (lg_FGA - lg_ORB + lg_TOV + 0.44 * lg_FTA)

Yep it's a constant with no team adjustments. The .44 constant is unexplained, unless that is what he thinks the proportionate value of a free throw attempt is in relation to turnovers, offensive, rebounds, and field goal attempts. Again, I'm unconvinced.

Chronz
03-27-2014, 04:25 PM
Yeah VOP is his constant but isn't it just a constant he's chosen, for reasons he hasn't said and with a value he hasn't stated? Not sure why I should take a stat like that seriously, especially when the output correlates with nothing in particular - it's just a positive number somewhere between 1 and ~35-40.

Even if he did have a pretty good way of determining his VOP constant, his adjustments for pace do not fix the issues inherent in having a constant for VOP. A possession for the Memphis Grizzlies is a lot more valuable than one for the Rockets - and pace of play does not factor that in.

You've lost me, hes laid out his methods. Some of the logic can be flimsy and it may be simplistic but the rationale isn't exactly out of nowhere. There is some statistical trends going into the weight of some stats. I say some because I cant recall why he values assists as 1/3 of a made basket, but I do recall the constantly upheld weight of an offensive rebound negating a missed shot. The major differences in all of them come in valuing the efficiency-usage trade off.

Not sure about that pace complaint, pace factor is a variable all these metrics attempt to cover last I checked

PatsSoxKnicks
03-27-2014, 04:28 PM
http://www.basketballprospectus.com/article.php?articleid=1985

Not sure how it stacks up today but he ran his own tests in 2009. Its interesting that he mentions a "tom thibs" metric, any idea if arguably the best coach in the game has his own pet stat?

Oh I was thinking of a different retrodiction one ran at a different site, think it was more recent. The name of the site is escaping me though. I think I did see that a long time ago.

I don't even know who the stats guy is for the Bulls. I feel like these guys all try to remain so anonymous (and there's probably some truth to that because they don't want to give anything away, a particular celts employee wouldn't even share how he goes about basic statistical methods in the presence of Neil, whose with the Hawks)

I bet he does have something though. Would make sense, question is if it's any good.

Chronz
03-27-2014, 04:30 PM
I'll answer my own question, here's the formula for VOP:

VOP = lg_PTS / (lg_FGA - lg_ORB + lg_TOV + 0.44 * lg_FTA)

Yep it's a constant with no team adjustments. The .44 constant is unexplained, unless that is what he thinks the proportionate value of a free throw attempt is in relation to turnovers, offensive, rebounds, and field goal attempts. Again, I'm unconvinced.

IIRC, thats for BBR's uPER, which is unadjusted. His entire methodology is explained in his books, though Im sure its been posted some where online.

flea
03-27-2014, 04:33 PM
Pace controls for total amount of possessions, not their value. It's not a straight-line correlation to say each additional possession decreases the value of the other possessions by X amount, but that's essentially what the statistic does.

If you want to be reductive and say, "PER measures what it says it measures in the way it says it measures it," then okay. But I'm used to a little more rigor in my advanced stats. WAR actually measures something, and you can test its assumptions with different values. It's not all that clear what assumptions PER is making, I suspect because the whole damn thing is a set of assumptions that were put in there with an eye on the results. In my profession, doing things with an eye on adjusting outcomes puts you on very shaky footing.

Chronz
03-27-2014, 04:44 PM
Pace controls for total amount of possessions, not their value. It's not a straight-line correlation to say each additional possession decreases the value of the other possessions by X amount, but that's essentially what the statistic does.

If you want to be reductive and say, "PER measures what it says it measures in the way it says it measures it," then okay. But I'm used to a little more rigor in my advanced stats. WAR actually measures something, and you can test its assumptions with different values. It's not all that clear what assumptions PER is making, I suspect because the whole damn thing is a set of assumptions that were put in there with an eye on the results. In my profession, doing things with an eye on adjusting outcomes puts you on very shaky footing.

Thats what any statistic factoring in pace of play will do. And I dont buy that it was done with an eye on the outcome.

Chronz
03-27-2014, 04:45 PM
And for the record, I would agree with you guys on WARP based analysis being best.

WadeKobe
03-27-2014, 04:49 PM
I do have a question Indy: Who do you think Morey would choose first in an open-draft scenario where cap space was a non issue. Kenneth Faried or Blake Griffin?


I still think you're being unfair to Berri, et al here.

First, let me state that I agree that te newer iteration of WP probabaly does still overvalue defensive rebounding a bit, but the adjustments they made were all based on the slew of data and theory criticizing their old WP. So, while I am always comfortable with improvement, it seems most Internet forum guys are t dealing empirically with the issue. That is, saying it overvalued rebounding, still, after the 2011 adjustments, would need to actually deal with data on the issue. I have yet to see it, even from guys on APBR forums. That is, all of the negative interaction with rebounding figures takes place pre-2011.

Tha isn't to say thy are correct, only to say that they have consistently dealt with empirical data and methodology -- and have been ridiculously transparent. I like that, and I don't see the same consistency from vocal internet critics, and I would like to. Hopefully that's a fair opinion/approach.

Likewise, even Berri, et al seem to -- if their rhetoric is indicative -- value Griffin more highly as a player than Faried.

Part of the issue is that you're only looking at numbers and not dealing with analysis. They've pretty consistently said that WP only tells "what", not "why" and that "why" is an intracal part of analysis. So, two players with similar numbers, while one takes on a much more important role in their team, is more valuable in almost any situation.

One of your issues earlier with numbers didn't deal with that , and also didn't deal with the way that plays itself out in the sum. That is, you lamented that WP48 finds jordan to be more productive (albeit insignificantly) than Durant, even though Durant plays more minutes, carries te offensive load, and this carries a higher total value expressed by WP, in which case Jordan's would likely drop significantly.

Again, this isn't to say WP is perfect, or even the best, just that it seems that most interaction with it is under informed and, as a result, often unfair.

Per minute numbers like WP48 are more useful in discussing minute allocation and rotations. So if you find a low minute second unit player with higher WP48 than a starter, WP does not suggest they are higher producers. Only that, per minute, against second unit players, they are very productive but there is likely a break-even point. That means he should quite possiblly be getting more minutes than he is if he is outproducing another bench player at his position. However, across how many minutes, and in what lineups, against what units could tha player remain that productive (or even good production to a lesser extent)? At what point does playing the player big minutes against starters cause them to become unproductive?

You have to deal with any stat I terms if what it is actually saying, and when there is so much commentary available from the makers of the stat, it seems unwise and unfair to pluck a bunch of numbers and criticisms them without dealing with the analysis and commentary available.

Also, I am always surprised that the biggest weakness of WP is rarely if ever discussed or criticized, and that is the methodology behind their defensive component (team adjustment).

WS utilizes DRtg if I remember correctly, which at least attempts to deal with the team's defensive performance while that player is on the court. WP takes the total defensive performance of the team across 140 minutes and divides it evenly per minute across the individual minutes played. That means weak defenders on good defensive teams, whose unit regularly struggles defensively and brigs the team's defense is down get hidden in the adjustment while better units are punished.

If this was fixed, would WP correlate as well with wins, or would they need to make further adjustments to bring the correlation back into focus?

Again, it just seems to me that most folks who discuss and criticize WP often do so from a very under informed position.

flea
03-27-2014, 04:49 PM
Thats what any statistic factoring in pace of play will do. And I dont buy that it was done with an eye on the outcome.

And that's a fine down-and-dirty way to do pace, but it doesn't address the problems of VOP constant. It's like saying park factors (also inherently flawed, like pace adjustments) in WAR also inform you on the value of fly balls over grounders. Not by a long shot, and certainly not by what you proclaim to measure.

I mean my point isn't all that astounding or anything. The value of a possession gained from an offensive rebound or steal is a lot greater than the value of a possession from an opponent's unforced error forcing play stoppage.

IndyRealist
03-27-2014, 08:23 PM
I do have a question Indy: Who do you think Morey would choose first in an open-draft scenario where cap space was a non issue. Kenneth Faried or Blake Griffin?
I myself would empirically select Griffin for impact, but Faried for value. Given that Morey is apparently targeting Melo I suspect a similar strategy.

Griffin adapts well to differing defenses and performs pretty evenly through most games. Faried is prone to highs and lows depending on the defender, because of his physical limitations.

Faried is the value pick, though, because any GM with half a brain is giving Griffin a max deal, where Faried can be had for substantially cheaper.

Regardless of the hypothetical, you'd have to pay them eventually.

IndyRealist
03-27-2014, 08:42 PM
Still waiting for IndyRealist to address the fact that contested eFG% correlates better with offensive efficiency than open FG%. Again, obviously based on the fact most shots are contested but don't you think most of those players who shoot a lot have that ability to hit contested shots at the highest rate? Probably why they are taking those shots. While their %'s might not be good, it's probably better for them to take that contested shot than another player who maybe doesn't have as good skill at making contested shots.

I doubt PER is attempting to factor that in but it inadvertently has I think. Of course probably best to look at the relationship between TS% and contested eFG% but I think there's something to that.

Didn't realize the question was addressed to me. I don't recall what it was in reference to, but I'll go back and look.

After reading the original post, I don't see how it's relevant to PER, which has a break-even point on 2pt FGs of 30.4%. What PER says is that "all players are better the more 2pt FGs they attempt" which is patently untrue. Even, as you reference, some players are better at taking contested FGs, that would only drive the break even point upwards because they are making a higher percentage, not lower it to an unreasonable level. If you shoot a poor percentage on the shot you are attempting, it is a detriment if you take that shot as opposed to giving it to a teammate with a higher percentage shot.

IndyRealist
03-27-2014, 09:53 PM
I do have a question Indy: Who do you think Morey would choose first in an open-draft scenario where cap space was a non issue. Kenneth Faried or Blake Griffin?

See, driving home from work pondering your question, it occurred to me the point of the question. You were trying to get me to say that because WP rates Faried higher than Griffin, that that should be the choice. And then everyone goes nuts saying "OH WP IS NONSENSE BLAH BLAH BLAH!" I choose not to look at metrics in a vacuum. Saying "<insert metric here> says player A is worse than player B so the metric is wrong" is exactly the same as saying "player A is better than player B because <insert metric here> said so". Taking any metric in a vacuum, whether cited as evidence or trying to discredit it, is asinine.

The use of metrics has to include knowing it's strengths and weaknesses. For WP, for instance, it uses a team defensive contribution which VASTLY skews the data when one player (i.e. Roy Hibbert) is disproportionately responsible for the defense. In no fair world does David West get the same defensive adjustment that Roy does. For PER, using it requires the caveat that it is skewed toward volume shooters regardless of shooting efficiency. So in PER it is a relatively valid comparison to look at Melo vs. Durant because they have similar USG% (32.8 vs. 32.8 actually), but completely invalid to compare either of them to Kyle Korver (14% USG).

JEDean89
03-27-2014, 10:03 PM
there are so many different ways to quantify what players on the floor do. imo you know a stat is legit if the best players in the league are at the top when you use it. the 5 best players in PER are KD, LBJ, K-Love, Unibrow and Boogie. Now are those the 5 best players in the league? LBJ and KD are but Love and Boogie are not top 5 players. AD very well could be, we need more time to tell. I think it's a good way to generally quantify a players basic stats, but its an advanced stat that doesn't take advanced stats into account. Bigs have a huge advantage too because of their FG%. IMO if a player has an awesome PER and plays great defense, he is a stud. If his PER sucks and his defense is good, he can still be legit, if his PER is awesome and his defense sucks, I don't want him because he's overpaid, if his PER is average and his defense is above average, I want him.

NBA_Starter
03-27-2014, 10:59 PM
I think that it is overrated.

PatsSoxKnicks
03-28-2014, 02:30 AM
Didn't realize the question was addressed to me. I don't recall what it was in reference to, but I'll go back and look.

After reading the original post, I don't see how it's relevant to PER, which has a break-even point on 2pt FGs of 30.4%. What PER says is that "all players are better the more 2pt FGs they attempt" which is patently untrue. Even, as you reference, some players are better at taking contested FGs, that would only drive the break even point upwards because they are making a higher percentage, not lower it to an unreasonable level. If you shoot a poor percentage on the shot you are attempting, it is a detriment if you take that shot as opposed to giving it to a teammate with a higher percentage shot.

No, I think you're still missing what I was saying. The whole "all players are better the more 2pt FGs they attempt" is I think (and actually I don't think because I don't think Hollinger was thinking of this when he came up with it) basically a proxy for players who attempt a ton of contested shots. The way the NBA is played, you have your "stars" and your non-stars. Those "stars" tend to take a lot of shots and are likely better at hitting contested shots. The non-stars are likely to take mostly open shots and aren't going to be shooting many contested shots. Because open shots have a higher league wide FG% then contested shots (and by a lot), you have those non-stars with better efficiencies. However, the "stars" are taking the tough contested shots that have to be taken by someone and are likely making it at a higher rate than a non-star would make it. But the non-star also never attempts those shots because the star is taking it. It's conterfactual data- attempting to account for the shots that the non-star doesn't take because the star is taking those tougher shots.

As to your last point, it's not that simple. Again if the shot is going to be contested- and a lot are, it's likely the "star" is better at shooting those contested shots. The non-star doesn't have to take that shot and so he maintains a higher efficiency. But the star suffers with a lower efficiency because he has to take that shot. He can hit that contested shot better than the non-star can.

Not sure if thats any clearer, if not I can maybe give an example. But you have to take into account the shots not taken by players, not just the shots taken.

FYI, I'm not just making up this concept here. This is something actual analytics guys in the NBA think about.

PatsSoxKnicks
03-28-2014, 02:37 AM
Yea, I’m with you, I’m hoping that one day things like rotations/help-defense, post defense, P&R defense, will be available for the public domain because I think they exist with statVU/player-tracking data but are withheld from teams for whatever reason….but then again breaking down stuff P&R defense and rotations/help–defense is relative to what offensive set the team is running/offensive philosophy, what type of defense does the team run/how they play the P&R…theres so many variables I don’t know if there will ever be a true quantitative stat on defense

SportVu is actually inferior to Vantage, which actually tracks basically every single event. Literally. Hard to understand without seeing the data though.

Also, look for a new version of RAPM in coming months (maybe weeks?) that uses the Vantage stats as a prior.

PatsSoxKnicks
03-28-2014, 02:51 AM
You've lost me, hes laid out his methods. Some of the logic can be flimsy and it may be simplistic but the rationale isn't exactly out of nowhere. There is some statistical trends going into the weight of some stats. I say some because I cant recall why he values assists as 1/3 of a made basket, but I do recall the constantly upheld weight of an offensive rebound negating a missed shot. The major differences in all of them come in valuing the efficiency-usage trade off.

Not sure about that pace complaint, pace factor is a variable all these metrics attempt to cover last I checked

Yeah, it's in his book. Think it has to do with there being 3 actions to an assisted shot- making the pass, getting open and making the shot. And so he gives the assister 1/3 credit since the shooter does 2/3s of the actions. I believe that's how it's described in the book.

IndyRealist
03-28-2014, 12:05 PM
No, I think you're still missing what I was saying. The whole "all players are better the more 2pt FGs they attempt" is I think (and actually I don't think because I don't think Hollinger was thinking of this when he came up with it) basically a proxy for players who attempt a ton of contested shots. The way the NBA is played, you have your "stars" and your non-stars. Those "stars" tend to take a lot of shots and are likely better at hitting contested shots. The non-stars are likely to take mostly open shots and aren't going to be shooting many contested shots. Because open shots have a higher league wide FG% then contested shots (and by a lot), you have those non-stars with better efficiencies. However, the "stars" are taking the tough contested shots that have to be taken by someone and are likely making it at a higher rate than a non-star would make it. But the non-star also never attempts those shots because the star is taking it. It's conterfactual data- attempting to account for the shots that the non-star doesn't take because the star is taking those tougher shots.

As to your last point, it's not that simple. Again if the shot is going to be contested- and a lot are, it's likely the "star" is better at shooting those contested shots. The non-star doesn't have to take that shot and so he maintains a higher efficiency. But the star suffers with a lower efficiency because he has to take that shot. He can hit that contested shot better than the non-star can.

Not sure if thats any clearer, if not I can maybe give an example. But you have to take into account the shots not taken by players, not just the shots taken.

FYI, I'm not just making up this concept here. This is something actual analytics guys in the NBA think about.

I get what you're saying, but my point is that it can't be good for EVERYONE to take the contested shot, because as you said the person who is good at it should take it over the person who isn't. PER sets the bar so low that anyone gets a net positive for taking shots. You have to take into consideration the average of their shot attempts, not just a subset. You have to include the fact that SOME shots will be open, and look at eFG as a whole rather than tailoring the formula for 80% of the shots and letting the other 20% skew the data. eFG is self scaling for shot distribution.

Jamiecballer
03-28-2014, 12:13 PM
I get what you're saying, but my point is that it can't be good for EVERYONE to take the contested shot, because as you said the person who is good at it should take it over the person who isn't. PER sets the bar so low that anyone gets a net positive for taking shots. You have to take into consideration the average of their shot attempts, not just a subset. You have to include the fact that SOME shots will be open, and look at eFG as a whole rather than tailoring the formula for 80% of the shots and letting the other 20% skew the data. eFG is self scaling for shot distribution.

are you 100 positive about this? i look at a guy like Iverson and think PER wasn't kind to him and he took mad shots compared to most everyone else. same with Anthony. i always assumed a big part of the reason his PER fell way short of what you might expect was the relative inefficiency compared to some of the other big stars.

these are just observations so i'm surprised to hear that if true.

IndyRealist
03-28-2014, 01:37 PM
are you 100 positive about this? i look at a guy like Iverson and think PER wasn't kind to him and he took mad shots compared to most everyone else. same with Anthony. i always assumed a big part of the reason his PER fell way short of what you might expect was the relative inefficiency compared to some of the other big stars.

these are just observations so i'm surprised to hear that if true.

Everyone does get a net positive for taking shots if they shoot over 31% on 2's and over 21% on 3's. Other players simply get a larger bump for actually being efficient.

The issue is not at the top, where virtually every stat agrees that Lebron, Durant, etc. are at the top, but rather in the middle and bottom where some players become overrated due to high USG rates, despite not being very productive. Conversely, some players are underrated due to low USG rates, despite shooting efficiently and posting average or above numbers in every other aspect of the box score.

Hollinger infamously (at least in Indiana) called out Brandon Rush as the worst starter in the NBA at the time. I think his PER was 9.6. However, he was an excellent defender, rebounder (for a SG), and crazy efficient 3pt shooter. He blocked shots, got steals, and didn't turn the ball over. He simply did not take a lot of shots on a team where he was really the 6th option despite being a starter. Virtually every other metric that did not rely on volume of shots rated him at least average, while some significantly above average.

PatsSoxKnicks
03-29-2014, 04:47 AM
I get what you're saying, but my point is that it can't be good for EVERYONE to take the contested shot, because as you said the person who is good at it should take it over the person who isn't. PER sets the bar so low that anyone gets a net positive for taking shots. You have to take into consideration the average of their shot attempts, not just a subset. You have to include the fact that SOME shots will be open, and look at eFG as a whole rather than tailoring the formula for 80% of the shots and letting the other 20% skew the data. eFG is self scaling for shot distribution.

Well I said earlier, I don't think PER was taking that into account necessarily because honestly, how could it when the data wasn't available? It is now and there's no reason to do subsets when obviously you can look at either player specific or the league as a whole. And again, you can look at eFG% for open and contested shots so why use total eFG% when you can be more specific? Frequency data is there too. And if you want to scale it- you use something like XPPS, which is basically a shot selection metric. Could factor in the shot clock too since eFG% is MUCH LOWER from 1-5 seconds.

I'll give an example though, Kobe is about 3-1 in contested to open FGAs. And he's top 5 in shots per touch so he clearly shoots a lot and a lot of those shots are contested. Who else takes those contested shots? And can they shoot whatever Kobe is shooting on those shots? And FYI, for contested 3s for example, you're talking about league average being like 32%, so a below average shooter is gonna be below that 30% mark you mentioned earlier. And if that player is taking a high% of shots near the end of the shot clock, that number will be even lower- near the low 20%'s possibly.

All that said, the 30% mark probably is too low because not all shots are threes. But I think he had some idea perhaps of trying to compensate for penalizing non-stars for shots not taken while rewarding stars for those extra contested shots they have to take because the alternative is a worse %. I think there may have been good intention behind it, but perhaps setting the bar a few % points too low. Although to be honest, I read the book and I don't even recall any of that coming up. If you read the methodology from his books, it actually does makes sense though (but it's not any sort of regression run or anything). Hard to explain unless you actually read it. And I try to remain open to different metrics until I read the methodology.

So for example, I can't really criticize Wins Produced since I need to go through the updated version. Intend to do that at some point, especially because I think Arturo is a pretty sharp guy and puts out quality analysis. I know he doesn't do Wins Produced but he's part of that crowd. I've heard others though completely turned off by the Berri crowd.

PatsSoxKnicks
03-29-2014, 05:10 AM
Hollinger infamously (at least in Indiana) called out Brandon Rush as the worst starter in the NBA at the time. I think his PER was 9.6. However, he was an excellent defender, rebounder (for a SG), and crazy efficient 3pt shooter. He blocked shots, got steals, and didn't turn the ball over. He simply did not take a lot of shots on a team where he was really the 6th option despite being a starter. Virtually every other metric that did not rely on volume of shots rated him at least average, while some significantly above average.

I'm 100% positive Hollinger wouldn't have called him out as such if he didn't think he was also bad defensively. Hollinger's admitted before that PER doesn't account for defense much. And he doesn't use PER as the be all end all. You're definitely short-selling him (and listen to others like Dean Oliver talk about him). He watches a ton of film, obviously now being with the Grizz but even before then when he was a writer (said he's up till like 3am watching film).

IndyRealist
03-29-2014, 11:38 AM
I'm 100% positive Hollinger wouldn't have called him out as such if he didn't think he was also bad defensively. Hollinger's admitted before that PER doesn't account for defense much. And he doesn't use PER as the be all end all. You're definitely short-selling him (and listen to others like Dean Oliver talk about him). He watches a ton of film, obviously now being with the Grizz but even before then when he was a writer (said he's up till like 3am watching film).
100% positive because you know Hollinger personally? This was like 2009 or so when he was still with ESPN, and still defending PER because HIS REPUTATION AND INCOME WAS DEPENDENT ON PEOPLE BELIEVING PER. If you look at his moves in Memphis, they are dramatically anti-PER.

"Brandon Rush has had a forgettable career thus far, but as we head into the home stretch he may be on his way to a milestone of sorts. Rush has a PER of just 9.96 in his second pro season, but somehow leads the Pacers in minutes played with 2,159 --157 more than the next closest player, Troy Murphy. If he manages to maintain his lead, he’ll claim the dubious distinction of being the worst player ever to lead his team in minutes."
http://espn.go.com/blog/truehoop/post/_/id/14569/a-lot-of-minutes-little-production


Well I said earlier, I don't think PER was taking that into account necessarily because honestly, how could it when the data wasn't available? It is now and there's no reason to do subsets when obviously you can look at either player specific or the league as a whole. And again, you can look at eFG% for open and contested shots so why use total eFG% when you can be more specific? Frequency data is there too. And if you want to scale it- you use something like XPPS, which is basically a shot selection metric. Could factor in the shot clock too since eFG% is MUCH LOWER from 1-5 seconds.

I'll give an example though, Kobe is about 3-1 in contested to open FGAs. And he's top 5 in shots per touch so he clearly shoots a lot and a lot of those shots are contested. Who else takes those contested shots? And can they shoot whatever Kobe is shooting on those shots? And FYI, for contested 3s for example, you're talking about league average being like 32%, so a below average shooter is gonna be below that 30% mark you mentioned earlier. And if that player is taking a high% of shots near the end of the shot clock, that number will be even lower- near the low 20%'s possibly.

All that said, the 30% mark probably is too low because not all shots are threes. But I think he had some idea perhaps of trying to compensate for penalizing non-stars for shots not taken while rewarding stars for those extra contested shots they have to take because the alternative is a worse %. I think there may have been good intention behind it, but perhaps setting the bar a few % points too low. Although to be honest, I read the book and I don't even recall any of that coming up. If you read the methodology from his books, it actually does makes sense though (but it's not any sort of regression run or anything). Hard to explain unless you actually read it. And I try to remain open to different metrics until I read the methodology.

So for example, I can't really criticize Wins Produced since I need to go through the updated version. Intend to do that at some point, especially because I think Arturo is a pretty sharp guy and puts out quality analysis. I know he doesn't do Wins Produced but he's part of that crowd. I've heard others though completely turned off by the Berri crowd.
30% is not the break even on 3's for PER, 21% is. 31% is the break even on 2's.

I'm not in any way sure of why we're discussing contested versus open FGs anymore. The reasoning behind it, as you said, has nothing to do with PER, WP, or virtually any other metric since they were created prior to video analysis.

As far as WP is concerned, the same arguments are constantly parroted by the APBR crowd without any understanding of what they're actually talking about. One person formed an intelligible argument and the rest, well, mindlessly agree because they don't understand the argument. Take the rebounding issue for instance. The big argument was that WP does not account for diminishing returns, and thus overvalued rebounds. The response was, "it would overcomplicate the formula for very little payoff". So APBR trolls flood boards like PSD and even wagesofwins.com with mindless parroting "well WP overvalues rebounds so it's nonsense!" nonstop. Finally, to shut them the heck up, the formula was modified to account for rebounding rate and diminishing returns and, guess what? It hardly changed a thing. A few players dropped a couple of percentage points, but for the most part Kenneth Faried is still a monster, and Andrea Bargnani is still bad. And still with the "WP overvalues rebounds" to this day.

People's problem with WP is that it tells them something they don't want to hear, that they are overly focused on raw point totals and not paying enough attention to scoring efficiency and nonscoring aspects of the game. Rebounds matter. Turnovers matter. Efficiency matters.

Ultimately, Berri said, "well if you're sure that your metric(s) is better, then publish in a peer reviewed academic journal where other professionals can critique your work". He's a bit of a jerk, but to my knowledge, that has not happened.

“Faced with the choice between changing one’s mind and proving that there is no need to do so, almost everyone gets busy on the proof.” John Kenneth Galbraith

torocan
03-29-2014, 12:25 PM
People's problem with WP is that it tells them something they don't want to hear, that they are overly focused on raw point totals and not paying enough attention to scoring efficiency and nonscoring aspects of the game. Rebounds matter. Turnovers matter. Efficiency matters.

Shhhh. That gets in the way of perception.

The funny thing about correlation is that when it correlates with something that people already believe, their first response is, "I don't need stats to tell me the obvious!"

However, when it flies in the face of convention, the response is typically more along the lines of, "Stats can be manipulated any way you want."

Sadly, there is a lack of examination of methodology and "lazy" use of stats in discussions. The vast majority of metrics have limitations, strengths and weaknesses. However I guess it's easier to say, "A is better than B because THIS stat said so! So there!" vs "Within the context of XXX, it seems that this stat raises questions about A's contributions vs B's contributions..."

PatsSoxKnicks
03-29-2014, 11:24 PM
100% positive because you know Hollinger personally? This was like 2009 or so when he was still with ESPN, and still defending PER because HIS REPUTATION AND INCOME WAS DEPENDENT ON PEOPLE BELIEVING PER. If you look at his moves in Memphis, they are dramatically anti-PER.

"Brandon Rush has had a forgettable career thus far, but as we head into the home stretch he may be on his way to a milestone of sorts. Rush has a PER of just 9.96 in his second pro season, but somehow leads the Pacers in minutes played with 2,159 --157 more than the next closest player, Troy Murphy. If he manages to maintain his lead, he’ll claim the dubious distinction of being the worst player ever to lead his team in minutes."
http://espn.go.com/blog/truehoop/post/_/id/14569/a-lot-of-minutes-little-production

I have met him yeah and that was after leaving ESPN. Had a good talk with him. But yeah, I'd agree with what you said- obviously at ESPN, he'd promote PER. But also, I still believe what I said earlier.



30% is not the break even on 3's for PER, 21% is. 31% is the break even on 2's.

I'm not in any way sure of why we're discussing contested versus open FGs anymore. The reasoning behind it, as you said, has nothing to do with PER, WP, or virtually any other metric since they were created prior to video analysis.

Fair enough. I do think this is the future of analytics though.



As far as WP is concerned, the same arguments are constantly parroted by the APBR crowd without any understanding of what they're actually talking about. One person formed an intelligible argument and the rest, well, mindlessly agree because they don't understand the argument. Take the rebounding issue for instance. The big argument was that WP does not account for diminishing returns, and thus overvalued rebounds. The response was, "it would overcomplicate the formula for very little payoff". So APBR trolls flood boards like PSD and even wagesofwins.com with mindless parroting "well WP overvalues rebounds so it's nonsense!" nonstop. Finally, to shut them the heck up, the formula was modified to account for rebounding rate and diminishing returns and, guess what? It hardly changed a thing. A few players dropped a couple of percentage points, but for the most part Kenneth Faried is still a monster, and Andrea Bargnani is still bad. And still with the "WP overvalues rebounds" to this day.

People's problem with WP is that it tells them something they don't want to hear, that they are overly focused on raw point totals and not paying enough attention to scoring efficiency and nonscoring aspects of the game. Rebounds matter. Turnovers matter. Efficiency matters.

Ultimately, Berri said, "well if you're sure that your metric(s) is better, then publish in a peer reviewed academic journal where other professionals can critique your work". He's a bit of a jerk, but to my knowledge, that has not happened.

“Faced with the choice between changing one’s mind and proving that there is no need to do so, almost everyone gets busy on the proof.” John Kenneth Galbraith

Honestly, I haven't seen much on WP in the APBR forum. No one from Berri's crowd ever bothers showing their face there and so no one does any analysis on it (i'm talking about the real analysts, not the people who just read everything and don't do anything). In any case, the best way to assess a metric is a simple retrodiction contest. How does the metric do when predicting? Unless that's not the intent of WP, if so, then I'm not really sure to the usefullness since NBA decision makers are interested in prediction.

And yeah, he does seem like a bit of a jerk. But then I don't know him, so I can't really say. But in regards to the peer reviewed professionals, I think there's a bit of a disconnect there. Whose the audience? Academics or NBA decision makers? I would think ultimately these metrics are designed for NBA decision makers. Also, I think Berri gives off the impression that scouting, video analysis doesn't matter, which is why I assume he's never gotten a consulting job with a team (unlike other pioneers like Dean Oliver, Ben Alamar, etc.). But thats speculation on my part, so not fair to him.

As I said though, I have got to meet Arturo, whose one of the primary wagesofwins guys. And he seems to be a nice guy, certainly puts out quality work and I think our views on mid-range shots probably align (though I don't know if I'd go as far, still think the player matters).

FYI, I believe in the retrodiction contest that an independent site ran, Myer's ASPM was the most predictive. RAPM did pretty well too I believe (tho not sure, have to find the link). Anyways, think our new statistical RAPM will improve the metric but we'll see.

IndyRealist
03-30-2014, 01:28 AM
I have met him yeah and that was after leaving ESPN. Had a good talk with him. But yeah, I'd agree with what you said- obviously at ESPN, he'd promote PER. But also, I still believe what I said earlier.



Fair enough. I do think this is the future of analytics though.



Honestly, I haven't seen much on WP in the APBR forum. No one from Berri's crowd ever bothers showing their face there and so no one does any analysis on it (i'm talking about the real analysts, not the people who just read everything and don't do anything). In any case, the best way to assess a metric is a simple retrodiction contest. How does the metric do when predicting? Unless that's not the intent of WP, if so, then I'm not really sure to the usefullness since NBA decision makers are interested in prediction.

And yeah, he does seem like a bit of a jerk. But then I don't know him, so I can't really say. But in regards to the peer reviewed professionals, I think there's a bit of a disconnect there. Whose the audience? Academics or NBA decision makers? I would think ultimately these metrics are designed for NBA decision makers. Also, I think Berri gives off the impression that scouting, video analysis doesn't matter, which is why I assume he's never gotten a consulting job with a team (unlike other pioneers like Dean Oliver, Ben Alamar, etc.). But thats speculation on my part, so not fair to him.

As I said though, I have got to meet Arturo, whose one of the primary wagesofwins guys. And he seems to be a nice guy, certainly puts out quality work and I think our views on mid-range shots probably align (though I don't know if I'd go as far, still think the player matters).

FYI, I believe in the retrodiction contest that an independent site ran, Myer's ASPM was the most predictive. RAPM did pretty well too I believe (tho not sure, have to find the link). Anyways, think our new statistical RAPM will improve the metric but we'll see.

Supposedly Berri has gotten offers from teams, but turned them down because he would have to sign a non-disclosure agreement, meaning he couldn't publish. One of the biggest concerns at Sloan this year, and of the WoW crowd, is that future analysis will be proprietary to the teams that develop them. Data will be held under lock and key and while certain teams would benefit, the field as a whole will virtually cease to exist. If you read about (or attended) Sloan this year, several of the presentations were proprietary, so they really couldn't talk in depth about what they were doing, let alone get feedback from educated peers who could improve their work. And more to the point of this conversation, we'll still be stuck with layfans thinking PER is cutting edge. Because while there are and will be much better metrics out there, the guys who developed them can't TALK about it.

PatsSoxKnicks
03-30-2014, 11:39 AM
Supposedly Berri has gotten offers from teams, but turned them down because he would have to sign a non-disclosure agreement, meaning he couldn't publish. One of the biggest concerns at Sloan this year, and of the WoW crowd, is that future analysis will be proprietary to the teams that develop them. Data will be held under lock and key and while certain teams would benefit, the field as a whole will virtually cease to exist. If you read about (or attended) Sloan this year, several of the presentations were proprietary, so they really couldn't talk in depth about what they were doing, let alone get feedback from educated peers who could improve their work. And more to the point of this conversation, we'll still be stuck with layfans thinking PER is cutting edge. Because while there are and will be much better metrics out there, the guys who developed them can't TALK about it.

I did attend Sloan. Did you? Kind of curious if anyone else here on PSD did.

Interesting, didn't know that about Berri. Though I'm not sure that I believe it. Arturo certainly was all ready to get his hands on the Vantage data (which requires an NDA). I think you definitely feel differently about that if you're actually in possession of the proprietary data. And for what it's worth, Vantage did try to make their data public but no one ponied up for the kickstarter (maybe their goal was a bit lofty though).

But yeah, as I said earlier- tons of secrecy among teams. I remember this one conversation I had at Sloan where a Celtics employee wouldn't even talk about which mathematical techniques he uses when evaluating a model because well a) he probably had no interest in sharing that with me but b) a Hawks employee was also there. It's kind of crazy the lengths to which teams go in regards to secrecy.

Edit: BTW, I think there's this perception from APBR people that the WoW crowd is against film/scouting use. And to be fair, it does kind of seem like they give that vibe off. Maybe not but seems that way.

IndyRealist
03-30-2014, 05:46 PM
I did attend Sloan. Did you? Kind of curious if anyone else here on PSD did.

Interesting, didn't know that about Berri. Though I'm not sure that I believe it. Arturo certainly was all ready to get his hands on the Vantage data (which requires an NDA). I think you definitely feel differently about that if you're actually in possession of the proprietary data. And for what it's worth, Vantage did try to make their data public but no one ponied up for the kickstarter (maybe their goal was a bit lofty though).

But yeah, as I said earlier- tons of secrecy among teams. I remember this one conversation I had at Sloan where a Celtics employee wouldn't even talk about which mathematical techniques he uses when evaluating a model because well a) he probably had no interest in sharing that with me but b) a Hawks employee was also there. It's kind of crazy the lengths to which teams go in regards to secrecy.

Edit: BTW, I think there's this perception from APBR people that the WoW crowd is against film/scouting use. And to be fair, it does kind of seem like they give that vibe off. Maybe not but seems that way.

What I've read from them is that people need to be wary of thinking that more data = better data. Without proper analysis having massive amounts of data to sift through will only obfuscate the important data in a sea of noise.

One article on SportVU, which I can attempt to find, attempted to show that Brook Lopez was a better rebounder than Reggie Evans. The analysis was that Brook Lopez travels farther for each rebound, by a substantial margin, than Evans meaning he covers a wider area. Which is exactly the wrong interpretation. In fact, Lopez travels further per rebound because he's out of position and so has to chase them down, whereas Evans blocks out and has a general notion of which way the rebound will bounce, and so he's in that area before the ball even hits the rim. Lopez has other responsibilities besides rebounding, to be sure, but suggesting that he's a better rebounder than Reggie Evans is exactly the kind of mistakes having too much data and too little analysis can cause.

Edit: here is the article http://hangtime.blogs.nba.com/2013/09/16/sportvu-adds-to-the-conversation/

Want to talk about rebounding? SportVU will tell you how many rebounding chances a player had, how many of his rebounds were contested or uncontested, and how much distance he travels for his rebounds. Reggie Evans led the league in rebounding percentage (the percentage of available rebounds that he grabbed while he was on the floor), but teammate Brook Lopez (in 18 games tracked by SportVU) actually converted a greater percentage of his rebound chances (63 percent vs. 62 percent) where he was in the vicinity of the ball. Furthermore, 54 percent of Lopez’s rebounds were contested, while only 31 percent of Evans’ were. And Lopez traveled 6.4 feet per rebound, while Evans traveled just 4.3 feet.