Can someone explain PER to me.
Moderators: bwgood77, zimpy27, infinite11285, Clav, Domejandro, ken6199, bisme37, Dirk, KingDavid, cupcakesnake
Can someone explain PER to me.
- Hendrix
- RealGM
- Posts: 17,030
- And1: 3,662
- Joined: May 30, 2007
- Location: London, Ontario
Can someone explain PER to me.
Actually I want to know what the problems with the PER stat are mostly. I was looking up stats, and noticed Calderon is 3rd out of all PG's in PER, and 13th out of all players in PER. While I think Calderon has played great for us this year I don't think he's near that good. I mean he's not better then a lot of the players under him.
So just wondering what the problems with that statistic are since I see a lot of people say that it has it's faults, but don't really know what they are.
So just wondering what the problems with that statistic are since I see a lot of people say that it has it's faults, but don't really know what they are.
oak2455 wrote:Do understand English???
-
- Banned User
- Posts: 1,721
- And1: 2
- Joined: Nov 19, 2007
- Location: Mr.Ferguson, meet Donovan. Donovan, Robert's neck.
-
- Banned User
- Posts: 1,721
- And1: 2
- Joined: Nov 19, 2007
- Location: Mr.Ferguson, meet Donovan. Donovan, Robert's neck.
- Yao_noodle
- Junior
- Posts: 429
- And1: 0
- Joined: Dec 28, 2004
-
- Forum Mod
- Posts: 52,226
- And1: 6,100
- Joined: Oct 31, 2004
- Location: Getting hit in the head
-
Regarding PER
-
- Forum Mod - Raptors
- Posts: 91,176
- And1: 30,870
- Joined: Oct 14, 2003
-
Regarding PER
PER is a descriptive stat; it has no ability to effectively convey information about man-on defense and it describes help defense only weakly.
But at its heart, PER is a summary of a player's contribution to the box score, rated and manipulated so as to convey as much subtext and context as possible. It is a flawed stat, like EVERY stat, but it is also often misunderstood and dealt with in the context of the inexplicable reaction to what is described as John Hollinger's "arrogance" or "presumption" or any of a variety of such terms.
Hollinger has provided a meaningful statistical tool and bases all of his work off of it because he is a statistician and that's how they work; you need to take it with a grain of salt but it has a lot more value than simple things like Tendex or the Lenovo efficiency stat or any such things.
The vast majority of the "contributions" of responses above mine reflect the sort of mass ignorance that plagues the basketball community on this topic.
Hollinger himself routinely and publicly acknowledges the weaknesses of his stat and the dependence of his work on PER; he's a niche analyst, no different than any other.
Do you like Bill Simmons or Marc Stein more? They're gimmick analysts just the same, only basing their commentary on subjective concepts rather than hard data, misleading as that can sometimes be.
But at its heart, PER is a summary of a player's contribution to the box score, rated and manipulated so as to convey as much subtext and context as possible. It is a flawed stat, like EVERY stat, but it is also often misunderstood and dealt with in the context of the inexplicable reaction to what is described as John Hollinger's "arrogance" or "presumption" or any of a variety of such terms.
Hollinger has provided a meaningful statistical tool and bases all of his work off of it because he is a statistician and that's how they work; you need to take it with a grain of salt but it has a lot more value than simple things like Tendex or the Lenovo efficiency stat or any such things.
The vast majority of the "contributions" of responses above mine reflect the sort of mass ignorance that plagues the basketball community on this topic.
Hollinger himself routinely and publicly acknowledges the weaknesses of his stat and the dependence of his work on PER; he's a niche analyst, no different than any other.
Do you like Bill Simmons or Marc Stein more? They're gimmick analysts just the same, only basing their commentary on subjective concepts rather than hard data, misleading as that can sometimes be.
- nate33
- Forum Mod - Wizards
- Posts: 69,865
- And1: 22,281
- Joined: Oct 28, 2002
PER is an honest attempt at assimilating all of the box score data and giving each stat a proper weighting. For example, a missed shot doesn't hurt your PER as much as a turnover because about 28% of missed shots end up in offensive rebounds. PER also adjusts for minutes and pace. PER is a very good, very accurate way of summarizing the statistical production of a player. (It has some minor flaws, but they pale in comparison to the huge flaws of other summary stats like Tendex and Efficiency.)
The problem with PER is that it is based on box score stats, and box score stats don't incorporate everything about the game. For example, box scores don't tell you who is a good position defender; they only tell you who gambles more for steals and blocks.
PER is a tool. It's a pretty good starting point for player comparisons because it's an objective standard. But after you look at PER, you then need to think about things that aren't encapsulated in box score data: intangibles, clutchness, compatibility with teammates, defensive rotations, setting picks, making the extra pass, spreading the floor with good outside shooting, etc.
The problem with PER is that it is based on box score stats, and box score stats don't incorporate everything about the game. For example, box scores don't tell you who is a good position defender; they only tell you who gambles more for steals and blocks.
PER is a tool. It's a pretty good starting point for player comparisons because it's an objective standard. But after you look at PER, you then need to think about things that aren't encapsulated in box score data: intangibles, clutchness, compatibility with teammates, defensive rotations, setting picks, making the extra pass, spreading the floor with good outside shooting, etc.
- Hendrix
- RealGM
- Posts: 17,030
- And1: 3,662
- Joined: May 30, 2007
- Location: London, Ontario
tsherkin wrote:PER is a descriptive stat; it has no ability to effectively convey information about man-on defense and it describes help defense only weakly.
But at its heart, PER is a summary of a player's contribution to the box score, rated and manipulated so as to convey as much subtext and context as possible. It is a flawed stat, like EVERY stat, but it is also often misunderstood and dealt with in the context of the inexplicable reaction to what is described as John Hollinger's "arrogance" or "presumption" or any of a variety of such terms.
Hollinger has provided a meaningful statistical tool and bases all of his work off of it because he is a statistician and that's how they work; you need to take it with a grain of salt but it has a lot more value than simple things like Tendex or the Lenovo efficiency stat or any such things.
The vast majority of the "contributions" of responses above mine reflect the sort of mass ignorance that plagues the basketball community on this topic.
Hollinger himself routinely and publicly acknowledges the weaknesses of his stat and the dependence of his work on PER; he's a niche analyst, no different than any other.
Do you like Bill Simmons or Marc Stein more? They're gimmick analysts just the same, only basing their commentary on subjective concepts rather than hard data, misleading as that can sometimes be.
nate33 wrote:PER is an honest attempt at assimilating all of the box score data and giving each stat a proper weighting. For example, a missed shot doesn't hurt your PER as much as a turnover because about 28% of missed shots end up in offensive rebounds. PER also adjusts for minutes and pace. PER is a very good, very accurate way of summarizing the statistical production of a player. (It has some minor flaws, but they pale in comparison to the huge flaws of other summary stats like Tendex and Efficiency.)
The problem with PER is that it is based on box score stats, and box score stats don't incorporate everything about the game. For example, box scores don't tell you who is a good position defender; they only tell you who gambles more for steals and blocks.
PER is a tool. It's a pretty good starting point for player comparisons because it's an objective standard. But after you look at PER, you then need to think about things that aren't encapsulated in box score data: intangibles, clutchness, compatibility with teammates, defensive rotations, setting picks, making the extra pass, spreading the floor with good outside shooting, etc.
Thanks a lot. Answers evreything.
oak2455 wrote:Do understand English???
- floppymoose
- Senior Mod - Warriors
- Posts: 59,232
- And1: 17,329
- Joined: Jun 22, 2003
- Location: Trust your election workers
If you are interested in seeing the limitations of PER, compare it to +-. If you compare a list of the top PER performers over, say, two seasons, with the top +- performers, you will notice differences in the lists. Thinking about those differences can expose the limitations of PER (and also of +-).
- kooldude
- Lead Assistant
- Posts: 4,823
- And1: 78
- Joined: Jul 08, 2007
floppymoose wrote:If you are interested in seeing the limitations of PER, compare it to +-. If you compare a list of the top PER performers over, say, two seasons, with the top +- performers, you will notice differences in the lists. Thinking about those differences can expose the limitations of PER (and also of +-).
any big names?
Warspite wrote:I still would take Mitch (Richmond) over just about any SG playing today. His peak is better than 2011 Kobe and with 90s rules hes better than Wade.
Jordan23Forever wrote:People are delusional.
-
- Assistant Coach
- Posts: 3,812
- And1: 1
- Joined: Dec 16, 2005
Addiing to the discussion:
The problem with PER isn't with the stat itself but with the way people interpret it. Some people seem to think higher PER = better in all cases which can easily be proven untrue. For instance, AI's best season was not the one in which he had the best PER. He won MVP and a lot of games because he adjusted his game radically to fit in with the strengths of the players around him.
The goal of a player is not to get higher PER but to win more games/championship and that many player sacrifice PER in order to accomplish that goal (see T-Mac in Houston, Nash/Finley in Dallas, Joe Johnson in PHX, KG in BOS etc.). Without recognition of that fact people will continue to treat PER as the end-all-and-be-all that it clearly isn't.
The problem with PER isn't with the stat itself but with the way people interpret it. Some people seem to think higher PER = better in all cases which can easily be proven untrue. For instance, AI's best season was not the one in which he had the best PER. He won MVP and a lot of games because he adjusted his game radically to fit in with the strengths of the players around him.
The goal of a player is not to get higher PER but to win more games/championship and that many player sacrifice PER in order to accomplish that goal (see T-Mac in Houston, Nash/Finley in Dallas, Joe Johnson in PHX, KG in BOS etc.). Without recognition of that fact people will continue to treat PER as the end-all-and-be-all that it clearly isn't.
"I'm sure they'll jump off the bandwagon. Then when we do get back on top, they're going to want to jump back on, and we're going to tell them there's no more room." - Kobe in March of 2005
- xcomputerman
- Pro Prospect
- Posts: 878
- And1: 0
- Joined: Dec 21, 2005
- Location: Houston, TX
- Contact:
Bgil wrote:Addiing to the discussion:
The problem with PER isn't with the stat itself but with the way people interpret it. Some people seem to think higher PER = better in all cases which can easily be proven untrue. For instance, AI's best season was not the one in which he had the best PER. He won MVP and a lot of games because he adjusted his game radically to fit in with the strengths of the players around him.
I think that boils down to people not really knowing what PER is. The casual fan sees Player X PER > Player Y PER ==> Player X > Player Y. The correct inference is that Player X is more productive on a per-minute basis than Player Y. I like the descriptions that tsherkin and nate33 gave, and like you rightly pointed out, a player doesn't have to put up a maximum PER stat for his team to win, and being as heavily box score-based as it is, it will never sufficiently highlight a player's intangibles (which is where +/- and on/off court statistics shine, for the most part).
The great thing about PER, though, is that with a sufficient enough sample size (in terms of minutes played), it almost always gives you a good measure of how "good" a player is, so for the OP, when you see Jose Calderon has a PER of 23.21, you had better believe it: he is one of the best point guards in the league.
http://fourpointers.com - We've got unlimited range, baby
- Ryoga Hibiki
- RealGM
- Posts: 12,430
- And1: 7,650
- Joined: Nov 14, 2001
- Location: Warszawa now, but from Northern Italy
I hate PER, because I think it oversimplifies the analysis trying to put together all the boxscore stats that are already inaccurate enought to evaluate a player.
I've yet to understand what PER measures and how to test it, honestly.
BTW, how does it work?
1) it's based on per minutes contributions, so 20-10 in 30 min is just as good as 10-5 in 15 min
2) measures the difference from the mean, you know 15 is the average contribution that season, so when you're comparing different players in different seasons their PER (wants to) measures how much they distanced themselves from their peers
3) it takes every stat in the boxscore and translates it in a change of the score. It starts calculating the average value of a possession (VOP, how many points are scored on average in a possession), so in this way gives a value to steals, turnovers (full VOP, change of possession), defensive and offensive rebounds (VOP - the probability a teamate would have gotten the rebound), missed shots (VOP - the probability to get an ORbd), etc
4) it becomes tricky when assists enter the discussion, cause you need to split the credit between the shooter and the passer. This is a very controversial point
5) points are given to the scorer minues a % that goes to the passer
This is how it works, roughly, if you ask me I think it overvalues scorers and bad defenders who gamble for steals and blocks.
Most important, using one number you lose the perspective on how each stat contributed to that and, so, you can't anymore use other data (like watching games) to properly weight it.
I've yet to understand what PER measures and how to test it, honestly.
BTW, how does it work?
1) it's based on per minutes contributions, so 20-10 in 30 min is just as good as 10-5 in 15 min
2) measures the difference from the mean, you know 15 is the average contribution that season, so when you're comparing different players in different seasons their PER (wants to) measures how much they distanced themselves from their peers
3) it takes every stat in the boxscore and translates it in a change of the score. It starts calculating the average value of a possession (VOP, how many points are scored on average in a possession), so in this way gives a value to steals, turnovers (full VOP, change of possession), defensive and offensive rebounds (VOP - the probability a teamate would have gotten the rebound), missed shots (VOP - the probability to get an ORbd), etc
4) it becomes tricky when assists enter the discussion, cause you need to split the credit between the shooter and the passer. This is a very controversial point
5) points are given to the scorer minues a % that goes to the passer
This is how it works, roughly, if you ask me I think it overvalues scorers and bad defenders who gamble for steals and blocks.
Most important, using one number you lose the perspective on how each stat contributed to that and, so, you can't anymore use other data (like watching games) to properly weight it.
- Hendrix
- RealGM
- Posts: 17,030
- And1: 3,662
- Joined: May 30, 2007
- Location: London, Ontario
Alright I pretty much get what everyone's saying. PER is a measure of statistics, but it leaves out a lot of analysis.
For example If someone were to say Player X is a PF that gets 10 rebounds a game, but the center, and SF on his team are terrible rebounders. Player Y gets 8 rebounds but his center, and SF are the best rebounders at their positions in the league. Therefor you can't conclude Player X > Player Y in rebounding.
PER doesn't really allow you to analize why 1 player might be higher or lower. It's just a bunch of stats that don't take into account variables compiled together to form 1 big stat that could be off. There could be some mis leading stats like the one I posted above all contributing to give a player X a higher PER then Player Y even though it isn't justified. When looking at PER you can't really disect it, and find out any real conclusions from it. Then throw in defence, statistical defense, making the pass that leads to the assist etc.... And it can be misleading. From a purely statistical point it's decent though no?
For example If someone were to say Player X is a PF that gets 10 rebounds a game, but the center, and SF on his team are terrible rebounders. Player Y gets 8 rebounds but his center, and SF are the best rebounders at their positions in the league. Therefor you can't conclude Player X > Player Y in rebounding.
PER doesn't really allow you to analize why 1 player might be higher or lower. It's just a bunch of stats that don't take into account variables compiled together to form 1 big stat that could be off. There could be some mis leading stats like the one I posted above all contributing to give a player X a higher PER then Player Y even though it isn't justified. When looking at PER you can't really disect it, and find out any real conclusions from it. Then throw in defence, statistical defense, making the pass that leads to the assist etc.... And it can be misleading. From a purely statistical point it's decent though no?
oak2455 wrote:Do understand English???
- Basileus777
- General Manager
- Posts: 7,819
- And1: 2,050
- Joined: Jul 13, 2007
- Location: New Jersey
-
Bgil wrote:Addiing to the discussion:
The problem with PER isn't with the stat itself but with the way people interpret it. Some people seem to think higher PER = better in all cases which can easily be proven untrue. For instance, AI's best season was not the one in which he had the best PER. He won MVP and a lot of games because he adjusted his game radically to fit in with the strengths of the players around him.
The goal of a player is not to get higher PER but to win more games/championship and that many player sacrifice PER in order to accomplish that goal (see T-Mac in Houston, Nash/Finley in Dallas, Joe Johnson in PHX, KG in BOS etc.). Without recognition of that fact people will continue to treat PER as the end-all-and-be-all that it clearly isn't.
This is all true, the problem is that Hollinger himself doesn't seem to recognize this. Hollinger regularly uses PER as a ranking system and as the sole judge of a player's worth.
PER is probably the best summary stat, but summary stats as a whole are deeply flawed.
- nate33
- Forum Mod - Wizards
- Posts: 69,865
- And1: 22,281
- Joined: Oct 28, 2002
Basileus777 wrote:-= original quote snipped =-
This is all true, the problem is that Hollinger himself doesn't seem to recognize this. Hollinger regularly uses PER as a ranking system and as the sole judge of a player's worth.
PER is probably the best summary stat, but summary stats as a whole are deeply flawed.
I'm sure Hollinger recognizes the weaknesses of PER. I think he leans so heavily on his PER ranking system in an effort to be a contrarian. He's trying to have a different take than the conventional wisdom. By sticking to PER, he is being completely objective in his analysis so that he is not influenced by his prejudices.
I'm sure he recognizes that David Lee is not better than Rasheed Wallace.
- Ryoga Hibiki
- RealGM
- Posts: 12,430
- And1: 7,650
- Joined: Nov 14, 2001
- Location: Warszawa now, but from Northern Italy
nate33 wrote:-= original quote snipped =-
I'm sure Hollinger recognizes the weaknesses of PER. I think he leans so heavily on his PER ranking system in an effort to be a contrarian. He's trying to have a different take than the conventional wisdom. By sticking to PER, he is being completely objective in his analysis so that he is not influenced by his prejudices.
I'm sure he recognizes that David Lee is not better than Rasheed Wallace.
well, I'm not so sure once you consider his articles and the way he justifies the numbers.
My feeling is that he considers his stat very accurate to measure offensive contributions, maybe a bit flawed against pure PGs, unable to take care of individual defence (Bowen example always comes out) and, working per minutes, helps players with poor stamina masking their flaw.
I'm expecting an answer like "Rasheed is a much better defender", not sure he'd consider Rasheed a better offensive player.
And if he did, than I wish he he explained me what the heck his stat his measuring.
And I've not even started discussing the merits of his assumptions, like on how he doesn't take away even a fraction of a VOP for every hit shot, prizing inefficient scorers, or how a FT point is worth like 30% more than one coming from a FG...