PER Explanation/Rondo

[vinnie]

I am glad I am an older guy and pay no attention to all of these computerized statistics. I see what I see and that is good enough for a middle-aged guy like me 

[Eja117]

I'd need to see things like what Bird and Jordan's pers were.  And Kevin McHale. and DJ.

Jordan is #1 in career PER.  LeBron is #2 and Shaq is #3.  Bird is 19, between KG and Oscar Robertson.  Paul Pierce 50.  Ed Macauley 55.  McHale 64.  Cousy 72. Ray Allen 81. Robert Parish 82. Bill Russell is 96, but it is admitted by Hollinger that defensive specialists are underrated by PER.  If Rondo had enough games for the career leader list on basketball-reference.com, he would be ranked in the 160s.  DJ isn't in the top 250.

How about Dennis Rodman and the Superloser Patrick Ewing? How about Scottie Pippen and Toine?

[LooseCannon]

I'd need to see things like what Bird and Jordan's pers were.  And Kevin McHale. and DJ.

Jordan is #1 in career PER.  LeBron is #2 and Shaq is #3.  Bird is 19, between KG and Oscar Robertson.  Paul Pierce 50.  Ed Macauley 55.  McHale 64.  Cousy 72. Ray Allen 81. Robert Parish 82. Bill Russell is 96, but it is admitted by Hollinger that defensive specialists are underrated by PER.  If Rondo had enough games for the career leader list on basketball-reference.com, he would be ranked in the 160s.  DJ isn't in the top 250.

How about Dennis Rodman and the Superloser Patrick Ewing? How about Scottie Pippen and Toine?

Toine is at 249 among NBA players, but falls out of the top 250 when you do combined ABA/NBA.  All rankings I gave were from the combined list where Pippen is #106, Ewing is #45, and Rodman is not on the list.

You can look up other names here.

[winsomme]

Full disclosure, I consider myself a huge fan of sabremetrics. I think that some advanced stats can say much more than your run of the mill PPG, FG% etc.

I consider myself a bit perplexed by Hollinger's PER - and I am a huge fan of his.

PER Rankings this year....

http://insider.espn.go.com/nba/hollinger/statistics?&action=login&appRedirect=http://insider.espn.go.com/nba/hollinger/statistics

Some things make sense: The top 5 in PER this year in the NBA are LeBron, Chris Paul, Dwight Howard, Wade and Kobe. There is a reasonable argument to be made that those five, in that order, are the best players in the NBA.

However, go further down the list. Rajon Rondo clocks in at #47. Players with better PERs than Rajon: Devin Harris, Elton Brand, Brook Lopez, Tyrus Thomas, Andre Miller, Nene and Ryan Anderson.

Those players aren't slouches by any means. However, judging from watching any of these players this season, have any of them played close to Rondo's level? I don't think so.

It strikes me as odd that Rondo is a bonafide All-Star and probably a top-20 player in the league and he can't crack the top 45 in PER. What about him doesn't translate?

the problem with advanced stats is that we ultimately decide whether a particular stat is useful  based on our overall subjective "sense."

Our perception trumps the stat and that's honestly how it should be. If you follow Rondo and the Cs you know that Rondo is vital to the team's success.

If you want to work backwards from there to determine that a stat is useful (ie wages of wins, etc), that's fine. but you're still working off your perception.

[Ersatz]

I won't bash Hollinger, because I like him and what he's trying to do. But it is worth pointing out that there's hardly an objective universal standard for these things. In essence, he already decided that MJ was the greatest player ever, so he devised PER to show how everyone measured as a distance from him. Similarly, he devised the Power Rankings at a time when the Spurs model of doing things was winning championships, so they became the de facto historical standard.

Of course I think his system doesn't capture what the Celtics do well. I think PER is predicated on measuring the contributions of super-duper stars operating more or less as the alpha dog on their team. Jordan and LeBron and Wade have an advantage because they have never been on teams with real point guards, so their usage rates and opportunities to distribute are not available to players like Paul and Ray, who play with a real point guard. In short, PER doesn't capture the value of players playing in a system. If you don't average 18 shots a game or have a 40% usage rate, you are not going to be a PER high earner.

PER doesn't measure a team that has a bunch of different options and takes turns exploiting their matchup advantages. And, as others have said, it undervalues the contributions made by defense.

[Ersatz]

Just poking around a bit on the all-time PER numbers: David Robinson is #4; Tim Duncan is #8. The Admiral was a great player, but is there anyone who knows anything about basketball who would, given an abstract choice between the two, choose Robinson over Duncan? I doubt it. (I could substitute Olajuwon, #15 all-time PER, for Duncan, and I still don't know that anyone would seriously pick Robinson.)

More relevantly to our beloved Cs, Vince Carter is #48, while Pierce is #50. Would anyone, not just Cs fans, ever pick VC over PP? Um, no.

Again, this isn't to bash Hollinger. It's just to say that PER is a blunt instrument and should be used with caution.

[LooseCannon]

Just poking around a bit on the all-time PER numbers: David Robinson is #4; Tim Duncan is #8. The Admiral was a great player, but is there anyone who knows anything about basketball who would, given an abstract choice between the two, choose Robinson over Duncan? I doubt it. (I could substitute Olajuwon, #15 all-time PER, for Duncan, and I still don't know that anyone would seriously pick Robinson.)

Keep in mind that PER is a rate based stat.  David Robinson played fewer games than Duncan or Olajawon, so his career has less value.  Who would you take between the three if you had to choose who had the best three- or four- or five- year peak?

You should also keep in mind that Robinson's best years statistically were before Tim Duncan was a Spur, so you can't judge their relative worth solely on who was better based on their time playing together.  The Admiral was already in his 30s when Duncan was drafted.

WS/48 is a pretty interesting stat.  It puts a bit more separation between Paul Pierce and Vince Carter.  It has Pierce as a somewhat better player on offense, but more than 75% of the difference between the two comes from PP's defensive superiority.  Of course, this stat has David Robinson at #2 all-time and Manu Ginobeli as tops among active players.

[BballTim]

Just poking around a bit on the all-time PER numbers: David Robinson is #4; Tim Duncan is #8. The Admiral was a great player, but is there anyone who knows anything about basketball who would, given an abstract choice between the two, choose Robinson over Duncan? I doubt it. (I could substitute Olajuwon, #15 all-time PER, for Duncan, and I still don't know that anyone would seriously pick Robinson.)

Keep in mind that PER is a rate based stat.  David Robinson played fewer games than Duncan or Olajawon, so his career has less value.  Who would you take between the three if you had to choose who had the best three- or four- or five- year peak?

You should also keep in mind that Robinson's best years statistically were before Tim Duncan was a Spur, so you can't judge their relative worth solely on who was better based on their time playing together.  The Admiral was already in his 30s when Duncan was drafted.

WS/48 is a pretty interesting stat.  It puts a bit more separation between Paul Pierce and Vince Carter.  It has Pierce as a somewhat better player on offense, but more than 75% of the difference between the two comes from PP's defensive superiority.  Of course, this stat has David Robinson at #2 all-time and Manu Ginobeli as tops among active players.

  Manu's numbers are inflated because he came into the league when he was 25.

[drza44]

Just poking around a bit on the all-time PER numbers: David Robinson is #4; Tim Duncan is #8. The Admiral was a great player, but is there anyone who knows anything about basketball who would, given an abstract choice between the two, choose Robinson over Duncan? I doubt it. (I could substitute Olajuwon, #15 all-time PER, for Duncan, and I still don't know that anyone would seriously pick Robinson.)

More relevantly to our beloved Cs, Vince Carter is #48, while Pierce is #50. Would anyone, not just Cs fans, ever pick VC over PP? Um, no.

Again, this isn't to bash Hollinger. It's just to say that PER is a blunt instrument and should be used with caution.

1) I agree with you, that PER is a blunt instrument and can't/shouldn't be used as a stand-alone rater.  There is enough variety out there that measures the game from different angles to get a much better picture with stats than you can get with PER alone, in addition to the obvious non-statistical analysis.

2) But, that said, your Robinson/Duncan/Olajuwon example doesn't really work because you were looking at regular season PER records...and in the regular season, Robinson likely WAS better than Duncan and Olajuwon.  He produced in the regular season in ways that Duncan and Hakeem never did, and he flat-out dominated the head-to-head battles with Hakeem in the regular season (similar overall stats, but Robinson shot a much higher percentage and his teams beat Hakeem's 30 - 12 in their 42 H-2-H matchups).

No, where Duncan and Olajuwon move above Robinson in most people's rankings is in the postseason where they were much more productive and had better success than DRob ever did.  And this is actually reflected in the postseason PER, where Duncan and Olajuwon are 5th and 6th all-time with Robinson in the teens.

Again, I don't think PER should ever be used as a standalone rater.  But this wasn't really a great example to illustrate that.

[footey]

PER tends to undervalue defense, especially perimeter defense, and overvalue volume scorers.  Some also think it doesn't value rebounding enough.  Hollinger freely admits that his formula undervalues defensive specialists.

As a guard who has a lot of value tied up in his defense and who doesn't take a lot of shots and especially doesn't take a lot of three-point shots, Rondo seems to be the sort of player who will be undervalued by PER.

PER is an attempt to do a quick method of distilling box score data into a single number for the purpose of comparison.  This thread is probably going to become another bash Hollinger thread, but it should be noted that insisting that a better formula should incorporate data not found in conventional box scores is probably an unfair criticism.

PER seems to undervalue assists too. Rondo is league leader, yet does not get much value in his ranking.