Chris Forsberg on Rondo's defense

[D.o.s.]

It is hard to understand why Rondo is more valuable to the Celtics than other teams, the team seems to win more without him....32nd best and there are not even 32 teams in the league

Ok. So. Let's break this down.

What is PER?

This is PER:

uPER = (1 / MP) *
     [ 3P
     + (2/3) * AST
     + (2 - factor * (team_AST / team_FG)) * FG
     + (FT *0.5 * (1 + (1 - (team_AST / team_FG)) + (2/3) * (team_AST / team_FG)))
     - VOP * TOV
     - VOP * DRB% * (FGA - FG)
     - VOP * 0.44 * (0.44 + (0.56 * DRB%)) * (FTA - FT)
     + VOP * (1 - DRB%) * (TRB - ORB)
     + VOP * DRB% * ORB
     + VOP * STL
     + VOP * DRB% * BLK
     - PF * ((lg_FT / lg_PF) - 0.44 * (lg_FTA / lg_PF) * VOP) ]

And then you take the uPER, or unadjusted PER, and multiply it by the league's pace adjustment:

pace adjustment = lg_Pace / team_Pace

and then multiply that number by (15 divided by the league's ajusted PER).
http://www.basketball-reference.com/about/per.html

Or, if you want, you can look at it like this:

[ FGM x 85.910

+ Steals x 53.897

+ 3PTM x 51.757

+ FTM x 46.845

+ Blocks x 39.190

+ Offensive_Reb x 39.190

+ Assists x 34.677

+ Defensive_Reb x 14.707

- Foul x 17.174

- FT_Miss x 20.091

- FG_Miss x 39.190

- TO x 53.897 ]

x (1 / Minutes).

Courtesy of one of the very few Bleacher Report articles worth reading (although that measurement doesn't adjust for pace, which actually affects the final PER rankings very little):
http://bleacherreport.com/articles/113144-cracking-the-code-how-to-calculate-hollingers-per-without-all-the-mess

So. If you look at the way PER works, you'll see that the things Rondo is good at, like rebounding and assists, are about as valuable to his final score as a missed shot. Does that make any sense to you? Do any of the relative weights given to any of the stats used in deriving PER seem anything other that arbitrary?

The answer to both of those should be "no." And, because the answer to both of those is "no," it would follow that using PER as any sort of list to determined whether Rondo is the 32nd or 400th or 5th best player in the league is completely and utterly insipid. QED.


edit: valuable here is used in the sense of weight, not in the sense of a positive contribution to his score.

[PickNRoll]

Playing with PER can actually be pretty fun.  Can we PLEASE sign this guy?



He ought to be really cheap with that PER!
(all per 36 minutes)

[mmmmm]

PER's a pretty flawed stat, but is there any advanced stat that tries to aggregate total value and doesn't have some of those oddball Barea/Shved type results?

If you have to use one, use Win Shares (or Win Shares per 48) with the qualifier for the Minutes Played leaderboard over the course of a season, or at least half of one, to get rid of some of the noise.

This is the list from last season:
http://bkref.com/tiny/LTQgN

That's a reasonable suggestion, but personally I'm not a fan of ANY of the attempts to 'roll up' the assessment of basketball players into one aggregated number.

The basic problem is this inevitably requires either an arbitrary weighting of elements or a regression fit to a pre-conception of relative merit (the latter being the problem with PER).

I don't mind using WS or PER or WOW numbers when comparing players within well-defined roles, such a their relative PER at a position on a given team.    But I think they break down rapidly when you try to compare players with vastly different roles on different teams.

Let's talk about PER for a minute:

PER claims to be an efficiency stat because it is normalized for minutes.  But it is not normalized for touch rate and so it blows up for players who get high touch rates compared to those with low touch rates.  Last year, for whatever reason, we gave Kris Humphies a ton of touches every minute he was ON the floor.   His touches/minute was something like 30% higher than Jeff Green's!   As a result of many touches, a player is naturally going to record more basketball events (even if he is not especially efficient with his touches).  Most importantly, more shots and thus more makes (efficient or not).   PER starts literally by incorporating the counts of all those events.  When you then divide by the minutes, then ... presto!  Kris Humphries ended up with a very nice PER of 18.2 -- THE HIGHEST ON THE TEAM!!!!

Who really believes that Kris Humphries was the very best player on the Celtics last season?  Anybody?  Anybody? Bueller?

Very useful, I suppose, if you are prepping a player for trade.

[Sixth Man]

PER's a pretty flawed stat, but is there any advanced stat that tries to aggregate total value and doesn't have some of those oddball Barea/Shved type results?

I just made a new one.  It's called PNR, for PickNRoll.  It's

points + rebounds + assists + (age/3)

I had to add age to make sure that Lebron was better than Kevin Love.  Here are the results for 2013-2014.  What do you guys think?



Did I just dethrone PER?

PER's a pretty flawed stat, but is there any advanced stat that tries to aggregate total value and doesn't have some of those oddball Barea/Shved type results?

If you have to use one, use Win Shares (or Win Shares per 48) with the qualifier for the Minutes Played leaderboard over the course of a season, or at least half of one, to get rid of some of the noise.

This is the list from last season:
http://bkref.com/tiny/LTQgN

Interesting, thanks.  I still see some odd results - like LaMarcus Aldridge being Portland's worst starter and Josh McRoberts and Chris Anderson over Tony Parker - but it seems like a decent fit overall. 

Every catch-all metric is flawed. That's sort of the point -- that one just happens to be the one that gets things wrong least often, in my experience.

Yes, terrific stuff.  The map will never be the same thing as the territory it represents, so we look for the one that is most accurate (or least inaccurate) to utilize.

The real problem is when we confuse the map with the territory, as if they were the same thing.

[bucknersrevenge]

Interesting though even Rondo acknowledged before the game that he hasn't been playing well defensively. He said he wasn't getting up into guys and wasn't fighting through the picks on defense enough. Said that as the Captain it had to begin with him. He is going to take more of the blame than the other players because that's comes with the job and it's true. And it wouldn't surprise me if film sessions with Stevens maybe pointed a few of those things out. People are rushing to Rondo's defense like it's sacrilege to suggest that his defense hasn't been good. Whatever the stat you want to use or not use, it is not without merit based what we've seen in the 4th quarter of games lately that Rondo has been a less effective PG on both ends of the pick n roll offensively and defensively. Maybe he will fix it. Maybe he won't.

[Chris22]

Rondo is not the player he once was.
He has gotten worse, much worse.

PER's a pretty flawed stat, but is there any advanced stat that tries to aggregate total value and doesn't have some of those oddball Barea/Shved type results?

If you have to use one, use Win Shares (or Win Shares per 48) with the qualifier for the Minutes Played leaderboard over the course of a season, or at least half of one, to get rid of some of the noise.

This is the list from last season:
http://bkref.com/tiny/LTQgN

That's a reasonable suggestion, but personally I'm not a fan of ANY of the attempts to 'roll up' the assessment of basketball players into one aggregated number.

The basic problem is this inevitably requires either an arbitrary weighting of elements or a regression fit to a pre-conception of relative merit (the latter being the problem with PER).

I don't mind using WS or PER or WOW numbers when comparing players within well-defined roles, such a their relative PER at a position on a given team.    But I think they break down rapidly when you try to compare players with vastly different roles on different teams.

Let's talk about PER for a minute:

PER claims to be an efficiency stat because it is normalized for minutes.  But it is not normalized for touch rate and so it blows up for players who get high touch rates compared to those with low touch rates.  Last year, for whatever reason, we gave Kris Humphies a ton of touches every minute he was ON the floor.   His touches/minute was something like 30% higher than Jeff Green's!   As a result of many touches, a player is naturally going to record more basketball events (even if he is not especially efficient with his touches).  Most importantly, more shots and thus more makes (efficient or not).   PER starts literally by incorporating the counts of all those events.  When you then divide by the minutes, then ... presto!  Kris Humphries ended up with a very nice PER of 18.2 -- THE HIGHEST ON THE TEAM!!!!

Who really believes that Kris Humphries was the very best player on the Celtics last season?  Anybody?  Anybody? Bueller?

Very useful, I suppose, if you are prepping a player for trade.

Yes, Humphries was the best player on the Celtics last year.

And he is helping Washington to a winning record this year.

Humphries had 20 rebounds tonight against the Lakers.

[Nerf DPOY]

Rondo is not the player he once was.
He has gotten worse, much worse.

PER's a pretty flawed stat, but is there any advanced stat that tries to aggregate total value and doesn't have some of those oddball Barea/Shved type results?

If you have to use one, use Win Shares (or Win Shares per 48) with the qualifier for the Minutes Played leaderboard over the course of a season, or at least half of one, to get rid of some of the noise.

This is the list from last season:
http://bkref.com/tiny/LTQgN

That's a reasonable suggestion, but personally I'm not a fan of ANY of the attempts to 'roll up' the assessment of basketball players into one aggregated number.

The basic problem is this inevitably requires either an arbitrary weighting of elements or a regression fit to a pre-conception of relative merit (the latter being the problem with PER).

I don't mind using WS or PER or WOW numbers when comparing players within well-defined roles, such a their relative PER at a position on a given team.    But I think they break down rapidly when you try to compare players with vastly different roles on different teams.

Let's talk about PER for a minute:

PER claims to be an efficiency stat because it is normalized for minutes.  But it is not normalized for touch rate and so it blows up for players who get high touch rates compared to those with low touch rates.  Last year, for whatever reason, we gave Kris Humphies a ton of touches every minute he was ON the floor.   His touches/minute was something like 30% higher than Jeff Green's!   As a result of many touches, a player is naturally going to record more basketball events (even if he is not especially efficient with his touches).  Most importantly, more shots and thus more makes (efficient or not).   PER starts literally by incorporating the counts of all those events.  When you then divide by the minutes, then ... presto!  Kris Humphries ended up with a very nice PER of 18.2 -- THE HIGHEST ON THE TEAM!!!!

Who really believes that Kris Humphries was the very best player on the Celtics last season?  Anybody?  Anybody? Bueller?

Very useful, I suppose, if you are prepping a player for trade.

Yes, Humphries was the best player on the Celtics last year.

And he is helping Washington to a winning record this year.

Humphries had 20 rebounds tonight against the Lakers.

[mmmmm]

Rondo is not the player he once was.
He has gotten worse, much worse.

PER's a pretty flawed stat, but is there any advanced stat that tries to aggregate total value and doesn't have some of those oddball Barea/Shved type results?

If you have to use one, use Win Shares (or Win Shares per 48) with the qualifier for the Minutes Played leaderboard over the course of a season, or at least half of one, to get rid of some of the noise.

This is the list from last season:
http://bkref.com/tiny/LTQgN

That's a reasonable suggestion, but personally I'm not a fan of ANY of the attempts to 'roll up' the assessment of basketball players into one aggregated number.

The basic problem is this inevitably requires either an arbitrary weighting of elements or a regression fit to a pre-conception of relative merit (the latter being the problem with PER).

I don't mind using WS or PER or WOW numbers when comparing players within well-defined roles, such a their relative PER at a position on a given team.    But I think they break down rapidly when you try to compare players with vastly different roles on different teams.

Let's talk about PER for a minute:

PER claims to be an efficiency stat because it is normalized for minutes.  But it is not normalized for touch rate and so it blows up for players who get high touch rates compared to those with low touch rates.  Last year, for whatever reason, we gave Kris Humphies a ton of touches every minute he was ON the floor.   His touches/minute was something like 30% higher than Jeff Green's!   As a result of many touches, a player is naturally going to record more basketball events (even if he is not especially efficient with his touches).  Most importantly, more shots and thus more makes (efficient or not).   PER starts literally by incorporating the counts of all those events.  When you then divide by the minutes, then ... presto!  Kris Humphries ended up with a very nice PER of 18.2 -- THE HIGHEST ON THE TEAM!!!!

Who really believes that Kris Humphries was the very best player on the Celtics last season?  Anybody?  Anybody? Bueller?

Very useful, I suppose, if you are prepping a player for trade.

Yes, Humphries was the best player on the Celtics last year.

And he is helping Washington to a winning record this year.

Humphries had 20 rebounds tonight against the Lakers.

[D.o.s.]

something something facts worldview something something

[Eddie20]

http://espn.go.com/blog/boston/celtics/post/_/id/4716256/rondo-on-criticism-not-weighing-on-me

Rondo has struggled to generate his own offense late in games when some of Boston's younger players tend to defer to him. In 143 minutes of fourth-quarter play this year, Rondo is 12-of-43 shooting (a team-worst 27.9 percent) with 30 rebounds, 42 assists and 19 turnovers. He is minus-47 in plus/minus in the fourth quarter.

[D.o.s.]

http://espn.go.com/blog/boston/celtics/post/_/id/4716256/rondo-on-criticism-not-weighing-on-me

Rondo has struggled to generate his own offense late in games when some of Boston's younger players tend to defer to him. In 143 minutes of fourth-quarter play this year, Rondo is 12-of-43 shooting (a team-worst 27.9 percent) with 30 rebounds, 42 assists and 19 turnovers. He is minus-47 in plus/minus in the fourth quarter.

I like how you bold things for emphasis.

[Eddie20]

http://espn.go.com/blog/boston/celtics/post/_/id/4716256/rondo-on-criticism-not-weighing-on-me

Rondo has struggled to generate his own offense late in games when some of Boston's younger players tend to defer to him. In 143 minutes of fourth-quarter play this year, Rondo is 12-of-43 shooting (a team-worst 27.9 percent) with 30 rebounds, 42 assists and 19 turnovers. He is minus-47 in plus/minus in the fourth quarter.

I like how you bold things for emphasis.

Didn't want you to think I was referring to his stellar rebounding.

Is this is a fair enough sample size? I remember you dismissing it a few weeks ago as a potential slump. It's now a trend.