Manu PER

[ShoogarBear]

Shoogar, can you make some of these posts on PER and statistical analysis on the APBR Board (link I gave above)? Could make for an interesting discussion on how to valuate (or not valuate) players?

(Seriously).

I've been to the APBR board several times (one of the gurus there is a Sonics fan and statistician I've known for almost 20 years), although I don't post there because a lot of the discussion tends to revolve around various tweaks to various formulas, and because I wanted to finish Dean Oliver's book before I waded in there. I got an email from him back in Sept asking me to post some stuff there, so maybe I will.

One of the problems with PER (and all individual evaluations) is that it is based on the idea that you can reduce multi-dimensional assesments (points, rebounds, assists, block) into a single all-encompassing parameter. That makes sense in baseball, because you can reasonably boil everything down into one of two completely separate jobs (either advancing baserunners or preventing that), and where the dependencies on your teammates are relatively minimal. In hoops I think the entire concept is flawed from the start because scoring and defending are interrelated and have an extreme number of teammate dependencies.

Comparison of individual basketall players doesn't make sense unless you want to do a specific one-on-one comparison of players at the same position, and even then you can't have an intelligent discussion unless you've actually seen them play. You can argue forever whether Shaq is or isn't better than Duncan and no statistic will ever be able to prove anything. With the possible exception of plus/minus (which is the only derived individual statistic which I believe has any hope of actually being meaningful), no statistic is ever going to tell you Bruce Bowen's worth.

IMO, what we should be doing is comparison of the TEAM factors that determine who wins. Number of wins and # of championships is an absolute (everyone agrees on who has the most wins and who won the championship) to which you can apply rigorous statistics, and, for example, prove whether or not defensive FG% is really more important than PPG allowed like Pop is always saying. This is what I was trying to do in this post: http://www.spurstalk.com/forums/showthread.php?t=61146&

The other reason I haven't posted in ABPR yet is I wanted to finish my update of this analysis and run it by my Sonic statistician friend.

[ShoogarBear]

Kind of hard to PROVE much of anything. Can anyone really prove much of anything? Its just a possible indicator of quality of play per minute. Its not a holy grail statistic -- it just seems to make more sense to use it over pure PPG, APG, RPG totals and Wins a player is a part of. I personally like the fact that it accounts for things like Pace, Team PPG etc.

I have no problem with that. It's making casual statements like "X is obviously better than Y" when there is no objective data to support it that drives me nuts.

Just because one losing team might have a higher PER than a winning team, doesn't mean its meaningless. Did Jordans 63 vs the Celtics prove to be worthless vs the Celtics because his team lost? Most people still would consider Jordan the inferior player to Bird even in those days, but he put up the same stats had a better PER than Bird, played better D. But the only difference was he had a weaker supporting cast, less experience with a good supporting cast.

I never said the concept of PER (per se) was meaningless (although I think there is a good deal of crock in Hollinger, who I don't think really understands numbers as well as he wants people to believe). In fact, I never said PER wasn't a better measure than PPG. I was just making the point that people need to have a better respect for what cons utes statistical proof.

(No offense, but parenthetically this whole conversation brings me back to the early days of the Church of Manu, when if you didn't say he was the greatest of all time, you were accused on not being a true Spurs fan. )

Ultimately it does all come down to opinion. But why is PER such a bad thing, when all most people use to judge player quality is pure output? PER just helps equalize for things such as Pace, Team PPG, and efficiency etc.....

Why is it such a bad thing?

PER is a bad thing only when people start tossing about absolutes like "it's clearly better than . . . " when they don't have a shred of evidence to support it.

[SpurOutofTownFan]

Thanks Nikos for making it clear to him. PER is just better than counting how many points a player scored and saying he's the best guy around. That is all. No need to write a wall of text on this subject.

[ShoogarBear]

Thanks Nikos for making it clear to him. PER is just better than counting how many points a player scored and saying he's the best guy around. That is all. No need to write a wall of text on this subject.

Are you clear on what cons utes an absolute yet?

[manustarting2gd]

So back to the thread, what is Manu's PER after the Wiz game?

[timvp]

Updated PER Standings
1 LeBron James, CLE 32.25
2 Manu Ginobili, SAS 28.94
3 Carlos Boozer, UTH 28.21
4 Dwight Howard, ORL 27.17
5 Chris Paul, NOR 27.15
6 Kevin Garnett, BOS 25.94
7 Kobe Bryant, LAL 25.87
8 Dirk Nowitzki, DAL 25.13
9 Tim Duncan, SAS 25.04
10 Tracy McGrady, HOU 24.57
11 Michael Redd, MIL 24.42
12 Steve Nash, PHO 24.39
13 Tony Parker, SAS 24.20






If you adjust PER for minutes played, the Big Three are amazingly close:

Adjusted PER
Tony Parker - 17.393
Manu Ginobili - 17.364
Tim Duncan - 17.319

To get the adjusted PER number, you multiply percentage of game played times PER.

[Mark in Austin]

Updated PER Standings
1 LeBron James, CLE 32.25
2 Manu Ginobili, SAS 28.94
3 Carlos Boozer, UTH 28.21
4 Dwight Howard, ORL 27.17
5 Chris Paul, NOR 27.15
6 Kevin Garnett, BOS 25.94
7 Kobe Bryant, LAL 25.87
8 Dirk Nowitzki, DAL 25.13
9 Tim Duncan, SAS 25.04
10 Tracy McGrady, HOU 24.57
11 Michael Redd, MIL 24.42
12 Steve Nash, PHO 24.39
13 Tony Parker, SAS 24.20






If you adjust PER for minutes played, the Big Three are amazingly close:

Adjusted PER
Tony Parker - 17.393
Manu Ginobili - 17.364
Tim Duncan - 17.319

To get the adjusted PER number, you multiply percentage of game played times PER.

Somwhere in America John Hollinger is having to change his shorts after seeing that info...

[diego]

i thought per already took into account minutes? wasnt that the main criticism of it?

anyways i started thinking, if 35 was such an exceptional avg for manu to have, what do single game PER's look like? for ex: does kobe's per in his 81pt game surpass duncan's almost quad double (or drob's real one for that matter)? that will tell me more about this stat than anything else. because if i look at those two stat lines, i will always be more impressed by the quad double type game.

[roycrikside]

I don't think PER is meant to be used or calculated that way. It's already a per minute stat. The production doesn't change based on time played in a game. If player A has a PER of 25 for 25 minutes, he isn't less of a player than Player B who has a PER of 24 in 30 minutes. His role is different perhaps, or the way the coach uses him, but not his overall quality.

I think the failure of TimVP's method can be proven by looking at LeBron's numbers. If he plays 90% of his team's minutes, then according to TimVP's adjusted PER, he'll widen the gap between himself and our big three. But all this does is punish our guys for being on such a good team that they can afford to have more blowouts and rest more. It doesn't mean that they're much lesser players individually, than James.

The beauty of Hollinger's formula is that it strips down everything else and just tells you how much a guy helps or hurts you per every minute he's on the floor. Messing with it any further devalues the formula.

[timvp]

i thought per already took into account minutes? wasnt that the main criticism of it?

PER is a per minute stat. Minutes per game isn't used in the equation.

anyways i started thinking, if 35 was such an exceptional avg for manu to have, what do single game PER's look like? for ex: does kobe's per in his 81pt game surpass duncan's almost quad double (or drob's real one for that matter)? that will tell me more about this stat than anything else. because if i look at those two stat lines, i will always be more impressed by the quad double type game.

That'd be an interesting comparison. As far as I know, Hollinger hasn't made the PER equation publicly known.

I don't think PER is meant to be used or calculated that way. It's already a per minute stat. The production doesn't change based on time played in a game. If player A has a PER of 25 for 25 minutes, he isn't less of a player than Player B who has a PER of 24 in 30 minutes. His role is different perhaps, or the way the coach uses him, but not his overall quality.

In that scenario, I think it'd be pretty safe to say that Player B is the superior player. A drop of one in PER for five extra MPG is more than a fair trade off.

Of course it depends on the specifics of why Player A doesn't average more MPG. Is it because of foul trouble, fatigue, poor defense or something similar that simply doesn't allow the player to average more minutes ... or is it simply due to other players at the same position also needing minutes on the same team?

For example, right now Jose Calderon and Stromile Swift both have a PER of around 21. While they both play far from starters minutes, it's for vastly different reasons. Calderon doesn't play more minutes due to TJ Ford also being on the same team, therefore limiting the available minutes at point guard. Swift doesn't play more minutes because he's a horrible defender, gets tired easy and is a foul prone player. In this scenario, you'd have to look at the context of the playing time to determine who is the better player. PER itself doesn't tell you that.

I think the failure of TimVP's method can be proven by looking at LeBron's numbers. If he plays 90% of his team's minutes, then according to TimVP's adjusted PER, he'll widen the gap between himself and our big three. But all this does is punish our guys for being on such a good team that they can afford to have more blowouts and rest more. It doesn't mean that they're much lesser players individually, than James.

Looking at an adjust version of PER is decently fair if you looking at it intra-team. Inter-team it doesn't make as much sense ... then again, PER itself isn't perfect to begin with.

The beauty of Hollinger's formula is that it strips down everything else and just tells you how much a guy helps or hurts you per every minute he's on the floor. Messing with it any further devalues the formula.

PER, like any other stat, needs to be looked at in context. You can't just blindly look at the PER number and then use that to come to all conclusions.

I still contend that the best use of PER is to find undervalued, young players who could produce well with increased playing time. Three examples of such players right now are Louis Williams, Andrew Bynum and Sean Williams. By watching each of those three players, you can see their talent. And PER backs that up as each of those three players are in the top 50 in the league in PER.

[FromWayDowntown]

Here's a primer on the PER calculation:

http://www.basketball-reference.com/about/per.html

[SpursIndonesia]

TiMVP is a great basketball mind.

But Jackie Butler was a PER phenom, wasn't he ?

[timvp]

Updated PER Standings
1 LeBron James, CLE 32.61
2 Manu Ginobili, SAS 29.96
3 Carlos Boozer, UTH 28.33
4 Chris Paul, NOR 27.06
5 Dwight Howard, ORL 26.80
6 Kobe Bryant, LAL 26.14
7 Kevin Garnett, BOS 25.41
8 Tim Duncan, SAS 25.21
9 Michael Redd, MIL 24.43
10 Steve Nash, PHO 24.35
11 Tony Parker, SAS 24.25

I want the Spurs to get three in the top ten just to see Hollinger's chode explode. Actually, I'm surprised that Hollinger hasn't written an article about the exploits for Ginobili, Duncan and Parker this year. Knowing Hollinger, now would be a good time to point to his magic PER stat as the reason why the Spurs have the best record in the West.

We are now more than 20% into the NBA season and Ginobili is still hovering around 30. If Ginobili can finish the year at his current PER number, it'd be the 12th best PER season all-time. And the fifth best for someone not named Chamberlain or Jordan.

Parker's number is also pretty impressive. Ginobili has been known as the PER master over the years but right now, Parker's PER is higher than Ginobili's PER ever was in the past. Duncan is also not half bad considering he's not even play up to his normal standards.

[timvp]

But Jackie Butler was a PER phenom, wasn't he ?

Yep.

Butler is the poster boy for why relying on PER too much is dangerous.

[exstatic]

Yep.

Butler is the poster boy for why relying on PER too much is dangerous.

What the heck is JButt up to these days, anyway? Where did he drift to after the Rox cut him loose?

[ArgSpursFan.]

What the heck is JButt up to these days, anyway? Where did he drift to after the Rox cut him loose?

He got a job @ McDonals, and now He can actually work and be a fat ass at the same time.
Something that the NBA didn't like from him I guess.

[Mark in Austin]

Yep.

Butler is the poster boy for why relying on PER too much is dangerous.

What Holinger refuses to acknowledge is that PER becomes more accruate as a predictor of impact/talent the more consistent minutes the player plays.

The problem w/ Butler was Holinger used the PER from his last season in NY, which he played limited minutes but was mildly effective when in. To turn around and apply that to project performance as a starter or major rotation player is so asinine I still can't believe he gets paid to come up with this crap.

[mikekim]

It strikes me that in a few years when he loses a step, Manu will probably end his career as a point guard, and may remain quite a useful player even when his explosiveness has diminished - he is just so smart and what a passer!

You know, that's not a bad idea. No need to put up with any Beno Udrihs, settle for Jacque Vaughns, or groom Darius Washingtons.

Just move Manu to back-up PG, officially. I know he fills this role sometimes. But in a few years, he could be there permanently.

[Deimosfobos]

Updated PER Standings:

1 LeBron James, CLE 32.52
2 Manu Ginobili, SAS 29.88
3 Carlos Boozer, UTH 28.26
4 Chris Paul, NOR 27.46
5 Dwight Howard, ORL 26.74
6 Kobe Bryant, LAL 26.07
7 Kevin Garnett, BOS 25.35
8 Tim Duncan, SAS 17 25.14
9 Steve Nash, PHO 24.28
10 Tony Parker, SAS 24.18

Our big 3 on the top ten

[Freeze]

1 LeBron James, CLE 32.64
2 Manu Ginobili, SAS 29.43
3 Carlos Boozer, UTH 28.28
4 Chris Paul, NOR 16 27.48
5 Dwight Howard, ORL 26.20
6 Kobe Bryant, LAL 25.52
7 Tony Parker, SAS 25.22
8 Tim Duncan, SAS 24.73
......
39 Brent Barry, SAS 20.05

[Freeze]

Even better !!!

Hollinger Stats: Player Efficiency Rating - International

1 Manu Ginobili, SAS 29.43
2 Tony Parker, SAS 25.22
3 Tim Duncan, SAS 24.73
4 Nash
5 Dirk
6 Yao



http://insider.espn.go.com/nba/holli...seasonType%3d2

[Deimosfobos]

Good moment for a bump

1 LeBron James, CLE 32.65
2 Manu Ginobili, SAS 30.65
3 Carlos Boozer, UTH 27.37
4 Chris Paul, NOR 27.37
5 Amare Stoudemire, PHO 26.64
6 Dwight Howard, ORL 25.82
7 Kevin Garnett, BOS 25.64
8 Kobe Bryant, LAL 25.64
9 Tim Duncan, SAS 24.54
10 Michael Redd, MIL 24.04

[PlaneFast]

How can Duncan's PER change without playing? PER is a per minute measurement. If he plays no minutes, then his per minute stats shouldn't change.

[Extra Stout]

PER can never be a complete measurement of a player. It only uses kept statistics. Hollinger has no means to measure individual defensive play, or altered shots, or defensive rotations that prevent a field goal attempt, or screens that lead to baskets, or charges drawn, etc.

[Tippecanoe]

How can Duncan's PER change without playing? PER is a per minute measurement. If he plays no minutes, then his per minute stats shouldn't change.

i was wondering the same thing. anyone know??