PDA

View Full Version : ESPN player efficiency ratings, projections for '06-'07



Slick Pinkham
10-18-2006, 10:35 AM
http://insider.espn.go.com/nba/hollinger/rankings

some points of discussion: JO ranked as the 28th best player in the league

Jax as the 26th best SG in the league

Speed
10-18-2006, 10:37 AM
http://insider.espn.go.com/nba/hollinger/rankings

some points of discussion: JO ranked as the 28th best player in the league

Jax as the 26th best SG in the league


I just saw this, is this a fantasy ranking, there is some Houston guy, I swear I've never heard of ranked way ahead of Al. I swear I thought I followed the NBA close until now.

Chuck Hayes = 11
Al = 28

AT POWER FORWARD, Wow!

Unclebuck
10-18-2006, 10:43 AM
Those don't appear to be fantasy ratings per se, but I don't believe in judging a player bassed on a series of stats and efficiency numbers


I had never heard of Chuck Hayes either, I guess he's an undrafted free agent the Rockets picked up last season

Speed
10-18-2006, 10:45 AM
Those don't appear to be fantasy ratings per se, but I don't believe in judging a player passed on a series of stats and efficiency numbers

Agreed, remember World (Lloyd) B Free, he'd probably project good on a stats ranking. Man I'm getting old.

Fool
10-18-2006, 11:13 AM
Appearantly Granger is already better than Tayshaun as well as half the starting SF in the NBA.

PacerFan31
10-18-2006, 11:44 AM
Appearantly Granger is already better than Tayshaun as well as half the starting SF in the NBA.

They also have Danny rated higher than Artest.

tinsley#11
10-18-2006, 12:12 PM
Chuck Hayes was an average player who played at University of Kentucky, nothing special. Al is def. better than him.

denyfizle
10-18-2006, 12:18 PM
I would've read this article more except that I saw Steve Nash ranked #19. Are you kidding me???? Without him there is no Phoenix Suns. Top 10 would be fine but almost missing the TOP 20??? Gimme a break.

ajbry
10-18-2006, 12:20 PM
I never take anything Hollinger does seriously, and this is certainly no exception. His precious statistically-based analysis is way off, as usual.

JayRedd
10-18-2006, 03:10 PM
This is all based on algorithms and statistical number crunching...Take it for what it's worth

Putnam
10-18-2006, 03:32 PM
You guys.

The PER is a pretty good summary statistic. Like any statistic, it precisely measures something, but not everything. There is no perfect statistic, and in this case the author makes no claim that the PER ranking is tantamount to a listing of NBA players in order from best to worst. Rather, it is a summary value that gets at a lot of what makes a player good. This one emphasizes output/time. Here's what goes into the Player Efficiency Rating


Two important things to remember about PER is that it's per-minute and pace-adjusted. It's a per-minute measure because that allows us to compare, say, Drew Gooden to Donyell Marshall, even though there is a wide disparity in the minutes they played. I also adjust each player's rating for his team's pace, so that players on a slow-paced team like Indiana aren't penalized just because their team's games have fewer possessions than those of a fast-paced team such as Phoenix.


The way to use statistics is not to glance at them and then say, "This does not conform to my presuppositions, therefore it is wrong." Or...


I never take anything Hollinger does seriously, and this is certainly no exception. His precious statistically-based analysis is way off, as usual.

The analysis is not "way off" at all. The PER measures what it claims to measure. Unless you can show that Hollinger made a math error in the calculations, you need to accept the results. Or ignore them if you'd rather. But don't say they are "way off" when in fact it is an accurate measure of what it claims to measure.

This way of quantifying performance favors players who account for a big share of their team's totals. That is why Kobe is way up there but Steve Nash isn't. Kobe hogs the ball, and the PER rewards that. But Kobe actually scored at a high clip, unlike another shooting guard who also hogged the ball but only had a FG% of .41.

This is an output/time rating. In that limited sence, it is a pretty good measure. It doesn't measure intangibles, or luck, or a player's importance to the team, or the success of the team, or whether a particular fan has a mancruch on a player.




In other words,


This is all based on algorithms and statistical number crunching...Take it for what it's worth

JayRedd
10-18-2006, 05:02 PM
Yeah...what Putnam said.

Not even Hollinger would suggest that Tim Duncan is the 6th Best PF in the League. That's just what it happens to say when he puts his little numbers into the formula from last year, and how the projected stats calculate for the upcoming season.

All numbers have there limitations, but they also can shed light on certain things if you look at them properly.

Fool
10-18-2006, 05:02 PM
The PER is a pretty good summary statistic. Like any statistic, it precisely measures something, but not everything. There is no perfect statistic, and in this case the author makes no claim that the PER ranking is tantamount to a listing of NBA players in order from best to worst. Rather, it is a summary value that gets at a lot of what makes a player good. This one emphasizes output/time. Here's what goes into the Player Efficiency Rating


This is not the first time Hollinger's PER has been presented as a topic. I think many of us already know its simply Hollinger's love fest with himself (though, I'm willing to admit that most would probably not label it exactly that way). While Hollinger might not openly claim his list to be the end-all be-all of rankings, his continued use of it to evaluate every player comparison he considers, along with the way he allows ESPN to present his "articles" is tantamount to doing so (that last clause is a stipulation I'm including only to be generous to Mr. Hollinger, but I don't believe its actually necessary). He's out to do one thing, promote his stat creation in the hopes it becomes the NBA version of Sabermetrics (that's flagrant opinion on my part but not ignorantly so). As such, addressing his ranking as though its offering more than the modest "here's one thing to look at as well" is not irrational IMO.

Regardless, I think most of the replies in this thread are treating the list correctly, as a very limited tool that is generally "way off" in its guidence for most useful things.

NuffSaid
10-18-2006, 05:38 PM
Appearantly Granger is already better than Tayshaun as well as half the starting SF in the NBA.

They also have Danny rated higher than Artest.

I would've read this article more except that I saw Steve Nash ranked #19. Are you kidding me???? Without him there is no Phoenix Suns. Top 10 would be fine but almost missing the TOP 20??? Gimme a break.
After reviewing Mr. Hollinger's player efficiency rankings and reading some comments such as the ones above, I can understand why some view the rankings as BS and share the same sentiments as the poster below...

I never take anything Hollinger does seriously, and this is certainly no exception. His precious statistically-based analysis is way off, as usual.
But, you have to try to understand how Hollinger came to his PER (Player Efficiency Rating) in the first place. There's a link on the first page at the top right above the PERs, "What is a PER?". If you click on that link, it will provide a brief on how Holliger came up with his figures. I think once people take the time to review how Holliger came up w/his ranking system most would agree the majority of the players are ranked quite fairly.

AesopRockOn
10-18-2006, 05:39 PM
Pau and Amare are way too high; even for projections, you should take number of games played into consideration. Lol at Chuck Hayes.

McKeyFan
10-18-2006, 05:47 PM
Agreed, remember World (Lloyd) B Free, he'd probably project good on a stats ranking. Man I'm getting old.

Anybody know where I could get a highlight reel of World B. Free, or at least a tape of a game he played in?

I've got a friend I nicknamed "World Free" but he doesn't know what I'm talking about. I'd like to enlighten him ;)

3rdStrike
10-18-2006, 07:10 PM
Hayes had some pretty impressive numbers all things considering.

40 G | 13.4 MPG | 3.7 PTS | 4.5 REB | .562 FG %

Just makes the Rockets look even dumber for trading Rudy Gay for Shane freakin Battier (the "WTF are you kidding me" moment of the draft).

Shade
10-18-2006, 07:46 PM
Artest as the #20 SF? Goodbye, credibility! :wave:

JayRedd
10-18-2006, 09:43 PM
Artest as the #20 SF? Goodbye, credibility! :wave:

Again...this is based on numbers.

Ron's numbers last year were not good

40.6% shooting on 15.5 attempts/game
30.8% 3pt shooting on 4.5 attempts/game
67.5% free-throw shooting on 5.4 attempts/game
only 3.8 apg and 5.1 rpg in 39.4 mpg
plus add in 2.3 TOs/game

Say what you want about statistics being nonsense or how his formula is way off, but Hollinger's numbers attempt to come up with a "Player Efficiency Rating" not a "How Good Are You Quotient".

And any way you look at those numbers, they certainly are not efficient.

Naptown_Seth
10-19-2006, 12:02 AM
This is all based on algorithms and statistical number crunching...Take it for what it's worth
Of course all algorithms and number crunchings are based on a person's OPINION of how those numbers should interact and be computated. Then the CPU just repeats the process quickly for many people.


Putnum is right to point out the methodology, but that doesn't make the stat a good stat outright. If results make any statician do a double-take (say Hayes better than Harrington) the next step is to review the method and perhaps refine it.

With the PER I think it's wrong to simply extend the per48/MIN in a linear fashion. Some sense of diminishing returns should be considered but at this point has not been. Similarly possessions per game determine a pace scalar that adjusts numbers, but this assumes that a player at a slow pace could perform just as well at a higher tempo game, and ignores the fact that perhaps the tempo is lower specifically because of a player's game, such as a spot up shooter who can only catch and shoot when open and is unable to effectively work as a main ball-handler for 30 mpg.

Certainly I wouldn't deny that more minutes and/or more possessions technically mean more chances for numbers, just that a linear increase is very possibly incorrect.


If you want a quality effort to do statistical analysis your one-stop shopping should be 82games.com, not Hollinger.

Putnam
10-19-2006, 09:13 AM
If you want a quality effort to do statistical analysis your one-stop shopping should be 82games.com, not Hollinger.


Agreed.

82games is very good. Many of their analyses are flawed, same as the PER, but they are very innovative.

Fool
10-19-2006, 10:14 AM
After reviewing Mr. Hollinger's player efficiency rankings and reading some comments such as the ones above, I can understand why some view the rankings as BS and share the same sentiments as the poster below...

But, you have to try to understand how Hollinger came to his PER (Player Efficiency Rating) in the first place. There's a link on the first page at the top right above the PERs, "What is a PER?". If you click on that link, it will provide a brief on how Holliger came up with his figures. I think once people take the time to review how Holliger came up w/his ranking system most would agree the majority of the players are ranked quite fairly.

I've posted on the message board he frequents. Going deeper doesn't make his PER better. And please, stop saying "he's not trying to say who's better". The guy LIVES by his PER creation. Its pretty much the only thing he considers WHENEVER he talks publically about players' worth.

JayRedd
10-19-2006, 11:03 AM
Unfortunately, and I think this is something that we are all savvy enough to understand, the actual value and sheer number of statistics in basketball in incredibly limited compared to both the MLB and NFL.

Things like OBP, SLG%, OPS, SB%, ERA and WHIP can often give you a very clear picture of how a player actually performed. I'd say the NFL is more limited in these of "quantifying" individuals, but rushing YPG, YPC, RAC, yards per reception, and QB completion % can give you a very good idea of how teams performed and some good insight into how well individual players did as well.

In the NBA, things like PPG, RPG, APG, SPG and BPG tell you things, but the player's role on his team, number of minutes player, number of shots taken, and the overall speed of the game have such a dramatic impact that compared Steve Nash to Sam Cassell with numbers becomes a pointless exercise with these traditional numbers.

Hollinger's equation tries to do that by minimizing all these outside factors. Does it succeed? Not exactly. But I wouldn't say that PER is an altogether failure. Surely, lots of things aren't accounted for such as actual defense, overall court awareness, situational play and countless other things that define a good basketball player.

But if you look at the PER ratings historically, the players who have the elite players who have the best seasons are usually over 25, the really good players are between 20 and 25, the quality starters are between 17-20, the rotation guys are between 13-17, and anyone below 13 is probably someone that the average sports fan has never even heard of (roughly).

So it does "work" in that sense and rarely will you see someone that's "way way off." Of course Chuck Hayes as the 11th best PF in the league is ridiculous, especially when Al is at 28. But last year's numbers came out at 18.07 and 16.08. Not a huge difference there. And when you consider that the #13 - #25 projected PFs are all between 16.00 and 19.00, I think even Hollinger himself would tell you that there is not a lot of diffence in overall talent in that 12 player group.

So, sure, I see why this number is dumb to a lot of people. I really don't think it's amazingly useful myself either. But it does provide a single number to give an overall comparison that can help broadly categorize show who is the elite of elite, who is an all star, who is really good, who is a quality starter, who is a bench guy and who you probably don't want on the court.

Think of this as QB Rating: An ultimately useless tool to discern the number 5 QB in the league from the number 8 guy, but something that, over time, generally gives a broad picture of who is great, who is good, who is average, etc.

Is Damon Huard really a better QB than Tom Brady? Of course not. And the fact that he is "rated" worse than Huard this year is obviously a traveshamockery. But that's what the numbers say. But like PER, the number's only tell some of the picture.

Take em for what they're worth. Which, I'd say is something, but not much as far as exact rankings.

Or better yet, like Naptown Seth says, check out Roland's work on 82games.com.

Naptown_Seth
10-19-2006, 03:18 PM
Agreed.

82games is very good. Many of their analyses are flawed, same as the PER, but they are very innovative.
And it's obvious that their overriding goal is to perfect statistical analysis, so they are always open about the flaw, even pro-active about finding those flaws, and make an effort to cross-reference other indexes and analyze the "accuracy/meaning" of their own final results.

Hollinger isn't trying to trick us, but at the same time he's a lot more comfortable parading out the results as if they have a major meaning.


Jay - I agree that the NBA is a poor "stat" sport, which is specifically why I never play roto-basketball despite playing tons of roto-MLB/NFL. I also think this is why the push toward PLUS-MINUS systems has been so strong. No coincidence that the +/- comes out of another very poor "stat" sport, hockey.

I'd really like to see some of the stats that Rick's staff came up with to make the case for Ron as DPOY.

I've toyed with trying to do a subjective "grading" of a game, giving players -1, 0, +1 (or similar) per possession based on doing what I thought was expected (just setting a screen, making an expected jumper, getting an unchallenged rebound) vs screwing up the play (not just a bad pass, but also a player messing up where he goes creating a missed pass, blowing a layup after a good assist to him, etc) vs making an above expected play (great assist even if the other player blows the shot, difficult play-saving FG after play breaks down, etc).

The problem is that this is highly subjective and frankly impossible to do real-time. If the Pacers hired me to break down tape, that would be different. ;)

But at least such a system would allow for a (-1) if you got beat off the dribble, even if your man then missed the shot when the PF challenged it, etc.


Back to 82games.com. Be careful if you go over there, especially if you are a stat geek. Appointments will be missed and lunches skipped as you browse some of the info they gather. :D

Putnam
10-19-2006, 04:12 PM
Naptown_Seth and JayRedd make great comments about how and why basketball is poor in data. Any game that is played rapidly with few stops and interruptions is bound to leave some important facets of the game out of the box score. One of the things 82games does, I think, is to rely on spotters who report extra details to them.

My question is this: What are the non-counted contributions that you would want to quantify if it were possible?

setting picks
boxing out
defensive stops (other than steals and blocks)
defensive coverage that prevents your man from a touch
defensive coverage that prevents a pass
joining on a double-team defensive effort
pressure that delays an inbounds pass in the halfcourt
spreading the floor
making a good pass before an assist
encouraging team mates
reading the floor and being in the right place
calling out information to team mates
being able to come in already warmed up
ability to score/perform in pressure situation (clutchness)

(what am I missing?)



With sincere respect to PD members who've played and coached and watched more games than I have, I think there is a tendency to overvalue the esoteric contributions. Boxing out is a skill that contributes, but it doesn't deserve to be rated as great a contribution as scoring or rebounding.

The first thing from the above list that I wish were better quantified would be passes that lead to a score. I'd give everybody an assist who handled the ball in a possession that resulted in a field goal. Second, missed field goals need to count against the player as well as makes count for him, through someting like a points-per-attempt index.
Third, I'd find some way to make points, rebounds and passes count more in a team victory than in a loss.

What do others think?

JayRedd
10-20-2006, 11:03 AM
And it's obvious that their overriding goal is to perfect statistical analysis, so they are always open about the flaw, even pro-active about finding those flaws, and make an effort to cross-reference other indexes and analyze the "accuracy/meaning" of their own final results.

Last year especially, they were really starting to look at a lot of very interesting and largely never-explored stat analysis. By getting "game charters" as volunteers to record and take notes on games, they were analyzing things like which players were the most effective scorers from different locations on the floor, and how "contested" rebounds are when shots were taken from different floor locations.

I found the second one very interesting in that 30% of the time, rebounds are completely uncontested. It's something I've always known somewhat I suppose from visual experience, but it really does make you question the value of the RPG stat. If a team has a bad shooting day, the fact that a big man got 12 boards doesn't necessarilly mean he's a good rebounder. It could mean he played a lot of minutes and while he was standing near the hoop a bunch of basketballs nearly hit him in the face so he extended his arms. Things like this help emphasize the "new", good stats like rebound rate and such.

I believe Roland (the guy that runs the site) is always looking for more "Game Charters" if any of you are interested.

Here's two of the ones I remember doing from last year.

www.82games.com/shotzones.htm (http://www.82games.com/shotzones.htm)
www.82games.com/rebounds.htm (http://www.82games.com/rebounds.htm)