IS PER the best way to evaluate players?

[calamity]

like all things, its good for some things and not very good for other things. not the end all and be all of course.

[chubby_1_kenobi]

Yes and no. PER doesnt measure defense so it hurts guys like Bruce Bowen who were valuable (in his prime). But it gives you a good baseline of a player. League average PER is 15.0

* A Year For the Ages: 35.0
* Runaway MVP Candidate: 30.0
* Strong MVP Candidate: 27.5
* Weak MVP Candidate: 25.0
* Bona fide All-Star: 22.5
* Borderline All-Star: 20.0
* Solid 2nd option: 18.0
* 3rd Banana: 16.5
* Pretty good player: 15.0
* In the rotation: 13.0
* Scrounging for minutes: 11.0
* Definitely renting: 9.0
* The Next Stop: DLeague 5.0

According to this, Derrick Rose isn't even a "weak MVP Candidate". I never understood the fascination with PER. Sure, it's a decent way to measure stats, but the NBA is far more than just stats. (Sorry Lebron)

Bulls fan wouldn't be saying this in the 90's.

P.S There is also something called anomalies.

Have anyone here ever even come close to coming up with a formula that made sense, and can be used? You can't be a idiot and come with something like PER. Yes it has flaws, but what doesn't?

The fact that it's a flawed formula make it idiotic to use it to prove anything. This is not a philosophical question, it's a freaking mathematical formula hence it's a cut and dry right or wrong. If a mathematical formula is flawed in any way, it is unusable.

[JCWalters]

It is the best STAT to use to evaluate players. But it is not the best WAY.

[chubby_1_kenobi]

Well thats why Rose is a weak MVP candidate. I never understood the guys who said that he was running away with it.

Maybe because Rose is putting up great numbers for a team that's only 2 games behind 1st place in the East. The Celtics have 4 allstars. The Heat have 3 allstars. Derrick Rose is carrying the Bulls with their 2nd and 3rd best players each missing significant amount of time.

Show me someone more deserving.

Im kidding. PER is also skewed in favour of guys who get to the free throw line so that hurts Rose.

PER is encouraging players to accumulate rebounds and assists because it's a surefire way to increase your PER. If someone wants to accumulate points, they risk a lower PER because a MISSED shot is recorded as a negative PER. But let's say if someone is busy chasing rebounds and never bother to attempt to challenge any shot in the paint, PER has no way to determine that. Even when it is clear that a big man who refuses to challenge shots in the name of defensive rebounds is a big detriment to your team defense, PER has absolutely no way of determining that as a negative. So why not do it? (if NBA players are encouraged to care about PER)

If a player accumulate assists by pounding and pounding the ball on the top of the key and stagnate the offense in the name of "dishing it to an open shooter", is there a negative impact to his PER? No. Again, it is proven that a team who runs a dynamic offense like the Celtics and the Spurs are going to do much much better in half court scenario compare to a team who implement "give the ball to a star player on top of the key and get out of the way" offense. So again, if there's no negative in doing that why not just do it all the time?

[Hangtime84]

It is the best STAT to use to evaluate players. But it is not the best WAY.

+1

[droponov]

The fact that it's a flawed formula make it idiotic to use it to prove anything. This is not a philosophical question, it's a freaking mathematical formula hence it's a cut and dry right or wrong. If a mathematical formula is flawed in any way, it is unusable.

Yeah. Let's stop using radars, for example. Like lots of other things - you can't even imagine - their technology is also based on estimation theory. From a mathematical standpoint, there's nothing wrong with PER.

I don't use PER though - or any of these compilation formulas, for that matter. I understand time is precious, but does it really take that much to look at the full stats of the players - at least those that are part of PER? To me, saying that "PER Is the best stat" is like prefering to read the Reader's Digest condensed version of a good book instead of the real book.

[chubby_1_kenobi]

It is the best STAT to use to evaluate players. But it is not the best WAY.

Use REAL statistics to evaluate players. The pure unprocessed data such as ppg, FG%, apg, rpg etc....
Don't use data who has been modified with arbitrary numbers like PER. PER is NBA's equivalent to food at a fast food joint. It tastes like crap, it looks horrible, it's unhealthy but it is the fastest way to fill your stomach.

[chubby_1_kenobi]

The fact that it's a flawed formula make it idiotic to use it to prove anything. This is not a philosophical question, it's a freaking mathematical formula hence it's a cut and dry right or wrong. If a mathematical formula is flawed in any way, it is unusable.

Yeah. Let's stop using radars, for example. Like lots of other things - you can't even imagine - their technology is also based on estimation theory. From a mathematical standpoint, there's nothing wrong with PER.

Hello? PER is not an estimation. It doesn't product result in the form of a range or a plus minus. It pinpoints a specific number with no room for errors.

From a mathematical standpoint, there's nothing wrong with PER.

PER is a ridiculously flawed stat. Why? Because it uses arbitrary weights.

Here are weights for PER in relation to points:

Pts: 1
Ast: 0.92
OReb: 0.85
DReb: 0.35
Stl: 1.2
Blk: 0.85
MissedFG: -0.85
MissedFT: -0.45
TOV: -1.20
PF: -0.41

There's nothing wrong with that? Really?

[JCWalters]

It is the best STAT to use to evaluate players. But it is not the best WAY.

Use REAL statistics to evaluate players. The pure unprocessed data such as ppg, FG%, apg, rpg etc....
Don't use data who has been modified with arbitrary numbers like PER. PER is NBA's equivalent to food at a fast food joint. It tastes like crap, it looks horrible, it's unhealthy but it is the fastest way to fill your stomach.

Your analogy is colorful, but it is just as empty as your "fast food." Please explain why PER "tastes like crap..." compared to the "unprocessed data."

On the issue of best stat to use, PER is just as good as the "unprocessed" data except for the weighing of each component. On that, it not just as good but it is very close. Plus, PER has the advantage of covering more ground than any single raw data category.

On the weighing, PER is pretty accurate. It does not use random amounts for weights. It goes by the fact that each basket is worth 2 or 3 points and each foul shot is worth 1 point. It is not arbitrary or random but instead logical and accurate. The only problem I have with the weighing that I can think of right now is the ASSIST component. Hollinger arbitrarily gives the assistor 1/3 of the credit and the finisher 2/3 of the credit irregardless of how good or bad the pass was or how good or bad the finisher was in getting open (creating). For example, a kick out from a Duncan post-up is worth more than a pass to a cutting Rip Hamilton. In the former, the passer (Duncan) did most of the creating. While, in the latter, the finisher (Rip) did most of the creating. But Hollinger would give both passes 1/3 and both finishes 2/3.

[JordansBulls]

Casual NBA fan here, so I am asking all the resident basketball experts if PER is the best way to evaluate talent and if player A is better then player B...

Or is it a little more complex and just going by PER is to simplistic.

Thanks in advance

It's good for evaluating if a guy produces a lot or no.

[YFZblu]

PER is OK to look at, but it's pretty rudimentary in terms of the data it pulls. Box score whores can be great PER players.

For example, Hakim Warrick has a higher PER than Grant Hill this season. Anyone who thinks Hakim Warrick is more productive than Grant Hill is out of their minds, which is why I prefer Adjusted Plus Minus among other things.

[chubby_1_kenobi]

On the weighing, PER is pretty accurate. It does not use random amounts for weights.

Yes, it does.

It goes by the fact that each basket is worth 2 or 3 points and each foul shot is worth 1 point.

And an assists is only 92% as useful as a point.
A defensive rebound is worth 35% of a point
An offensive rebound is worth 85% of a point

and my own personal favorite;
A steal is worth 120% of a point (it actually values a steal more than any other category )

Where do those numbers come from, I wonder?

It is not arbitrary or random but instead logical and accurate

Explain the logic please.

[chubby_1_kenobi]

PER is OK to look at, but it's pretty rudimentary in terms of the data it pulls. Box score whores can be great PER players.

For example, Hakim Warrick has a higher PER than Grant Hill this season. Anyone who thinks Hakim Warrick is more productive than Grant Hill is out of their minds, which is why I prefer Adjusted Plus Minus among other things.

Hakim Warrick does 2 things that boosts up his PER. Rebound and finish in the paint a lot. I'm glad you brought up Warrick and Hill think as PER values a wing player FG% equal as a post player FG% (Oooh since A center if shooting 60% from the field, why not let him take all the shot nonsense)

Can you imagine if an entire team is constructed from players who cares about their PER. Everyone would want to either drive or post up, who'd want to shoot and risk lowering their PER?

[droponov]

The fact that it's a flawed formula make it idiotic to use it to prove anything. This is not a philosophical question, it's a freaking mathematical formula hence it's a cut and dry right or wrong. If a mathematical formula is flawed in any way, it is unusable.

Yeah. Let's stop using radars, for example. Like lots of other things - you can't even imagine - their technology is also based on estimation theory. From a mathematical standpoint, there's nothing wrong with PER.

Hello? PER is not an estimation. It doesn't product result in the form of a range or a plus minus. It pinpoints a specific number with no room for errors.

So what? Read the posts I was replying to. And PER is not an estimation in a formal sense but not because it doesn't product a result in the form of a range. That would be quite easy to do. That's not what defines if there's randomness hence a non-deterministic problem.

From a mathematical standpoint, there's nothing wrong with PER.

PER is a ridiculously flawed stat. Why? Because it uses arbitrary weights.

Here are weights for PER in relation to points:

Pts: 1
Ast: 0.92
OReb: 0.85
DReb: 0.35
Stl: 1.2
Blk: 0.85
MissedFG: -0.85
MissedFT: -0.45
TOV: -1.20
PF: -0.41

There's nothing wrong with that? Really?[/quote]

From a mathematical perspective? Nothing at all. There's mathematical tractability. There are lots of stuff where you need to use subjective assumptions - including many estimators. That doesn't make things invalid.

As I''ve said, I don't use PER, but the process is basically the same: I get the statistics and I weight them to form an overall idea about the performance of the player. I don't think that using the same weights for every player (or even for the same player in different moments) or expressing the result with a single numeric value is useful or interesting at all, but in an intuitive, heuristic way, what we do when we look at those raw stats and process them in our brain to form an opinion about the value of the player is exactly what bothers you so much.

[uncarved block]

Yes and no. PER doesnt measure defense so it hurts guys like Bruce Bowen who were valuable (in his prime). But it gives you a good baseline of a player. League average PER is 15.0

* A Year For the Ages: 35.0
* Runaway MVP Candidate: 30.0
* Strong MVP Candidate: 27.5
* Weak MVP Candidate: 25.0
* Bona fide All-Star: 22.5 * Borderline All-Star: 20.0
* Solid 2nd option: 18.0
* 3rd Banana: 16.5
* Pretty good player: 15.0
* In the rotation: 13.0
* Scrounging for minutes: 11.0
* Definitely renting: 9.0
* The Next Stop: DLeague 5.0

Kevin Martin says hi

[YFZblu]

PER is OK to look at, but it's pretty rudimentary in terms of the data it pulls. Box score whores can be great PER players.

For example, Hakim Warrick has a higher PER than Grant Hill this season. Anyone who thinks Hakim Warrick is more productive than Grant Hill is out of their minds, which is why I prefer Adjusted Plus Minus among other things.

Hakim Warrick does 2 things that boosts up his PER. Rebound and finish in the paint a lot. I'm glad you brought up Warrick and Hill think as PER values a wing player FG% equal as a post player FG% (Oooh since A center if shooting 60% from the field, why not let him take all the shot nonsense)

Can you imagine if an entire team is constructed from players who cares about their PER. Everyone would want to either drive or post up, who'd want to shoot and risk lowering their PER?

Yeah, pretty much. I'd be interested in finding out if PER has ever been used during a contract negotiation:

Agent: "My guy has the best PER on the team! You can't low ball him like that!"

I would put PER on par with single game unadjusted plus/minus, only even more misleading. I don't really like it much, only because there are more complete advanced tools available for statheads.

[chubby_1_kenobi]

So what? Read the posts I was replying to. And PER is not an estimation in a formal sense but not because it doesn't product a result in the form of a range. That would be quite easy to do. That's not what defines if there's randomness hence a non-deterministic problem.

As long as the result of a formula is a figure instead of a range, it is NOT an estimation.

From a mathematical perspective? Nothing at all. There's mathematical tractability. There are lots of stuff where you need to use subjective assumptions - including many estimators. That doesn't make things invalid.

Repeat, PER is not an estimation.

what we do when we look at those raw stats and process them in our brain to form an opinion about the value of the player is exactly what bothers you so much.

An opinion that comes from observation and raw data, I have no problem with.

An opinion that stems from PER and PER only, I despise (before you say no one here does that, may I point there's one bald sob under ESPN payroll who fits that description to a T)

[paul]

Good god no, it's one of the most flawed statistics in all of sports, particularly when it's used as some kind of overall player ranking tool.

PER stands for PLAYER EFFICIENCY RATING, and that is absolutely all it does. It rates players by their efficiency, and it does it pretty poorly. Some heavy volume can make up for a lack of efficiency, but if you are a low usage / good rebounding rate / efficient scorer (i.e. can catch and dunk) you are a GOD to per, even if you play 10mpg and can't shoot from further than 3 feet. Conversely if you have a poor TS you can absolutely forget about it, regardless of anything else you do well. It also completely overlooks defense as an apparently unnecessary skill of the game.

It sucks.

[droponov]

So what? Read the posts I was replying to. And PER is not an estimation in a formal sense but not because it doesn't product a result in the form of a range. That would be quite easy to do. That's not what defines if there's randomness hence a non-deterministic problem.

As long as the result of a formula is a figure instead of a range, it is NOT an estimation.

From a mathematical perspective? Nothing at all. There's mathematical tractability. There are lots of stuff where you need to use subjective assumptions - including many estimators. That doesn't make things invalid.

Repeat, PER is not an estimation.

what we do when we look at those raw stats and process them in our brain to form an opinion about the value of the player is exactly what bothers you so much.

An opinion that comes from observation and raw data, I have no problem with.

An opinion that stems from PER and PER only, I despise (before you say no one here does that, may I point there's one bald sob under ESPN payroll who fits that description to a T)

I really don't understand your fixation with the estimation thing, that, as explained, I only used to refute that guy's view about what is right or wrong in mathematics - the failure to achieve the exact true value doesn't make estimators wrong, so it was a good example. You may repeat "PER is not an estimation" how many times you want, but that doesn't make it mathematically invalid, it'd be a non-sequitur, even if true (which depends a lot on semantics).

And again, I have no idea where you studied math, but the thesis that an estimator needs to produce estimates in the form of a range is absolute crap. An interval estimator will produce a range of values; point estimators will produce single value estimates.

A sample mean is an estimator. How often do you see a mean presented in the form of a range of values? Single-value estimates aren't as informative as confidence intervals, but they have their place in statistics. THis issue is lateral at best for the topic at hand, but people are already ignorant of basic mathematics, so there's no need to the likes of you spewing misinformation around.

---

As for the issue at hand, I'm glad to see you backtracked and now your problem is with opinions "that stems from PER and PER only". I read enough of Hollinger to know that your characterization of him is absolutely false though. He relies on PER way too much for my liking, but to say his opinions stem from PER only may be a popular caricature but not something grounded in reality. And this is coming from a guy who, as I've said, has no use for PER (or other linear weight formulas, or for that matter for any sort of summarized player rating metrics) whatsoever.

[chubby_1_kenobi]

I really don't understand your fixation with the estimation thing, that, as explained, I only used to refute that guy's view about what is right or wrong in mathematics - the failure to achieve the exact true value doesn't make estimators wrong, so it was a good example. You may repeat "PER is not an estimation" how many times you want, but that doesn't make it mathematically invalid, it'd be a non-sequitur, even if true (which depends a lot on semantics).

I never said it's mathematically invalid, I said it's "flawed" in context of measuring player A vs player B.
The failure to achieve the exact true value is not even of significance when it comes to PER, why? Because it is not even capable of determining what a true value is in terms of comparing one statistical measurement to others, hence this is created out of thin air:

Here are weights for PER in relation to points:

Pts: 1
Ast: 0.92
OReb: 0.85
DReb: 0.35
Stl: 1.2
Blk: 0.85
MissedFG: -0.85
MissedFT: -0.45
TOV: -1.20
PF: -0.41

You get it now? This is exactly why PER has eliminated the possibility of it being an estimate. It relies on specific weight that has no merit to come up with a figure that has no room for error mathematically but only because that room has been filled with numbers which origin are unclear.

And again, I have no idea where you studied math, but the thesis that an estimator needs to produce estimates in the form of a range is absolute crap. An interval estimator will produce a range of values; point estimators will produce single value estimates.

Now, who's arguing semantics? An interval estimator and a point estimator might be presented differently but it still points to a posterior distributions inside a certain parameter. When I say that PER is not an estimate because it's not presented in a range, I'm being very layman about it. Should I say that "PER is not an estimate because it doesn't point to a posterior distributions" just to satisfy some people's superficial ism?

As for the issue at hand, I'm glad to see you backtracked and now your problem is with opinions "that stems from PER and PER only"

Backtrack when?

I read enough of Hollinger to know that your characterization of him is absolutely false though. He relies on PER way too much for my liking, but to say his opinions stem from PER only may be a popular caricature but not something grounded in reality.

I've read enough of him to know that he has no idea that his formula is flawed and shamelessly using it as basis of everything in his argument.