Arguments Against Advanced Stats?

[murraycraven]

Most stat is suck - true statement

[brelic]

Most stat is suck - true statement

 

Absolutely. When I watch the Flyers tonight, I won't give two hoots about stats. I just wanna see Umberger pummel Hartnell. We need a least SOME sort of validation that the trade wasn't a complete waste, right??

 

EDIT - also, are we in some sort of cruel time warp where we keep playing Columbus??

[pilldoc]

I don't think anyone said otherwise. It's not that numbers lie, like the cliche goes - it's that numbers are just numbers. They're decontextualized.

 

So in your case, if a certain medication says that 10% of patients with diabetes suffer from side effect A, and you know that the patient has already exhibited side effects with a similar medication in the past due to his diabetes, you're probably going to recommend against taking that medication. The numbers don't lie - 10% suffer side effects. But you have additional information that contextualizes it.

 

Do you mean a patient suffering from, your example diabetes is suffering side effects from a drug they are using for diabetes?  Or do you mean they are already suffering from a complication of diabetes, such as eye problems, diabetic neuropathy, increased cardiac risks such as Hypertension, Stroke, increased cholesterol levels and a new medication could aggravate one of those conditions?  In general one does try to avoid medications that aggravate certain conditions, however, there are times when that is truly unavoidable and that is where the risk-benefit ratio comes into play.  It  happens alot more than one realizes especially in today's world where many patient's have many co-morbidities and this is unavoidable.

 

It works both ways.  For example, patient A has increased hypertension, but also suffers from asthma.  Beta-Blockers are a good choice for hypertension, but can exacerbate the patient's asthma, therefore we try to avoid these agents as first line therapy.

 

Patient B suffers from muscle weakness, however, Patient B also has very high cholesterol levels.  Your statin drugs are 1st line therapy but unfortunately may cause muscle myalgia (weakness).  Risk Benafit...what is worse, having muscle weakness or high cholesterol which could lead to severe cardiac events.

 

Finally Patient C, suffers from diabetes.  Steriods are know to cause fluxuations in blood sugar levels, especially those who have diabetes.  Patient C suffers a severe spinal cord injury.  Intial treatment of choice is high dose steroids.  This is given despite the known risk of steroids and diabetes.

 

I guess the point I'm trying to make is there are exceptions to every rule.  We throw out the numbers and make treatment options based on what the patient condition presents.  It could  be detrimental to the patient if we went strinctly on percentages

 

As this relates to hockey, I get your point.  I have no problem with the Corsi / Fenwick models as long as everyone know what their limitations and faults are.  No one should truly take Corsi numbers as the final say in a players ability.  If I am repeating what has already been said,  I apologize.

 

 

I thought this thread was discussing whether or not advanced statistics were valid models, not whether or not they should form 100% of your decision-making.

 

If that was the intent of the thread, then my apologies, I misread the intent. 

 

see comments in red.....

 

I knew I was getting in over my head.

[radoran]

Thanks for actually proving my point.

 

You mean "we agree"

 

Again, why is so hard for the adv stats crowd to get what the other side is saying. Use the stats, just don't rely on them as the gospel. That's all I'm saying yet I keep getting arguments that "this" is 75% correct and playoffs show this and that guy had 38.234175 and half % of possession time so he sucks and if you don't believe it your an idiot.

 

I'm pretty sure this is exactly what I am saying.

 

It appears they may be right. Not all the time but we will talk of this again later in the year as the realities get fleshed out. Later on...

 

I believe this is exactly what I have been saying. It's a useful tool for a season but not for individual games.

 

Not sure if you are arguing and if so, who with, but it doesn't appear to be me...

 

really quickly, if possible, how can corsi be used to "evaluate" an individual player. there seems to be a lot of macro factors to drill through. I can see where team corsi can tell a guy that can't watch all the games which teams seem to have the puck more, I don't see where Corsi can be used to point to a individual players weakness or strengths.

 

I believe I have said exactly that.

 

With the potential caveat that seeing how individual players affect other players' Corsi could be an additional factor in player evaluation.

 

And, again, not over the course of a game but over the course of a season and, to a lesser extent, a series of games.

 

It is a tool, and what matters is how you use that tool.

 

It's not Gospel or magic and it is never the be-all and end-all of a player evaluation.

 

That's not to say it has absolutely no value whatsoever.

[brelic]

Do you mean a patient suffering from, your example diabetes is suffering side effects from a drug they are using for diabetes?  Or do you mean they are already suffering from a complication of diabetes, such as eye problems, diabetic neuropathy, increased cardiac risks such as Hypertension, Stroke, increased cholesterol levels and a new medication could aggravate one of those conditions?  In general one does try to avoid medications that aggravate certain conditions, however, there are times when that is truly unavoidable and that is where the risk-benefit ratio comes into play.  It  happens alot more than one realizes especially in today's world where many patient's have many co-morbidities and this is unavoidable.
 
It works both ways.  For example, patient A has increased hypertension, but also suffers from asthma.  Beta-Blockers are a good choice for hypertension, but can exacerbate the patient's asthma, therefore we try to avoid these agents as first line therapy.
 
Patient B suffers from muscle weakness, however, Patient B also has very high cholesterol levels.  Your statin drugs are 1st line therapy but unfortunately may cause muscle myalgia (weakness).  Risk Benafit...what is worse, having muscle weakness or high cholesterol which could lead to severe cardiac events.
 
Finally Patient C, suffers from diabetes.  Steriods are know to cause fluxuations in blood sugar levels, especially those who have diabetes.  Patient C suffers a severe spinal cord injury.  Intial treatment of choice is high dose steroids.  This is given despite the known risk of steroids and diabetes.
 
I guess the point I'm trying to make is there are exceptions to every rule.  We throw out the numbers and make treatment options based on what the patient condition presents.  It could  be detrimental to the patient if we went strinctly on percentages
 
As this relates to hockey, I get your point.  I have no problem with the Corsi / Fenwick models as long as everyone know what their limitations and faults are.  No one should truly take Corsi numbers as the final say in a players ability.  If I am repeating what has already been said,  I apologize.

 

Ha! I should have stayed away from straying into your field of expertise

 

I guess all of the above demonstrates the point I was making. A given medication will have certain side effects and contraindications based on clinical trials, underlying conditions, health conditions, predispositions, etc. So you take that information PLUS what you know about the other conditions and interactions PLUS the patient's history PLUS whatever other factors. All of them have statistics and numbers attached to them, and you make a decision based on all of those inputs. 

 

If that was the intent of the thread, then my apologies, I misread the intent. 

 

I wasn't criticizing... just making sure that *I* didn't misread the intent. I haven't seen a poster say they should be used 100% of the time (or form 100% of decision-making). But maybe someone did and I missed it.

[pilldoc]

Ha! I should have stayed away from straying into your field of expertise   LOL and now you know better....  (just kidding)

 

I guess all of the above demonstrates the point I was making. A given medication will have certain side effects and contraindications based on clinical trials, underlying conditions, health conditions, predispositions, etc. So you take that information PLUS what you know about the other conditions and interactions PLUS the patient's history PLUS whatever other factors. All of them have statistics and numbers attached to them, and you make a decision based on all of those inputs. 

 

 

 

 

I wasn't criticizing... just making sure that *I* didn't misread the intent. I haven't seen a poster say they should be used 100% of the time (or form 100% of decision-making). But maybe someone did and I missed it.

 

No worries Brel... I am in total agreement with you too.  I too thought someone mentioned that Corsi scores should be used close to 100%, if they didn't then that is my bad.  After 12 pages, the replies seem to run together!

 

But yes..I agree with your entire response.

 

On a side note....serious quesion..just curious are there any advanced stats used on or used exclusively for goalies??  (Note: this question is not just for Brel but for anyone else with additional insight) - thanks

[brelic]

On a side note....serious quesion..just curious are there any advanced stats used on or used exclusively for goalies??  (Note: this question is not just for Brel but for anyone else with additional insight) - thanks

 

Good question. Not that I'm aware of other than the stats we use regularly.

 

To me, goalies are like the lawn mowers in Plants vs. Zombies - the absolute last line of defense. They can bail you out occasionally, but if you don't protect them most of the time, the zombies are getting in

[JackStraw]

To me, goalies are like the lawn mowers in Plants vs. Zombies - the absolute last line of defense. They can bail you out occasionally, but if you don't protect them most of the time, the zombies are getting in

 

What about zombie plants though?

 

[pilldoc]

Good question. Not that I'm aware of other than the stats we use regularly.

 

 

 

well seek and ye shall find...aka google is our friend.  Seriously this is what I have found.  I have no idea how merit it is, butI'm skimming throught he article now.

 

http://ingoalmag.com/analysis/gsaa-essential-statistic-evaluating-goaltenders/

 

Advanced statistical knowledge in sports has increased rapidly in the last two decades. Hockey has been a step behind most other sports, but is starting to catch up with the rise of possession statistics such as the wildly popular “Corsi” percentage. The sheer volume of information that is collected now can be enough to make your head spin, but a lot of it is extremely valuable when it comes to evaluating individual performance.

 

When it comes to evaluating a goaltender, three statistics in particular have been looked at throughout most of history. Those three stats are wins, goals against average and save percentage. While all three categories still have some merit, they have become somewhat outdated. Wins and goals against average are better looked at as a team statistic and is actually a very poor way to evaluate an individual goaltender’s performance. Even save percentage isn’t perfect. It can be skewed depending on the quality of team that the goaltender plays for.

 

In baseball, batting average, home runs and runs batted in were the “big three” categories for years. They have been effectively replaced by percentage stats, and statistics that compare a player to a “replacement player.” Essentially, if that player was replaced by a minor leaguer, would that minor league player perform better, or worse? It’s called “WAR” which stands for Wins Above Replacement.

 

There is a relatively new statistic for hockey that has been made available by the folks at Hockey-Reference. It’s similar to baseball’s WAR, and it is called “GSAA” – Goals Saved Above Average. You take the league’s average save percentage and apply it to the amount of shots a particular goalie has faced. You get a number of goals that the average goalie in that league would have surrendered if they faced the same number of shots as the goaltender in question. That number gets compared to the number of goals surrendered by that goaltender, and a plus/minus is created. If a goalie is in the positive, that is how many goals they have saved compared to a league-average goalie. If they are in the negative, then it is safe to assume that they are performing worse than how a league-average goaltender would perform in the same situation.

 

Here are the top ten goaltenders in GSAA in the 2013-14 NHL season, as provided by Hockey-Reference:

 

 

For the full list of GSAA leaders, click here.

 

Every stat has pros and cons, which is why you must look at the entire picture. GSAA does a lot of things better than other stats, but is still not perfect. Here is a breakdown of what it does well, ways that it is biased, and how it can be improved:

 

What GSAA Does Well

• It is a very good stat at equalizing goalies across the league, regardless of the team that they play for. For example, a goalie that has a .925 save percentage and has faced a lot more shots than average is actually playing better than a goalie that has a .930 save percentage and has faced less shots. The first goalie has faced more scoring opportunities, and has saved more goals from going in based on their quality of play.

• It tells you how much a team relies on their goaltending to win games. If a team gives up a lot of shots, but their goalie is continually bailing them out, their goalie will have a very high GSAA number. Those teams are more likely to struggle if their goalie goes into a slump or gets injured. Teams that succeed despite having a goaltender that is in the middle of the pack (or worse) in GSAA are actually more stable, because if their goalie slumps or gets injured, they still have a very good chance to win using a replacement netminder.

• It gives you a physical number of goals saved, rather than a percentage. It’s a stat that can be very shocking. For example, Ben Bishop has saved almost 24 goals from being scored on the Tampa Bay Lightning in 44 games. A number that large will draw a lot of attention. It is a great stat to prove Bishop’s worth to the Lightning. 24 goals saved is a VERY significant number.

What GSAA Does Not Do Well

• Goalies that play more games will accumulate more goals saved. If the stat was expressed as GSAA/per 60 minutes, it would be even more accurate. You would have to turn it into a percentage, though. Only goalies that have played a certain number of games would be able to qualify.

• It does not take penalty killing into account. On average this season in the NHL, goalies have seen a 4.4% drop in save percentage while on the penalty kill compared to even strength. This is a major problem for some goalies that play for a team that is constantly killing penalties. Undisciplined teams will have goalies with a lower save percentage and a lower GSAA as a result.

• It does not take fatigue into account. Saves that are made after the 30 shot mark should be worth more, because goalies that face more shots than the league average will be more tired, and will make less saves due to the poor quality of the team in front of them giving up more scoring attempts.

• It also does not take shot quality into account. This can be related to the number of penalties a certain team commits, or even a team’s quality of defence, but that would be very difficult for any statistic to quantify.

GSAA is definitely not a perfect statistic, but it is one of the best ones available at the moment when it comes to analyzing goaltenders. It has flaws, but it is more accurate than save percentage and should become more widely used.

[pilldoc]

What about zombie plants though?

 

 

looks like Snider when Vinny is serving them popcorn and peanuts!   

[BillyInFourC]

the identical weaknesses exist with corsi and fenwick

 

No they don't have identical weaknesses, as was pointed out, goal +/- is swamped by shooting/save percentage variance. 

Shot differentials have demonstrable predictive value at both the individual and team level, goal +/- doesn't. That predictive value is derived directly from moving the analysis from goals(relatively rare, infrequent events) to shot attempts(extremely common) events. The collection method might be the same, but you are flat out ignoring the key difference in the on ice events being tracked. 

You keep going back to the well of the 9 other skaters on the ice issue... as if this is something that's denied or never addressed by the stat community? Like this is some sort of profound revelation that nobody has ever considered? 

 

Almost all of the efforts in the last few years have been focused a stripping away the contributions of those other skaters and contextual factors. If you want to ignore those efforts, or if you have some fundamental problem with them, write a paper about it and put it up for review. 

 

The 5 guys on the other team, matter a lot in small samples, but matter very little in large ones. Because over the course of a season, you face players of varying degrees of skill level. 

 

The 4 other guys on your team matter a lot because your teammates tend to stay the same over a full season. But with can suss out individual impacts by looking at corsi rel, how you perform relative to those teammates over time, and your impact on teammates numbers over time.

 

You can hem and haw about the other 9 guys on the ice, but at the end of the day, do you really disagree with what Corsi has to say about who is carrying the mail on the Flyers roster this season?  http://goo.gl/oUw3tP Where has this tremendous amount of false data skewed the breakdown? Point to the player whose numbers and usage factors aren't doing him justice? Umberger? Lecavalier? The guys that stat analysts would have told you to bench or demote....and the coaching staff has proceeded to bench or demote? 

 

Couple of points I'd like to reiterate and hammer home. 

 

1. No one is arguing this stat is the end all be all. 

2. Shot metrics are not perfect, never have been, never will be. 

3. It is a useful objective tool to aid scouts and organizations in removing certain human biases that can creep into player evaluations. It is not a replacement for traditional scouting, it is not a replacement for watching the game, and no sane person has ever suggested that. 

4. Numbers should always been put in the proper context. 

If your contention is that this data has nothing valuable to offer to a hockey organization, I just don't see how you can support that claim. The stats would have been a red flag for the Flyers brass to stay away from Umberger and Lecavalier, or MacDonald. Two of those guys have been benched, one seemingly permanently, and I doubt there is any Flyers fan who could argue with a straight face that R.J. has performed above expectations. 

The stats would have told you that Voracek was ready for a breakout, or that Giroux was very quietly the best player on the Flyers in 2010. (Season afterwards he became 1C, season after that he was given the keys to the franchise with the Richards and Carter trades) 

At the team level, the stats would have told you that LA was a good bet in 2011-2012. Or that Colorado was due for a crash this season when their shooting/save percentages cam back to earth. Same thing for the Maple Leafs last season, or the Minnesota Wild before that. 
 

But hey, if you don't like the stats, don't use them?

 

 

 

 

 

 

[idahophilly]

I love how the tune has changed. Anyway, no one said we don't like the stats. Many reacted to the things said here as well as in other threads and forums. Case in point: No one has said they hate or would not use the stats as stated above. Pls don't over extrapolate what others say. You may respond if you wish and have the last word. It's not worth my time, especially when it's a moving target. You are entitled to your opinion. I would have gone with a smaller footprint at the 9 posting point but you did say it was a feisty group in our PM. Anyway, so long as you are having fun...

[Podein25]

Mrs Podein has a stats degree. I ran this thread past her to get her take. Her response, and I quote verbatim (f-bombs redacted):

 

"stats shmats, you can't over-think hockey: take the shortest route to the puck and arrive in ill-humor"  

[JackStraw]

Deleted 

[JackStraw]

take the shortest route to the puck and arrive in ill-humor

 

That's brilliant. Why has no one ever thought of that before???

[Podein25]

Why has no one ever thought of that before???

 

 

Poor understanding of geometry maybe.

[BillyInFourC]

I love how the tune has changed.

 

How so? It's been a contentious discussion with aziz from the start. 

Anyway, no one said we don't like the stats. Many reacted to the things said here as well as in other threads and forums. Case in point: No one has said they hate or would not use the stats as stated above.

 

I think aziz as state quite clearly that he doesn't like corsi, particularly when applied to individuals. I never used the word "hate". 

 

Anyway, so long as you are having fun...

 

It's a contentious discussion, that doesn't mean it has been an unproductive one. 

I hope I've at least corrected some of the popular misconceptions about the stats and how they are used, the biggest one being that stat guys think the stat is the end all be all of evaluations.

[JackStraw]

Poor understanding of geometry maybe.

 

Ahh, but the strawberries that's... that's where I had them. They laughed at me and made jokes but I proved beyond the shadow of a doubt and with... geometric logic... that a duplicate key to the wardroom icebox DID exist

[Podein25]

Ahh, but the strawberries that's... that's where I had them. They laughed at me and made jokes but I proved beyond the shadow of a doubt and with... geometric logic... that a duplicate key to the wardroom icebox DID exist

 

 

 

Ah, Old Yellowstain

[TropicalFruitGirl26]

@pilldoc

 

Very nice on that GSAA thing.

I honestly didn't know that existed.....probably because likeyou said, it is still relatively new.

 

I studied that table you provided a bit and read the pros and cons and I must say I am intrigued by that.

In trying to personally evaluate hockey players, I find myself looking at all sorts of stats, situations, and of course, watching them play when I can (having the Center Ice package, working from home and some down time when kids are in school, hubby is at work, and I have no other pressing matters on hand helps there!).....but always understanding that no single observation or stat can tell the whole story...as it should be.

 

But this GSAA is interesting to me.

You mentioned that it sorta paints a picture of how much a team may rely on goalies to bail them out, and as an example, just looking at that table of the top goalies in the GSAA category, I must say, it rings pretty true.

 

Bishop, Bernier, Scrivens, and Varlamov for example DO play for teams that tend to give up many shots on goal (the respective teams are more offense oriented, have a bit looser defensive coverage than some others, and rely heavily on their goaltender to bail them out of missed defensive assignments) and the teams represented by those 4 goalies WOULD suffer should the goaltender go down with injury or underperform.

Craig Anderson isn't on that top list for last season, but I would throw him in that category with the Sens as well...though I understand he missed quite a bit of time last season and underperformed when he WAS in.....kinda proving the GSAA theory a bit, no?

 

Ben Bishop going down with injury for the playoffs last season certainly is an argument for the GSAA stat as well.

 

I also see some goaltenders on that list, that quite frankly, I am a bit surprised to see on there.

Tuuka Rask for example, Josh Harding, Carey Price, and maybe even Cam Talbot.

Why?

Because the teams those 'tenders play for tend to be a bit more defensively responsible and DON'T give up as many scoring chances that the goalie would need to constantly bail the team out of.

Then again, if someone argues that, let's say Tuuka Rask for example, is good ONLY because he plays behind a better Boston defense, one could look at that GSAA stat and say, "Well, Boston does have a good defense, but when they mess up, Rask is there to make the saves he needs to".

 

Rask may STILL not thrive as much under the kind of defensive coverage, say, the Lightning provide as opposed to the Bruins, but looking at this, it is no longer simply a matter of saying "Rask is only good because his defense is elite".

Obviously, even elite D's breakdown and if Rask were just an 'average' goalie, then he would still probably let in more goals than he should...and this shows that he still makes stops when he needs to.

Same could be said of Harding and Talbot, though in Talbot's case, since he is primarily a backup, his exposure to such situations are more limited.......one of the flaws, I suppose of this particular stat.

 

Again, all very interesting and different ways to look how certain goalies perform aside from just GAA and S%.....flaws and all.

[pilldoc]

@TropicalFruitGirl26

 

Thanks for reading it.  It was pretty interesting.  With all this alk around advanced hockey stats it made be wonder if there were any for goaltenders.  So a quick google search produce the article you read.  Another good stat is if you click on the full list, they list another stat.  Quality start.  It starts with SV% > than the league average and or a SV% of 88.5% when facing 20 or less SOG.

 

Surprising Varlamov was #1 with 43 out of 59 quality starts.  Next was Rask, Bishop and then Price.  For comparison's sake, Mason was #9 with 33 out of 58 (56%).  MAF was rounds out the top 10 with 32 out of 62 (51%). 

 

Another interesting fact, Niemi from SJ was tied with the most games started at 62, yet his Qualiity Starts was only 48%.  It was a wonder SJ did as well as they did.

 

yeah the more I look the more interesting it becomes.  Glad you enjoyed them.

 

EDIT...one more interesting thing regarding Niemi, he was middle of the pack as far as quality starts go...yet he was tied for 2nd in overall wins.  So is that a credit to Niemi or a credit to the Sharks for overcoming a mediocre goalie.  Even though Niemi has a cup, based on stats, is he just a mediocre, middle of the pack netminder?

[J0e Th0rnton]

@TropicalFruitGirl26

 

Thanks for reading it.  It was pretty interesting.  With all this alk around advanced hockey stats it made be wonder if there were any for goaltenders.  So a quick google search produce the article you read.  Another good stat is if you click on the full list, they list another stat.  Quality start.  It starts with SV% > than the league average and or a SV% of 88.5% when facing 20 or less SOG.

 

Surprising Varlamov was #1 with 43 out of 59 quality starts.  Next was Rask, Bishop and then Price.  For comparison's sake, Mason was #9 with 33 out of 58 (56%).  MAF was rounds out the top 10 with 32 out of 62 (51%). 

 

Another interesting fact, Niemi from SJ was tied with the most games started at 62, yet his Qualiity Starts was only 48%.  It was a wonder SJ did as well as they did.

 

yeah the more I look the more interesting it becomes.  Glad you enjoyed them.

 

EDIT...one more interesting thing regarding Niemi, he was middle of the pack as far as quality starts go...yet he was tied for 2nd in overall wins.  So is that a credit to Niemi or a credit to the Sharks for overcoming a mediocre goalie.  Even though Niemi has a cup, based on stats, is he just a mediocre, middle of the pack netminder?

Chicago won a cup. Niemi was a beneficiary of great defense. Chicago knew how to limit shots to low shots well. All 7 goals San Jose Scored against Niemi when they were swept were shots they were able to get away from defense and shoot high. Because Niemi is utterly awful at stopping any shot 18 inches off the ice. Sharks defense of course, is not as polished as Chicago's to say the least. Pass east west and shoot high on Niemi and it is almost certainly going to set the goal light off.

 

Take it from a guy who has seen him wear teal for 40 playoff games and get zero shutouts. He is never going to steal a game for us. He also handles the puck so badly he usually helps the other team an gives it to them. I'd have to say, at least 35% of the time Niemi touches the puck to play it, it ends up in the back of our own net. He goes to play the puck because he thinks it is the right play, but then has no idea where to pass it and handles it like a live grenade.

[fanaticV3.0]

Yes and since every player on the ice gets a plus or a minus after each shot-at-net, Corsi awards plus marks to players even if they did nothing to help create the shot-at net, and the system assigns minus marks to players who made no mistake on the shot-at-net against.

 

 

This is exactly what the +/- stat does as well. The only reason I argue against +/- so much and not the Corsi is because nobody takes that Corsi bullsh-t seriously, whereas people actually use +/- in when making a case (which is mind-blowing to me). They're both flawed, to the point of being irrelevant, stats. Stats are supposed to be a statement of fact. If it doesn't tell the full story for all players, it is of no use to me.

[Podein25]

This is exactly what the +/- stat does as well. The only reason I argue against +/- so much and not the Corsi is because nobody takes that Corsi bullsh-t seriously, whereas people actually use +/- in when making a case (which is mind-blowing to me). They're both flawed, to the point of being irrelevant, stats. Stats are supposed to be a statement of fact. If it doesn't tell the full story for all players, it is of no use to me.

 

 

I'm not sure where all the mind-blowing is coming from. Nobody I know ever took plus/minus as anything but a proxy. By definition, a proxy in statistics is "a variable that is not in itself directly relevant, but that serves in place of an unobservable or immeasurable variable.  In order for a variable to be a good proxy, it must have a close correlation, not necessarily linear or positive, with the variable of interest."

 

So, we could debate whether plus/minus is a good or bad proxy, which is kind of what we've been doing in this thread, since Corsi also is a proxy as far as I can tell. But maybe we could dispense with the hyperbole and straw-man arguments?

 

One of the things that made plus/minus a good proxy, I would submit, is precisely it's simplicity - you were either on the ice or you weren't when a goal was scored. Kids just learning the game could understand it easily. Even the dumbest hockey players can count.

 

In my experience, plus/minus served a very useful accountability function within a team, from the time I was very young to this day. Because, regardless of it's flaws, to be a consistently minus hockey player, over the course of a season, is to be either extraordinarily unlucky or, more likely, lazy and indifferent to the defensive side of the puck. And while a player could point to a particular play where they were unlucky and got a minus, I could always point to the several plays where they failed to pick up a man, didn't try, acted selfishly etc etc. In short, I could point to several plays where their actions were directly responsible for the puck going in our net. In all those cases they not only deserved a minus (plus a slap in the head), but that minus was a very good proxy indeed. 

[JackStraw]

a proxy in statistics is "a variable that is not in itself directly relevant, but that serves in place of an unobservable or immeasurable variable.

 

You got that from Mrs P, admit it!

 

Are we talking unknown unobservables/immeasurables or known unobservables/immeasurables?

 

I think +/- is a useful stat, but like all stats context is key. It's probably worth pointing out (in case anyone doesn't know it) that +/- was devised by Montreal's coaching staff back in the '50s so it's not as if it was invented by some geek who didn't know anything about hockey. Also, if you look at the list of career +/- leaders it's a pretty impressive group. So there must be something to it.

 

Top 5: Season high

• Bobby Orr, 1970–71, +124
• Larry Robinson, 1976–77, +120
• Wayne Gretzky, 1984–85, +98
• Dallas Smith 1970–71, +94 (Orr's defense partner, so there's maybe a bit of an assist due here  )
• Guy Lafleur, 1976–77, +89

Top 5: Career high

• Larry Robinson, career, +730
• Bobby Orr, career, +597
• Ray Bourque, career, +528
• Wayne Gretzky, career, +518
• Bobby Clarke, career, +506