Plus/Minus the Stat that tastes like Pie

[LGR4GM]

Have at it. Trpm. Regular PM. Whatever you want about plus minus. 

[SwampD]

I think that when a player is on the ice and the team scores, that's good.

I think that when a player is on the ice and the other team scores, well, that's bad.

[Randall Flagg]

In the proper thread,

 

 

The problem I have with all of your arguments when it comes to this kind of stuff is that you basically just say, "no, you are wrong and here's some number to prove it," yet you yourself have said that there are still a lot of thing that take place on the ice that there are no measurements for. Like a nice D pass out of the zone after stealing the puck, or a slick pass of the boards in the neutral zone to a rushing forward, which might result in a goal that leads to no points for those players, yet would not have happened had they not made those plays.

 

I know that +/- and, well, wins/losses, gets scoffed at around here sometimes because they are course measurements. but just because there are finer measurements out there, doesn't mean they should just be completely discounted.

 

I would love to see a comparison of goal diff/possession/standings for the past couple of years.

 

Also, if I'm a coach, and a player is on the ice, and is a +40,… even if he "does nothing to make that happen",… you better believe I'm going to keep putting him on the ice.

I don't think I do this at all. For the last major crusade I was on, I routinely qualified literally every point I made. I invented my own thing to count so as not to only rely on numbers you can look up. I also spent $50 and 25 hours collecting video to appease the eye test, immediately dumping the software when I was done with it. And to be honest there's no way I'm doing that again if it's just going to get interpreted the way it has on several occasions later on.

 

Even in this argument, I've said at least twice that my main issue with pi is not the number he's using, but his utter refusal to qualify it or to bring anything else to the analysis, he stands by it as a universal absolute judging metric the way you accuse me of doing, even though you can go back to the other thread and see me stress that you need to combine numbers with visual analysis as much as possible. Maybe my head's up my own ass, but I really think you're misrepresenting how I argue. 

 

There are 100 other posters that have used +/- this season in their discussions and pi is the only one I post in response to because of his unique posting style and because of the brashness/absoluteness that you claim to disdain but watch him do passively.

 

Team goal differential and winning absolutely goes hand - in - hand, which is why I made the last post that I did in the last thread - for teams, goal differential is a good way to tell which teams are the best this year. But to get a good goal differential, you have to score and allow goals, and how many goals a player has this year does a poorer job statistically at predicting how many goals they'll have next year than the stats I mention, which is why they were invented and why they are used. It has been shown that shot attempts are a great proxy for possession and that possession metrics just do a better job. They aren't perfect, I don't claim they are, I don't claim you can fill out a corsi bracket and guarantee a cup winner. Sports wouldn't be sports if you could do that, and furthermore, somebody who likes stats generally understands the concept of sample size and that 7 games is a small one. That's why I laughed at a recent post I saw that said somethign like "See, the Sabres outshot their opponents for three games but lost, fancy stats are overrated"...I mean come on, do people really think that people who like statistics think the world works in such a way that a good metric will guarantee a winner of every game for all time? 

 

If you're a coach you don't need to look at +/-, you watch your players play, practice. You see every second of every game and have the right mind to analyze it. There's no ###### way a coach says in a vacuum "well, this player has this number, better keep putting him out there". He uses what he synthesizes in his brain as a result of watching, and then MAYBE checks the stats too if he's  a nerd, but plenty of them openly trash doing that. 

 

I've already mentioned this too, but I'll say it again. If you watch every team all the time, you don't need stats as much, but we naturally don't have time for that and rely on numbers to tell us stories about, say, Nutivaara on Columbus. Ain't nobody gonna tell me they know anything about this kid. The only beef I have here is when pi says "his TRPM is this so he must be good", and if you are telling the truth about your problem being "here's a number, you're wrong" then this would have bothered you at some point. My counter is to say "look, when you use metrics, developed statistical principles (which when applied to other fields, let us do quantum mechanics or thermodynamics or find the Higgs Boson etc.) tell you that these things do a better job, so if you want your argument to be taken seriously AND want to use numbers, it is better to use a bunch of these stats and couple it to an analysis that involves the actual hockey we see". 

Which I try as hard as possible to do in my essays, but apparently gets hand-waved away as a "muh stats" argument. 

Edited March 28, 2018 by Randall Flagg

[woods-racer]

Flagg....

 

:worthy:

[SwampD]

In the proper thread,

 

I don't think I do this at all. For the last major crusade I was on, I routinely qualified literally every point I made. I invented my own thing to count so as not to only rely on numbers you can look up. I also spent $50 and 25 hours collecting video to appease the eye test, immediately dumping the software when I was done with it. And to be honest there's no way I'm doing that again if it's just going to get interpreted the way it has on several occasions later on.

 

Even in this argument, I've said at least twice that my main issue with pi is not the number he's using, but his utter refusal to qualify it or to bring anything else to the analysis, he stands by it as a universal absolute judging metric the way you accuse me of doing, even though you can go back to the other thread and see me stress that you need to combine numbers with visual analysis as much as possible. Maybe my head's up my own ass, but I really think you're misrepresenting how I argue. 

 

There are 100 other posters that have used +/- this season in their discussions and pi is the only one I post in response to because of his unique posting style and because of the brashness/absoluteness that you claim to disdain but watch him do passively.

 

Team goal differential and winning absolutely goes hand - in - hand, which is why I made the last post that I did in the last thread - for teams, goal differential is a good way to tell which teams are the best this year. But to get a good goal differential, you have to score and allow goals, and how many goals a player has this year does a poorer job statistically at predicting how many goals they'll have next year than the stats I mention, which is why they were invented and why they are used. It has been shown that shot attempts are a great proxy for possession and that possession metrics just do a better job. They aren't perfect, I don't claim they are, I don't claim you can fill out a corsi bracket and guarantee a cup winner. Sports wouldn't be sports if you could do that, and furthermore, somebody who likes stats generally understands the concept of sample size and that 7 games is a small one. That's why I laughed at a recent post I saw that said somethign like "See, the Sabres outshot their opponents for three games but lost, fancy stats are overrated"...I mean come on, do people really think that people who like statistics think the world works in such a way that a good metric will guarantee a winner of every game for all time? 

 

If you're a coach you don't need to look at +/-, you watch your players play, practice. You see every second of every game and have the right mind to analyze it. There's no ###### way a coach says in a vacuum "well, this player has this number, better keep putting him out there". He uses what he synthesizes in his brain as a result of watching, and then MAYBE checks the stats too if he's  a nerd, but plenty of them openly trash doing that. 

 

I've already mentioned this too, but I'll say it again. If you watch every team all the time, you don't need stats as much, but we naturally don't have time for that and rely on numbers to tell us stories about, say, Nutivaara on Columbus. Ain't nobody gonna tell me they know anything about this kid. The only beef I have here is when pi says "his TRPM is this so he must be good", and if you are telling the truth about your problem being "here's a number, you're wrong" then this would have bothered you at some point. My counter is to say "look, when you use metrics, developed statistical principles (which when applied to other fields, let us do quantum mechanics or thermodynamics or find the Higgs Boson etc.) tell you that these things do a better job, so if you want your argument to be taken seriously AND want to use numbers, it is better to use a bunch of these stats and couple it to an analysis that involves the actual hockey we see". 

 

Which I try as hard as possible to do in my essays, but apparently gets hand-waved away as a "muh stats" argument. 

 

Like myself, I think you just need to remember to stop arguing with people who have found religion.

Edited March 28, 2018 by SwampD

[pi2000]

Games are decided by goals. Scoring a goal is a quantifiable statistic. Preventing a goal is just as important as scoring a goal... so how do you quantify it? Plus/minus.

[WildCard]

Games are decided by goals. Scoring a goal is a quantifiable statistic. Preventing a goal is just as important as scoring a goal... so how do you quantify it? Plus/minus.

Light is given by the sun. The sun exists in space. So why is space dark? 

[SwampD]

Light is given by the sun. The sun exists in space. So why is space dark? 

Buckyballs.

:lol:

[North Buffalo]

Games are decided by goals. Scoring a goal is a quantifiable statistic. Preventing a goal is just as important as scoring a goal... so how do you quantify it? Plus/minus.

There is not enough info in that stat. Agree with Flagg... ? Ill need to dig out my old stat books but confidence levels. Also look at political campaign stats and how often they are wrong because of sample sizes and measuring errors, in this case accounting for talent of lines played against, goalie in net and probably other factors making predictive inferences difficult at best.

[pi2000]

There is not enough info in that stat. Agree with Flagg... ? Ill need to dig out my old stat books but confidence levels. Also look at political campaign stats and how often they are wrong because of sample sizes and measuring errors, in this case accounting for talent of lines played against, goalie in net and probably other factors making predictive inferences difficult at best.

 

Exactly.   

 

Which is why I came up with TRpm.   

[North Buffalo]

Exactly.

 

Which is why I came up with TRpm.

 

Appreciate the attempt and think it provides a glimpse, but as Flagg implied a shortage of info can really skew these stats short term. TRpm may be a better stat as a longer term predictor after a few years of data, but hard to use for a developing player. Edited March 28, 2018 by North Buffalo

[pi2000]

Appreciate the attempt and think it provides a glimpse, but as Flagg implied a shortage of info can really skew these stats short term. TRpm may be a better stat as a longer term predictor after a few years of data, but hard to use for a developing player.

 

Shortage of info?  We have an entire season of data, how much do you consider acceptable?  

 

The Stanley Cup isn't awarded to the team who played best over the past 2, 3, 4, or 5 seasons.    You have one season, 82 games to make the playoffs.   Statistics accumulated over that timeframe matter.

 

I use TRpm to illustrate which players contributed to the team's overall success, or lack thereof.    And by "success" I mean wins and losses, which are quantified by how many goals you score and how many you give up... NOT SAT%.    TRpm doesn't predict a players development, but it can tell you if the player is contributing to the team's success (wins/losses).   

 

Casual fans don't appreciate how important it is to play sound defensive hockey.    Making a quality play to prevent a goal is just as important as scoring a goal, but it's not something that can be easily quantified.   As a former player, when matching up against top offensive players, your job is to keep them to the outside.   Let them shoot from distance and bad angles instead of challenging them one on one.    This skews SAT%.    Just because you're giving up shot attempts, doesn't mean you're playing poor defensively.     

 

Somebody brought up league standings based on goal diff vs. SAT%...   have a look yourself:

 

https://www.nhl.com/standings/2017/league- sorting by goal diff moves maybe a few teams up and down 2-3 spots at most, but the standings are almost perfectly aligned with goal team goal differential.

 

SAT% - http://www.nhl.com/stats/team?report=teampercentages&reportType=season&seasonFrom=20172018&seasonTo=20172018&gameType=2&filter=gamesPlayed,gte,1&sort=shotAttemptsPctg- 3 of the top 4 teams aren't even going to make the playoffs!  

 

 

I've also repeatedly said that TRpm differences of single digits between players isn't very telling because of variables like matchups and zone starts, etc, but when you have a guy like ROR who takes just about all dzone draws with a +10 TRpm and a guy like Okposo with a -14 TRpm, it's a significant data point that shouldn't be ignored.    

[WildCard]

Your change in avatar should be a ban-able offense, pi

[Sabres Fan in NS]

Your change in avatar should be a ban-able offense, pi

 

You idiot ( ;) I mean no disrespect ... )

 

Anything to get him to change it from that terrible one is good, eh?

 

I actually welcome the new avatar, but noticed that pi does not have the trade mark showing on it.

[pi2000]

You idiot ( ;) I mean no disrespect ... )

 

Anything to get him to change it from that terrible one is good, eh?

 

I actually welcome the new avatar, but noticed that pi does not have the trade mark showing on it.

 

ahhhH!  fixed.  hahaha

[LGR4GM]

So plus minus matters going forward? It tells us something?

Edited March 28, 2018 by Skurk Liger

[North Buffalo]

So plus minus matters going forward? It tells us something?

Again it is only a snap shot in time, unless like ROR on faceoffs he has been doing it a long time. Guys who play against best offensive guys or penalty kills are intuitively gonna give up more goals. Guys who are sheltered to go against less stong defensemen will generally have higher plus minus if they have good offensive skills. Overall team + - means more to me than individuals over 1 season. Eg Stafford comes to mind in a contract year. Teams get suckered into giving these guys big contracts for 1 seasons performance v consistent over and I am not sure where the significance increases 2 or 3 years that plus minus becomes more telling. Dadnov is a good example; does he keep it up, does playoff hockey effect his performance and do any other guys have good playoffs but mediocre regular season. Ovi has been inconsistent in the playoffs. Does TRpm mean much with one year of data or can a player adjust and get better or worse. Again, why Flagg suggests there is a data set issue. Not that it doesnt mean anything, but its significance may not mean as much as we would all like. It would be easier. More data is always better. Figuring out the right weighting is as well. Can you tell me what weighting you use. You probably already stated it but humor me and remind me.

[Randall Flagg]

Like myself, I think you just need to remember to stop arguing with people who have found religion.

As a member of the church of Moulson I am offended by this comment.

[mjd1001]

I think plus-minus has been bashed a little too much by the analytics crowd. Is it a perfect stat? No.  Can there be a lot of factors that skew it quite a bit?  Yes. But for 80% of the league, I think it can be a good too, not just discarded.

 

I hear and read some people saying it is a flawed stat because it doesn't agree with what they want it to. Case and point? Ryan O'Reilly.  He is a big minus player, and I have read people on this board saying it is a flawed stat because O'Reilly is not a 'minus' talent player because of his defensive play. However, earlier in the year, I have seen quite a few goals scored against the Sabres where he got a minus...where the goal was a result of a turnover of his...OR with a little more effort backchecking (or a little more skating speed) he could have reached the eventual goal scorer...but instead he was 2-3 feet directly behind the goal scorer when the shot was taken.

 

I do know this for sure...if I was in the front office of an NHL team, I would come up with my own +/- stat.  Every goal that was scored, I'd assign a % (out of 100) and assign that to every one of my players on the ice...shooting..passing..winning a battle....what percentage of that 'goal' were they responsible for.    ON the minus side of things, I would do the same...then at the end of the year, I'd tally up the totals and use that as a tool internally.

Edited March 28, 2018 by mjd1001

[Randall Flagg]

I think plus-minus has been bashed a little too much by the analytics crowd. Is it a perfect stat? No.  Can there be a lot of factors that skew it quite a bit?  Yes. But for 80% of the league, I think it can be a good too, not just discarded.

 

I hear and read some people saying it is a flawed stat because it doesn't agree with what they want it to. Case and point? Ryan O'Reilly.  He is a big minus player, and I have read people on this board saying it is a flawed stat because O'Reilly is not a 'minus' talent player because of his defensive play. However, earlier in the year, I have seen quite a few goals scored against the Sabres where he got a minus...where the goal was a result of a turnover of his...OR with a little more effort backchecking (or a little more skating speed) he could have reached the eventual goal scorer...but instead he was 2-3 feet directly behind the goal scorer when the shot was taken.

Then I think you're building a strawman and not talking to actual "members of the analytics crowd" (of which I am not). Because they can tell you the exact results of the studies done on large data sets and what they say about the predictability of the stats. 

 

"It doesn't agree with what they want to"

No, they generally use it to help shape how they view the event. There is one group in this "debate" that is seeking to support preconceived conclusions, and it isn't the one taking large data sets and running analyses with them with proper controls and then publishing their results.

[pi2000]

I think plus-minus has been bashed a little too much by the analytics crowd. Is it a perfect stat? No. Can there be a lot of factors that skew it quite a bit? Yes. But for 80% of the league, I think it can be a good too, not just discarded.

 

I hear and read some people saying it is a flawed stat because it doesn't agree with what they want it to. Case and point? Ryan O'Reilly. He is a big minus player, and I have read people on this board saying it is a flawed stat because O'Reilly is not a 'minus' talent player because of his defensive play. However, earlier in the year, I have seen quite a few goals scored against the Sabres where he got a minus...where the goal was a result of a turnover of his...OR with a little more effort backchecking (or a little more skating speed) he could have reached the eventual goal scorer...but instead he was 2-3 feet directly behind the goal scorer when the shot was taken.

This why I came up TRpm. Last season ROR was a +23, this season he's a +17 I think, somewhere in the high teens. He's not the reason this team sucks... which is obvious, but TRpm helps to confirm that. Edited March 28, 2018 by pi2000

[mjd1001]

This why I came up TRpm. Last season ROR was a +23, this season he's a +17 I think, somewhere in the high teens. He's not the reason this team sucks... which is obvious, but TRpm helps to confirm that.

What exactly is TRpm..how is it calculated? IF you have a link to a prior thread explaining it, I'd like to learn more.

[pi2000]

What exactly is TRpm..how is it calculated? IF you have a link to a prior thread explaining it, I'd like to learn more.

https://forums.sabrespace.com/topic/24217-sabres-prospects-2016-17/?p=963912

[SwampD]

Then I think you're building a strawman and not talking to actual "members of the analytics crowd" (of which I am not). Because they can tell you the exact results of the studies done on large data sets and what they say about the predictability of the stats. 

 

"It doesn't agree with what they want to"

No, they generally use it to help shape how they view the event. There is one group in this "debate" that is seeking to support preconceived conclusions, and it isn't the one taking large data sets and running analyses with them with proper controls and then publishing their results.

This kinda hits on another issue I have with AS and maybe I should bring it up in that thread. Can't they both be right?

 

We always hear about needing a larger sample size, over several seasons. We also hear about how momentum isn't real. A stats guy would tell me that the world is smoother than a cueball, but I would think that the momentum of falling off a 1000ft cliff would be quite real.

 

I believe that there are more coarse stats over smaller sample sizes (even very small like a period or even a couple of shifts) that actually can be predictive. Maybe they are just predictive for a shorter amout of time. They are real though. I think.

Fractal Stats.

Edited March 28, 2018 by SwampD

[Randall Flagg]

This kinda hits on another issue I have with AS and maybe I should bring it up in that thread. Can't they both be right?

 

We always hear about needing a larger sample size, over several seasons. We also hear about how momentum isn't real. A stats guy would tell me that the world is smoother than a cueball, but I would think that the momentum of falling off a 1000ft cliff would be quite real.

 

I believe that there are more coarse stats over smaller sample sizes (even very small like a period or even a couple of shifts) that actually can be predictive. Maybe they are just predictive for a shorter amout of time. They are real though. I think.

Fractal Stats.

"they" as in "advanced" versus "non-advanced" stats? I mean yeah they both have successes. I would rather have plus-minus than no numbers at all, if I wanted numbers to look at. 

 

The rest of that, I don't necessarily feel I 

 

(a.) represent the mythical entity that is the stats community

(b.) know enough

 

to address, which is why True needs to get his butt in here.

 

Though I'm cautious of the way you phrase "we often hear about how we need large sample sizes". 

 

The fact is, noise dampens in bigger sample sizes which is why they tried starting to track things that happen 30 times per game instead of 2. It just becomes a stronger data set no matter what you're measuring. I don't think there's anything more to it than that. 

 

 

Edited March 28, 2018 by Randall Flagg