In defense of Plus/Minus as a stat

[mjd1001]

Ok, I'll start out with saying its not perfect. Its had to compare plus-minus of a player from one team to another. It is not a perfect stat.

But, it is what it is. It is a measure (not counting special teams) that shows how many goals for you were on the ice for vs how many goals against you were on the ice for. Its useful in some, maybe many ways. It doesn't tell the whole story, but it can be a starting point to look at certain players and ask "why"?

With that said, look at the Sabres currently.  Who are the best plus-minus players on the team?

Tuch +11, Bryam +7, McLeod +6, Thompson +4, Dahlin +4, Zucker +3, Greenway +3.

Who are the worst?

Quinn -11, Peterka -5, Bryson -4, Cozens -4.

The positive guys are ones who produce offensively AND are not a consistent source of awful defensive plays.  

The negative guys (Quinn, Cozens, and Bryson to a lesser extent) are guys who play either a defensively unaware game, or at worst a selfish game. 

 

I guess my point is if I looked at the team play BEFORE looking at the stats, who helps the team the most without hurting it?  I'd put Thompson, Dahlin and Tuch at the top of the list (maybe greenway before he got hurt).

And if I said, who is constantly hurting the team, who makes plays that ACTIVELY makes this team worse? I'd put Quinn and Cozens up there.

So, plus-minus is not perfect, but to me it tells more of a picture than many people who have discarded it as a state say.

[LGR4GM]

It tells us the +/- even strength goal differential and you are correct, it should not be used to compare across teams. The issue is there are just more stable and better stats that give us a better understanding. For example xGF%. Jack Quinn is 7th on the team (out of 15 forwards) in xGF% at 49.44 which is still below 50% so he needs to be better but he is about even at getting and giving x goals. However, and I couldn't find the exact number because I forgot where to look, Quinn is -4 or 5 expected goals to date. His plus minus does reflect that but at 5 on 5 he is actually not a major negative. He just hasn't scored to go with his chances which we could argue is the only thing that matters. 

Further +/- does not account for the teammates you have. Everyone wants to crap on Quinn because he isn't scoring, which is fair btw. However a closer look tells me that while Quinn isn't doing great right now, Cozens is a ***** boat anchor tied to his neck. Quinn with Cozens has an xGF of 44.92% but you put Quinn with anyone else and it jumps up to 54.95xGF% and Cozens stays right at 44.53%.  They have a 2nd line center problem and Quinn isn't playing well enough this year to compensate. If I were Buffalo, I would be seriously considering whether or not we should keep Cozens. 

Basically +/- tells us a bit but there are more stats out there that correlate to good or bad hockey better. Quinn's on ice sh% is 4.32 which is worse than every single player not named Jiri Kulich. Because he isn't scoring goals, his line isn't scoring goals, and subsequently his +/- is awful.

Our best forwards are Tage, Tuch, and Benson. They drive quality chances in favor of the Sabres. Peterka and Zucker are still looking okay there too. All these stats are at 5v5 where the majority of +/- comes from. 

[GASabresIUFAN]

Don't forget what bad goaltending does to these stats.  Innocent plays with guys like Georgiev in goal end up in the back of the net.  

 

[LGR4GM]

Sidenote: Zach Benson is going to be amazing in this league for a long time. When you put him with Cozens and Quinn, they jump up to a 56.98xgf%. Thats shocking because their HDCF is awful at 35.29. Also, Benson rocks a 57.77xgf with everyone else not named Cozens/Quinn. The dude does 2 things better than anyone else on this team, gets possession of pucks, puts them in areas where teammates should be scoring. 

[GASabresIUFAN]

1 minute ago, LGR4GM said:

Sidenote: Zach Benson is going to be amazing in this league for a long time. When you put him with Cozens and Quinn, they jump up to a 56.98xgf%. Thats shocking because their HDCF is awful at 35.29. Also, Benson rocks a 57.77xgf with everyone else not named Cozens/Quinn. The dude does 2 things better than anyone else on this team, gets possession of pucks, puts them in areas where teammates should be scoring. 

I'm in full agreement here.  He still needs to get stronger on his skates, but that is vastly improved year over year.  

[Pimlach]

11 minutes ago, LGR4GM said:

It tells us the +/- even strength goal differential and you are correct, it should not be used to compare across teams. The issue is there are just more stable and better stats that give us a better understanding. For example xGF%. Jack Quinn is 7th on the team (out of 15 forwards) in xGF% at 49.44 which is still below 50% so he needs to be better but he is about even at getting and giving x goals. However, and I couldn't find the exact number because I forgot where to look, Quinn is -4 or 5 expected goals to date. His plus minus does reflect that but at 5 on 5 he is actually not a major negative. He just hasn't scored to go with his chances which we could argue is the only thing that matters. 

Further +/- does not account for the teammates you have. Everyone wants to crap on Quinn because he isn't scoring, which is fair btw. However a closer look tells me that while Quinn isn't doing great right now, Cozens is a ***** boat anchor tied to his neck. Quinn with Cozens has an xGF of 44.92% but you put Quinn with anyone else and it jumps up to 54.95xGF% and Cozens stays right at 44.53%.  They have a 2nd line center problem and Quinn isn't playing well enough this year to compensate. If I were Buffalo, I would be seriously considering whether or not we should keep Cozens. 

Basically +/- tells us a bit but there are more stats out there that correlate to good or bad hockey better. Quinn's on ice sh% is 4.32 which is worse than every single player not named Jiri Kulich. Because he isn't scoring goals, his line isn't scoring goals, and subsequently his +/- is awful.

Our best forwards are Tage, Tuch, and Benson. They drive quality chances in favor of the Sabres. Peterka and Zucker are still looking okay there too. All these stats are at 5v5 where the majority of +/- comes from. 

I have been saying since preseason, we have a 2C problem.   We are weak up the middle.  This also hurts the PP, not having a playmaking center forces us to do the same things over and over.

Quinn has his share of problems and playing with Cozens exacerbates them.   But the catch 22 is that Quinn is not scoring with any of the centers, so putting him with Tage or McLeod hasn't helped anything.  

[ska-T Palmtown]

58 minutes ago, LGR4GM said:

It tells us the +/- even strength goal differential and you are correct, it should not be used to compare across teams. The issue is there are just more stable and better stats that give us a better understanding. For example xGF%. Jack Quinn is 7th on the team (out of 15 forwards) in xGF% at 49.44 which is still below 50% so he needs to be better but he is about even at getting and giving x goals. However, and I couldn't find the exact number because I forgot where to look, Quinn is -4 or 5 expected goals to date. His plus minus does reflect that but at 5 on 5 he is actually not a major negative. He just hasn't scored to go with his chances which we could argue is the only thing that matters. 

My bone to pick with the xGF% family of stats is that is looks at the "average" shooting percent across the league for the spot on the ice where the shots are taken and does simple math to say player X and their linemates get these chances and give up these other chances. It does not account for the actual skill of the person taking the shots, etc. So, things can be skewed one way or the other, given that a MacKinnon vs Lafferty getting the same scoring chance is way different.

Again, not saying it does not have its spot - but it also requires several grains of salt and other information to be truly useful*

*the author fully reserves the right to have completely misunderstood xGF%, etc and thus further cluttered the thread with non-sensical babble.

[triumph_communes]

I’ve only ever looked at +\- for defenseman and third/fourth liners, but not sure what pitfalls that puts me into. 

[Jorcus]

1 hour ago, LGR4GM said:

It tells us the +/- even strength goal differential and you are correct, it should not be used to compare across teams. The issue is there are just more stable and better stats that give us a better understanding. For example xGF%. Jack Quinn is 7th on the team (out of 15 forwards) in xGF% at 49.44 which is still below 50% so he needs to be better but he is about even at getting and giving x goals. However, and I couldn't find the exact number because I forgot where to look, Quinn is -4 or 5 expected goals to date. His plus minus does reflect that but at 5 on 5 he is actually not a major negative. He just hasn't scored to go with his chances which we could argue is the only thing that matters. 

Further +/- does not account for the teammates you have. Everyone wants to crap on Quinn because he isn't scoring, which is fair btw. However a closer look tells me that while Quinn isn't doing great right now, Cozens is a ***** boat anchor tied to his neck. Quinn with Cozens has an xGF of 44.92% but you put Quinn with anyone else and it jumps up to 54.95xGF% and Cozens stays right at 44.53%.  They have a 2nd line center problem and Quinn isn't playing well enough this year to compensate. If I were Buffalo, I would be seriously considering whether or not we should keep Cozens. 

Basically +/- tells us a bit but there are more stats out there that correlate to good or bad hockey better. Quinn's on ice sh% is 4.32 which is worse than every single player not named Jiri Kulich. Because he isn't scoring goals, his line isn't scoring goals, and subsequently his +/- is awful.

Our best forwards are Tage, Tuch, and Benson. They drive quality chances in favor of the Sabres. Peterka and Zucker are still looking okay there too. All these stats are at 5v5 where the majority of +/- comes from. 

 

I just wish I could see what your computer is telling you. Quinn maybe taking shots from good spots but he is not taking good shots. You can blame Cozens for a lot of things this year but I would not put Quinn's poor play on him. Quinn has rarely found open Ice, he only has 43 shots on goal which is around 1.8 a game, with a 3.9 shot attempt per game average. I don't know where Quinn's shot has gone, he can be an effective sniper when right but It has to start with getting open enough to hit the net. He has been better at finding some spots of late but where is the real production?

[Jorcus]

One place where I really like to look at plus minus is players in the AHL. Most young prospects tend to start off very negative even if they are scoring a lot. You want to see who is coming up look at the improvement in plus minus stats and you will see who is playing defense well enough for a call up.    

[SabreFinn]

1 hour ago, LGR4GM said:

It tells us the +/- even strength goal differential and you are correct, it should not be used to compare across teams. The issue is there are just more stable and better stats that give us a better understanding. For example xGF%. Jack Quinn is 7th on the team (out of 15 forwards) in xGF% at 49.44 which is still below 50% so he needs to be better but he is about even at getting and giving x goals. However, and I couldn't find the exact number because I forgot where to look, Quinn is -4 or 5 expected goals to date. His plus minus does reflect that but at 5 on 5 he is actually not a major negative. He just hasn't scored to go with his chances which we could argue is the only thing that matters. 

Further +/- does not account for the teammates you have. Everyone wants to crap on Quinn because he isn't scoring, which is fair btw. However a closer look tells me that while Quinn isn't doing great right now, Cozens is a ***** boat anchor tied to his neck. Quinn with Cozens has an xGF of 44.92% but you put Quinn with anyone else and it jumps up to 54.95xGF% and Cozens stays right at 44.53%.  They have a 2nd line center problem and Quinn isn't playing well enough this year to compensate. If I were Buffalo, I would be seriously considering whether or not we should keep Cozens. 

Basically +/- tells us a bit but there are more stats out there that correlate to good or bad hockey better. Quinn's on ice sh% is 4.32 which is worse than every single player not named Jiri Kulich. Because he isn't scoring goals, his line isn't scoring goals, and subsequently his +/- is awful.

Our best forwards are Tage, Tuch, and Benson. They drive quality chances in favor of the Sabres. Peterka and Zucker are still looking okay there too. All these stats are at 5v5 where the majority of +/- comes from. 

I have had big hopes for Cozens, but I am more and more thinking he needs to be traded. But a big questions is where and what do we get back?

[mjd1001]

1 hour ago, Jorcus said:

 

I just wish I could see what your computer is telling you. Quinn maybe taking shots from good spots but he is not taking good shots. You can blame Cozens for a lot of things this year but I would not put Quinn's poor play on him. Quinn has rarely found open Ice, he only has 43 shots on goal which is around 1.8 a game, with a 3.9 shot attempt per game average. I don't know where Quinn's shot has gone, he can be an effective sniper when right but It has to start with getting open enough to hit the net. He has been better at finding some spots of late but where is the real production?

That is my problem with the xGF% and related stats.  It measures where the shots are taken from if I am correct, but are the 'good shots'?  I shot that you have time to set up for, maybe there is a screen in front of the net but is from a typical 'bad' area will be rated higher than a shot that is taken from a supposed better area but has no screen or is rushed.   Also, as others have said, it doesn't measure a particular players shooting percentage.  Tage is, has been, and will always likely be a much better shooter than Cozens, yet some of the advanced stats will not show that.

So it takes me back to plus-minus.  Yeah, there are flaws and exceptions with it. But on this forum goals expected and other advanced stats that are similar are used often as the end-all-to-be-all to judge a player, and they can be just as flawed as plus-minus.  At least plus-minus measures something that actually HAS happened, not something that could happen.

If there was only a single stat that somehow incorporated plus-minus, weight that plus-minus with relation to where you were on the ice, and merge that with expected goals for AND shooting percentage.

[LGR4GM]

+/- doesn't correlate well to good v bad play. It's more flawed than expected goals.

[mjd1001]

47 minutes ago, LGR4GM said:

+/- doesn't correlate well to good v bad play. It's more flawed than expected goals.

I guess we disagree on that one. When comparing players on the same team, I think expected goals is much worse.  They are different stats, they measure different things. But just as plus-minus can get something totally wrong, so can expected goals.  A few times over the past few years expected goals was used to say how bad of a game Bryam had or how good of a game Cozens had, and It was obvious to many of us on this board it was way off base.

Edited December 5 by mjd1001

[LGR4GM]

6 minutes ago, mjd1001 said:

I guess we disagree on that one. When comparing players on the same team, I think expected goals is much worse.  They are different stats, they measure different things. But just as plus-minus can get something totally wrong, so can expected goals.  A few times over the past few years expected goals was used to say how bad of a game Bryam had or how good of a game Cozens had, and It was obvious to many of us on this board it was way off base.

It's not. Statistically it simply isn't worse than +/- and that's a fact. 

It shouldn't be used for game by game comparison impo.

[mjd1001]

12 minutes ago, LGR4GM said:

It's not. Statistically it simply isn't worse than +/- and that's a fact. 

It shouldn't be used for game by game comparison impo.

It is not a fact for EVERY situation.  How can you say something is a fact, just because you want it to be?

I have personally found that when I evaluate a player on a team, when using plus-minus with a little 'common sense eye test', it works better than the expected goals that people throw out on this forum.

But hey, I guess the definition of what a 'fact' is has to run through you now?  Um..nope.

[EM88]

Not to break up the argument between the two of you but what I have read lately on here is that expected goals does not take into consideration shooting percentage? Is that correct?

Because I also read that one of the best indicators of a team's success that equates to wins and losses over the long term is shooting percentage for plus save percentage. Not even strength only, but overall.

If both of those things are true, then the expected goal stat would be very fundamentally flawed as it does not take into consideration one particular statistic that over the years is one of the best factors of wins and losses.

Edited December 5 by EM88

[PerreaultForever]

Stats are stats and they all provide information and they all have flaws in terms of providing a complete picture. 

When it comes to Quinn anyone can see he is performing poorly and has an overall negative impact on the team. If you can't see that you don't get hockey at all. 

Now you can argue whether Cozens makes Quinn worse or Quinn makes Cozens worse and so forth but those are nuances. Neither one of them is playing well and Cozens is most definitely not playing up to his contract which is perhaps the most significant statistic. The expected standard has to be measured against their cost in order to have a successful team. 

Benson, on the other hand, is obviously playing exceptionally well for his age and time here. You don't need any stats to see that either. 

In terms of economic value, Sabres worst performers are Cozens and Power. Best value is Benson. 

[ska-T Palmtown]

6 minutes ago, EM88 said:

Not to break up the argument between the two of you but what I have read lately on here is that expected goals does not take into consideration shooting percentage? Is that correct?

Because I also read that one of the best indicators of a team's success that equates to wins and losses over the long term is shooting percentage for minus shooting percentage allowed, meaning save percentage.

If both of those things are true, then the expected goal stat would be very fundamentally flawed as it does not take into consideration one particular statistic that over the years is one of the best factors of wins and losses.

I'll jump back into the fray and attempt to elaborate my understanding and personal opinions. (I was late for a meeting earlier and had to rush, lol)

Both +/- and any of the xGF% stats are a combination of what has happened while that player is on the ice, in total. So, if two players hop on and off the ice at the exact same time every shift, every game, their numbers would be identical - even if one has 60 goals and the other has 2. So, if a player has an xGF% of say, 45%, that means if you look at all the shots taken by both teams while that player is on the ice and assign them the NHL average shooting % (someone jump in here if it is from that location or overall - i think it is from that location?), it would say that Player X's team should score 45% of the goals that occur when they are on the ice. Which, if you are a shutdown line that regularly faces the other team's top line - that is pretty good. But a less than 50% xGF% tends to mean you give up better chances than you get.

Just personally, where I find that a little lacking for judging Player X, let's say they are McJesus and are just regularly burying goal after goal - but their linemates are a bunch of Cozens and Quinns - getting good shots, but not many, and certainly not burying many. Player X's xGF% might be pretty high, but their actual GF% may be low - so it was sorta my point that not too many stats are good on their own. I guess over a full season, you would expect xGF% and GF% to normalize to a team's abilities?

[JoeSchmoe]

1 hour ago, LGR4GM said:

+/- doesn't correlate well to good v bad play. It's more flawed than expected goals.

I think over a short sample size xGF% is better. 

Over a larger sample size where slumps and streaks work their way out, +/- or actual GF% is better.

Both need to take into account linemates, matchups, O vs D zone starts, etc.

[ska-T Palmtown]

8 minutes ago, PerreaultForever said:

Stats are stats and they all provide information and they all have flaws in terms of providing a complete picture. 

When it comes to Quinn anyone can see he is performing poorly and has an overall negative impact on the team. If you can't see that you don't get hockey at all. 

Now you can argue whether Cozens makes Quinn worse or Quinn makes Cozens worse and so forth but those are nuances. Neither one of them is playing well and Cozens is most definitely not playing up to his contract which is perhaps the most significant statistic. The expected standard has to be measured against their cost in order to have a successful team. 

Benson, on the other hand, is obviously playing exceptionally well for his age and time here. You don't need any stats to see that either. 

In terms of economic value, Sabres worst performers are Cozens and Power. Best value is Benson. 

I will first say that I am a HUGE Benson fan for all the reasons you said and what most people have ever praised him for. He is gonna be a "glue" guy on a really good team some day (Lord, let it be in Buffalo!!) 

Just to be contrarian, (why not?), on the score sheet I would not say Benson has contributed significantly. 3G and 4A and a +/-0  in 19 games gets most other players a pretty "meh" reaction (obviously ignoring his age and ELC). Personally, I feel his impact is harder to measure since he is, at this stage in his promising young career, an "effort" guy. I think it is just to say that I think we'd cut Quinn the same slack if he were the tenacious forechecker that was lifting sticks and swiping pucks but maybe not lighting up the scoreboard. But as you said at $7M ... Cozens needs to either win the Selke or be at least responsible in is own end and still put up 60+ points. Maybe he gets to be Nylander/Skinner (allergic to his own zone) if puts up 90 points?

[LGR4GM]

31 minutes ago, mjd1001 said:

It is not a fact for EVERY situation.  How can you say something is a fact, just because you want it to be?

I have personally found that when I evaluate a player on a team, when using plus-minus with a little 'common sense eye test', it works better than the expected goals that people throw out on this forum.

But hey, I guess the definition of what a 'fact' is has to run through you now?  Um..nope.

This encapsulates most of the internet these days. You "feel" something else is better because you understand it better. Granted,  there's literally statistical testing that nhl teams use for this. There's all the public data that demonstrates xgf correlates better to on ice performance than +- but it's 2024, where everything's made up and the facts don't matter. That's right the facts are like a diet on Christmas. 

So "umm nope" your misinformation back to ignorance mountain. It's a statistical fact xgf is a better indicator of on ice performance when compared to +- whether your feelings are different or not. It has nothing to do with me at all. 

Here's a great xgf talk and it mentions the weaknesses. You should read it or take some classes on stats and probability https://insidetherink.com/inside-the-stats-expected-goals/

 

8 minutes ago, JoeSchmoe said:

I think over a short sample size xGF% is better. 

Over a larger sample size where slumps and streaks work their way out, +/- or actual GF% is better.

Both need to take into account linemates, matchups, O vs D zone starts, etc.

It's not. 

Edited December 5 by LGR4GM

[LGR4GM]

You know what. ***** it. 

[ska-T Palmtown]

3 minutes ago, LGR4GM said:

You know what. ***** it. 

haha - wait what? This is one of most interesting discussions in a while! Certainly better than all the doom and gloom Sabres threads 🙂

That article was sweet - being an engineer, I would be in the "stat nerd" part of one of those 4-way graphs, for sure. I like knowing the nuance of things - and I find outliers (like how in 21-22 Matthews obliterated expectations) to be fascinating. It is also fun to find negative outliers to try to pin point where things go wrong. Player? Model? A little bit of both? Good stuff!

[PerreaultForever]

4 hours ago, ska-T Chitown said:

I will first say that I am a HUGE Benson fan for all the reasons you said and what most people have ever praised him for. He is gonna be a "glue" guy on a really good team some day (Lord, let it be in Buffalo!!) 

Just to be contrarian, (why not?), on the score sheet I would not say Benson has contributed significantly. 3G and 4A and a +/-0  in 19 games gets most other players a pretty "meh" reaction (obviously ignoring his age and ELC). Personally, I feel his impact is harder to measure since he is, at this stage in his promising young career, an "effort" guy. I think it is just to say that I think we'd cut Quinn the same slack if he were the tenacious forechecker that was lifting sticks and swiping pucks but maybe not lighting up the scoreboard. But as you said at $7M ... Cozens needs to either win the Selke or be at least responsible in is own end and still put up 60+ points. Maybe he gets to be Nylander/Skinner (allergic to his own zone) if puts up 90 points?

Keep in mind Benson goes to the net and makes things happen whether he gets on the scoreboard or not. If they all played like that we'd be in first place.