Plus / Minus -- the Stat that Inherently Punishes Skilled Forwards

[The Panther]

Okay, so I know this isn't news to anyone on the History Forum, but Plus/Minus is wack!!

It seems to me that it is time to garbage-bin this near-useless statistic in favor of an advanced-stat that takes more things into account.

Anyway, getting back to standard plus/minus. Why do I say it's wack? Well, everyone knows that the stat is mainly dependent on team-success. Also, this stat has the potential to punish good players with bad team/line-mates and reward average players with good team/linemates.

We all know those facts. But above all, I think Plus/Minus is a stat that is inherently nonsensical, especially in regard to high-scoring forwards (and the lesser number of high-scoring defencemen). Why? Because Plus/Minus inherently handicaps those high-scoring players. It just makes no sense.

This is what I can't get my head around: If a team, say, trades for a high-scoring Power-play specialist -- the Dave Andreychuk / Tim Kerr type, if you will -- and that new player successfully achieves exactly what the team asked him to do, Plus/Minus will effectively punish him for it. This is because his main contribution to the team won't be factored in. He'll earn no 'plus' for doing his job well and scoring.

While this handicap is moderate for 2nd and 3rd-line type of players, it's going to massively handicap most top-3 forwards on teams, and also some high-scoring defencemen.

While I concur that it's not right to 'minus' players who are on-ice for Power-play goals against, it seems totally wrong to not 'plus' players who score on Power-plays. At the very least, I think the NHL should consider altering the stat so that Power-play goals for count for perhaps 1/2 a normal score, or something like that. Let's give Tim Kerr his due!

The other failed aspect of this stat is that it punishes the team's best defensive players -- the 'stay-at-home defencemen', if you will. Since those players aren't concerned with scoring and have the most difficult match-ups night after night, it's quite expected that on average teams (and even some good teams) the shut-down D-men will have more goals scored against them than for. Obviously, plus/minus cannot take that into account, but it somehow seems to me that the stat is almost completely useless in any comparison of forwards to defencemen.

Now, for some examples. Let's look at 4 historical teams, and their top-10 plus/minus players (who played 50+ games). The table shows the players (ranked by official plus/minus), then the official plus/minus itself, then the Difference of TOTAL goals scored for minus TOTAL goals scored against ("raw" plus/minus), and then finally the plus/minus difference between actual and official:


Boston Bruins 1970-71

Rank | Player | Official +/- | "Raw" +/- | Difference
1 | B. Orr| +124| +173|49 more
2 | D. Smith| +94| +70|24 less
3 | P. Esposito| +71 |+142|71 more
4 | K. Hodge |+71| +89|18 more
5 | W. Cashman|+59| +78|19 more
6 | Ed Westfall |+58 |+24|34 less
7 | D. Awrey |+40| +29|11 less
8 | D. Sanderson|+39 |+13|26 less
9 | T. Green| +37| +26|11 less
10| J. Bucyk| +36| +106|70 more

Montreal Canadiens 1976-77

Rank | Player | Official +/- | "Raw" +/- | Difference
1|L. Robinson |+120|+144| 24 more
2 |G. Lafleur | +89| +136|47 more
3 |S. Shutt | +88| +120|32 more
4 |S. Savard | +79| +75|4 less
5| J. Lemaire |+70| +90|20 more
6| G. Lapointe | +69| +86|17 more
7| B. Nyrop | +42| +39|3 less
8| R. Houle | +39| +46|7 more
9| P. Mahovlich | +36| +59|23 more
10| P. Bouchard | +33| +30|3 less

Edmonton Oilers 1984-85

Rank | Player | Official +/- | "Raw" +/- | Difference
1 | W. Gretzky| +98| +122|24 more
2 | J. Kurri| +76| +93|17 more
3 | M. Krushelnyski|+56| +87|31 more
4 | P. Coffey |+55| +91|36 more
5 | C. Huddy |+50| +66|16 more
6 | R. Gregg |+27| +24|3 less
7 | D. Jackson |+27| +17|10 less
8 | G. Anderson |+24| +58|34 more
9 | L. Fogolin |+16| -13|29 less
10 | K. Lowe |+9| -13|22 less

Pittsburgh Penguins 1992-93

Rank | Player | Official +/- | "Raw" +/- | Difference
1 | M. Lemieux|+55 |+114|59 more
2 | L. Murphy |+45 |+88|43 more
3 | U. Samuelsson|+36|+16|20 less
4 | J. Jagr| +30| +70|40 more
5 | R. Tocchet |+28| +82|54 more
6 | K. Samuelsson|+25| -3|28 less
7 | S. McEachern|+21| +28|7 more
8| J. Mullen| +19| +33|14 more
9| K. Stevens| +17| +86|69 more
10| J. Paek |+13 |+1|12 less

Now, of course the "raw" plus/minus I'm showing (and calculating differences with) is itself kind of wacky, because it is counting power-play goals scored against, which, as I stated, I don't think Plus/Minus should do. But still, even with that major flaw, I believe the 'raw' stats give us a much more accurate picture of the player-value of forwards in particular.

Regarding Boston, the "raw" plus/minus really shows the dependence of Dallas Smith's plus-rating on Bobby Orr. In official stats, Orr is only 30+ higher than Smith, but in "raw" stats, Orr is 103+ higher than Smith. Ed Westfall and Phil Esposito provide good examples of how misleading Plus/Minus can be in overall on-ice value. By looking at the official plus/minus, Westfall is merely 13+ lower than Esposito, despite Espo's scoring almost 100 more points than Westfall. That just seems weird, and might make you think Esposito wasn't doing enough on the back-end. However, by 'raw' plus/minus, we see that Esposito was actually +142 and Westfall +24. Now the difference in relative player-value is clear. However, the limitation is clear, too -- Derek Sanderson comes out a measly +13 in raw plus/minus, playing for one of the most dominant teams ever. The reason is likely to do with his assignments as a checking forward against other teams' top players, something that Plus/Minus in any form just can't bear out.

With Montreal, the relative value of Lafleur and Robinson becomes closer, as Robinson's raw plus/minus is only slightly higher than Lafleur's. The Plus/Minus problem of linemates remains, however, as I don't know if Steve Shutt's raw plus/minus deserves to be as close to Guy's and Larry's as it is.

An interesting point of Edmonton's is that this team appears to be less dependent on the power-play for its high-octane offense than the other three teams studied (this is the year Gretzky scored 73 goals, and only 8 were on the power-play!). I say this because the difference between official and 'raw' plus/minuses are less than the other teams here. Here we see perfect examples of the defensive defencemen getting hammered down, however, by raw Plus/Minus. (I'm sure that wouldn't be the case, however, if we could erase the power-play goals scored against from their stats.)

And Pittsburgh's 'raw' plus/minuses show more clearly the relative player value of high-scoring Murphy, Tocchet, and Stevens. Lemieux's dominance, often on the power-play, also goes into sharper focus.



So, what about Tim Kerr, since I mentioned him off the top? If you can believe it, Kerr went into the books as a -5 with Philly in 1985-86 (his teammate, Mark Howe, was +85). You might ask, how could any player, on a great team, score 58 goals and end up -5? The answer, obviously, is Kerr's power-play 'dependence'. In 'raw' plus/minus, Kerr was a solid +73, and this would more accurately show his relative player value. (In Kerr's case, since he didn't kill penalties, this +73 doesn't change at all when power-play goals against are discounted.)

What of other great players' stats if the power-play goals FOR are counted, and the power-play goals against are NOT?:

Player | Season | "Raw" +/- not counting PPG against
B. Orr |1970-71|+203
Phil Esposito|1970-71| +149
Guy Lafleur|1976-77|+136 (he was never on-ice for a PP goal against)
Larry Robinson|1976-77|+154
Wayne Gretzky|1984-85|+159
Jari Kurri|1984-85|+114
Mario Lemieux|1992-93|+137

Hmm... but those stats seem a little extreme, and maybe too favorable to the power-play/high scorers.

What do you think of my suggestion of altering it so that a power-play goal counted for 0.5 of a regular goal in plus/minus? I think that might be a fairer way of using this stat for forwards than the way it is now. I mean, something is wrong if a 58-goal scorer on a great team is a raw +73 and ends up -5 officially.

Thoughts?

 

[MXD]

Well, the said Kerr/Andreychuk player doesn't get a "+"... but he gets a point instead.
+/- does a very good job at what it purports to measure.

It's when people are using the stat to measure something it doesn't measure that it becomes a bit useless.

 

[blogofmike]

Plus minus doesn't account for a player's PP performance.

This leaves out contributions of great PP players, but I wouldn't say that the new number shows that Dallas Smith was dependent on Orr for his Plus/Minus (he was, but the number doesn't prove it.) The number only shows that Boston had 5 guys who were on the PP ALL the time. Dallas was not one of them, as Stanfield took the point and Esposito took Stanfield's linemates. If Smith were on the PP, the new number would be just as Orr-skewed.

The way to separate Smith from Orr would be to differentiate between points and goals on-ice for. Something like ESPOINTS - (0.52 * ESGA) so that the value of a Plus (goal you scored or assisted on) produced a plus at a comparable rate to a minus (ES goal against) though the multiplier would change with assists per goal in that year.

 

[tony d]

Meh, plus/minus is a useless stat, really shouldn't be used to measure how good a player is.

 

[billybudd]

I'm not grasping where skilled forwards are getting "punished," exactly. Lack of credit for something someone may or may not deserve credit for is an absence of praise, not a "punishment."

The issue with your way is that if you're counting powerplay plus minus, you also need to be counting PK plus minus. If a guy like, idk, Jay Pandolfo would end every season as a big minus because he played so many short-handed minutes and few to no powerplay minutes, that would be a punishment.

Plus minus isn't meant to measure overall contributions (no one number can), but to measure differentials when all else is equal. It doesn't do a wonderful job of that due to things like variance, differing usage, strength of opposition, team factors (etc), but as a general rule of thumb, it works well enough at the end points as long as you don't take it too seriously. There's a good chance the worst plus minus player on the worst plus minus team is among the worst two-way players in the sport (and vice versa).

I mean, that Ulf Samuelsson would have better differentials than Kevin "I'll play offense, everything else is Tommy B's problem" Stevens isn't exactly a surprise to me. Ditto Kurri/Anderson.

 

[authentic]

I'm not grasping where skilled forwards are getting "punished," exactly. Lack of credit for something someone may or may not deserve credit for is an absence of praise, not a "punishment."

The issue with your way is that if you're counting powerplay plus minus, you also need to be counting PK plus minus. If a guy like, idk, Jay Pandolfo would end every season as a big minus because he played so many short-handed minutes and few to no powerplay minutes, that would be a punishment.

Plus minus isn't meant to measure overall contributions (no one number can), but to measure differentials when all else is equal. It doesn't do a wonderful job of that due to things like variance, differing usage, strength of opposition, team factors (etc), but as a general rule of thumb, it works well enough at the end points as long as you don't take it too seriously. There's a good chance the worst plus minus player on the worst plus minus team is among the worst two-way players in the sport (and vice versa).

I mean, that Ulf Samuelsson would have better differentials than Kevin "I'll play offense, everything else is Tommy B's problem" Stevens isn't exactly a surprise to me. Ditto Kurri/Anderson.

Yeah this is all basically how I feel about +/- as well. There are far better stats to look at like adjusted +/-, many different advanced stats, etc. It's no coincidence however that a lot of the best two way or overall players usually fair well in the +/- category.

No one is going to make the argument that Kunitz is a more effective even strength player than Crosby this season. At the same time it makes sense that a guy like Kopitar is 2nd overall in the league, to Toffoli whom he plays with quite often by a slim margin, while also being the go to forward on his team offensively and defensively while taking on the toughest assignments in terms of line matching.

 

[Dennis Bonvie]

So, plus/minus inherently punishes skilled forwards.

Or, plus/minus inherently punishes one dimensional players?

I'd say the latter. And I'm pretty sure that's what it was designed to do.

 

[Black Gold Extractor]

Plus-minus would improved by using it in only an even-strength context (or we should keep track of a new measure, "EV +/-", or something like that). It's not perfect, but it's the only stat that has tried to measure play on both ends of the rink going back to 67-68.

"Raw +/-" overly benefits players who play on the powerplay. On average, it's roughly 2.5 times easier to score per minute on the powerplay than at evens. Furthermore, most skilled forwards play only on powerplays but not the penalty kill. A defensive forward defending a 5-on-3 gets a minus when a goal goes in? A skilled forward getting a plus while playing on a 5-on-3 man advantage? Sorry, but that's a huge imbalance.

We already reward players who score on the powerplay (or at even strength, or shorthanded) with points.

 

[The Panther]

Surprised at some of these responses, so far.

Don't get me wrong, I find plus/minus an interesting stat in its various forms (hence this thread), but I just don't think the NHL should be using it as an 'official' stat in its current method of determination.

I strongly disagree with the suggestion (above) that plus/minus "inherently punishes one dimensional forwards". There is actually nothing 'punishable' about being a one-dimensional forward. Plenty of Stanley Cup teams are filled with one-dimensional role players who play their role to perfection. My problem with current plus/minus is that it concludes that a guy with 58 goals on a great team who is overall +73 is "officially" minus 5. The same goes for Mario in 1985-86, as one example of a high-skilled forward on a so-so team: He is in reality +73 but officially minus 5. Lemieux was obviously the MVP of his team, but ends up with the 5th or 6th worst plus/minus on it. Does no one agree with me that that renders this stat meaningless?

I guess my question would be: Since the MVP of NHL teams is ending up a minus, what exactly is plus/minus measuring??

 

[Black Gold Extractor]

Surprised at some of these responses, so far.

Don't get me wrong, I find plus/minus an interesting stat in its various forms (hence this thread), but I just don't think the NHL should be using it as an 'official' stat in its current method of determination.

I strongly disagree with the suggestion (above) that plus/minus "inherently punishes one dimensional forwards". There is actually nothing 'punishable' about being a one-dimensional forward. Plenty of Stanley Cup teams are filled with one-dimensional role players who play their role to perfection. My problem with current plus/minus is that it concludes that a guy with 58 goals on a great team who is overall +73 is "officially" minus 5. The same goes for Mario in 1985-86, as one example of a high-skilled forward on a so-so team: He is in reality +73 but officially minus 5. Lemieux was obviously the MVP of his team, but ends up with the 5th or 6th worst plus/minus on it. Does no one agree with me that that renders this stat meaningless?

I guess my question would be: Since the MVP of NHL teams is ending up a minus, what exactly is plus/minus measuring??

Since Lemieux doesn't seem to have played PK in 85-86, I'd say that plus-minus is what it says it is: when not given a man-advantage, more goals were scored against than for his team. The Penguins scored 208 EV goals and gave up 210 EV goals. The Penguins also gave up 17 SH goals, so the Penguins were 208 GF/227 GA without a man-advantage.

Because Hockey-Reference is silly and doesn't list SH goals for and against, I'll estimate Lemieux as being 9 SH goals against when on ice.

This also means that at even-strength, Lemieux probably had an "EV +/-" of +3 or so. Basically, facing the toughest competition, Lemieux was more or less a slight net benefit in his sophomore year on a team that was at a slight minus at even strength.

Lemieux isn't really +73, because that's an unfair rating of his net benefit with respect to his offense versus his defense. That year, he didn't have any SH time. He was on the 1st PP line.

I know what you're thinking: the Penguins had a net positive goal differential thanks to their powerplay. I agree. Special teams do matter, and Lemieux is one of the best ever at it.

The bigger issue is that there is absolutely no real measure of defensive play except for plus-minus; however, with plus-minus, we can kind of work backwards to guess at a player's defensive prowess. Yes, the rest of the team also plays a role in a player's plus-minus, but we can always measure a player's plus-minus relative to either the team average or the player's likely linemates if we can somehow do that (a "Rel +/-").

So if a player who is 2nd in NHL scoring is only slightly positive on a team that's a slight negative at even-strength, then we can guess that his defensive game leaves something to be desired.

That's the (admittedly flawed and limited) value of plus-minus. It lights a candle in a dark windowless room with regards to defensive/two-way play.

EDIT: And let's be honest: it's still better than having to resort to Selke votes and newspaper clippings before 67-68.

 

[reckoning]

The same goes for Mario in 1985-86, as one example of a high-skilled forward on a so-so team: He is in reality +73 but officially minus 5. Lemieux was obviously the MVP of his team, but ends up with the 5th or 6th worst plus/minus on it. Does no one agree with me that that renders this stat meaningless?

Let me use an analogy: In 1983-84 Rod Langway won the Norris for best defenceman, and finished second in the Hart voting to Gretzky. He was obviously Washington's MVP, but he finished tied for 210th overall in points in the league. Should that render points a meaningless statistic?

A stat measures only what it measures, nothing else. In the case of +/-, it's the difference between even-strength goals for and against that were scored when a player was on the ice. That's all it says. Whether one thinks that's useful or useless information is up to them, but the statistic doesn't claim anything more than it is.

Your suggestion of adding powerplay goals would make a flawed stat even worse. It's almost a certainty that each teams best defensive forward would have one of their worst ratings (if you kill penalties, but never play on the powerplay, your mark can only go down, not up, no matter how good you are).

In the case of Tim Kerr, his even-strength production fell off significantly in 85-86. He was outpointed by many Flyers at ES (including McCrimmon of all people). Does that make even-strength points an unfair statistic?

 

[Chili]

I made my points on this topic here.

Some food for thought...

Tom Preissing finished second one year in plus minus...he was the Sens sixth defenceman.

Edit/Correction: Preissing finished tied for third in 2006/07 with Nik Lidstrom (plus 40) behind Thomas Vanek (+47) and Daniel Alfredsson (+42).

 

[Nalens Oga]

I made my points on this topic here.

Some food for thought...

Tom Preissing finished second one year in plus minus...he was the Sens sixth defenceman.

So does that suggest that coaches over-shielded ****** players to the detriment of their better players and should have been shielding the better players more?

 

[Chili]

So does that suggest that coaches over-shielded ****** players to the detriment of their better players and should have been shielding the better players more?

I don't even try to interpret the number.

I always find it interesting when I hear someone needs to improve their plus minus.

The best way to do that is to get traded to a better team.

 

[Canadiens1958]

Plus / Minus

Plus/minus is a holdover stat from the early days of the NHL when there was hardly any substitution. Starters played most of not the complete game. Easy to calculate almost a by default stat.

Once rosters expanded to include multiple lines and defensive pairings, teams would track and interpret their own numbers.

The NHL, under Ron Andrews started releasing generalized +/- data with the 1967 expansion. These stats had little relation to how teams interpreted their own data. But it kept the fans interested and helped the NHL teams in contract negotiations.

Today it is a nostalgia type stat that allows overviews covering a period of close to fifty seasons. Nothing more, nothing less.

 

[The Panther]

Basically, facing the toughest competition, Lemieux was more or less a slight net benefit in his sophomore year on a team that was at a slight minus at even strength.

Lemieux isn't really +73, because that's an unfair rating of his net benefit with respect to his offense versus his defense. That year, he didn't have any SH time. He was on the 1st PP line.

I know what you're thinking: the Penguins had a net positive goal differential thanks to their powerplay. I agree. Special teams do matter, and Lemieux is one of the best ever at it.

Right, but it's not only that -- it's also the fact that Lemieux, as the #1 P.P. guy on a so-so team, probably was on the ice for 95% of the Pens' power-play time during 1985-86. The Pens in 1985-86 had 425 power-play opportunities, so let's guess'timate they spent 850 minutes, give or take, with the man advantage. That means Mario Lemieux maybe spent the equivalent of 12 full games (or, if you like, 24 games with 30 minutes' ice-time per night), on the power-play. Is it right that for those 24 games' worth of ice-time, he is ineligible to get a 'plus' rating? For 25% of his ice-time on the season -- when he doing his job most successfully -- he is ineligible.

with plus-minus, we can kind of work backwards to guess at a player's defensive prowess.

I don't think we can do that unless we know ice-time, player style, and other things. But I guess my question is: in the era of advanced stats, why should we (fans) have to work backwards to arrive at a number? Surely there's a more equitable and mathematically sound way to arrive at a number that is more meaningful than 'plus/minus'?

So if a player who is 2nd in NHL scoring is only slightly positive on a team that's a slight negative at even-strength, then we can guess that his defensive game leaves something to be desired.

Can we? I'm not so sure. When Gretzky was 19 he was +41 on a losing team. Can we guess that Gretzky was defensively sound?

I know you're aware of these flaws I point out, but I just think there could very easily be a much more useful stat to make 'official' and to replace plus/minus with.

Let me use an analogy: In 1983-84 Rod Langway won the Norris for best defenceman, and finished second in the Hart voting to Gretzky. He was obviously Washington's MVP, but he finished tied for 210th overall in points in the league. Should that render points a meaningless statistic?

Well, points are points. Nobody looks at raw points and evaluates players. But many, many people do look at plus/minus and draw conclusions about players' overall value to their team, and I say this is bogus.

A stat measures only what it measures, nothing else. In the case of +/-, it's the difference between even-strength goals for and against that were scored when a player was on the ice.

I get what you're saying, but if a stat is just a stat, why do we 'officially' not count power-play goals against?

Your suggestion of adding powerplay goals would make a flawed stat even worse.

Well, I'm just throwing the idea around of (for example) counting a power-play goal as a 1/2 +, instead of not counting it at all. There are various other ways to improve this stat, I think. My suggestion would be no more complicated than the current system, which, as I say, is already 'tweaked' in that it's not counting power-play goals against.

For example, in 1985-86, when Kerr finished -5, he was on the ice for 78 (!) power-play goals scored by Philly, none of which count in his plus/minus. By my tentatively suggested system, he would get 39 'plus' points for those goals, and so his -5 would have instead been +34. I think peak Tim Kerr, scoring 58 goals for a strong team, deserves roughly a +34 more than he deserves a -5. (Again, I'm sure there are better ways to calculate this than my suggestion, but I'm just throwing out one idea.)

It's almost a certainty that each teams best defensive forward would have one of their worst ratings (if you kill penalties, but never play on the powerplay, your mark can only go down, not up, no matter how good you are).

Perhaps you misunderstood me. I wasn't suggesting to count power-play goals against. I was only suggesting to count power-play goals for (and then maybe as a 1/2).

Plus/minus is a holdover stat from the early days of the NHL when there was hardly any substitution. Starters played most of not the complete game. Easy to calculate almost a by default stat.

Once rosters expanded to include multiple lines and defensive pairings, teams would track and interpret their own numbers.

The NHL, under Ron Andrews started releasing generalized +/- data with the 1967 expansion. These stats had little relation to how teams interpreted their own data. But it kept the fans interested and helped the NHL teams in contract negotiations.

Today it is a nostalgia type stat that allows overviews covering a period of close to fifty seasons. Nothing more, nothing less.

Very interesting, thanks.

 

[billybudd]

Surprised at some of these responses, so far.

Don't get me wrong, I find plus/minus an interesting stat in its various forms (hence this thread), but I just don't think the NHL should be using it as an 'official' stat in its current method of determination.

I strongly disagree with the suggestion (above) that plus/minus "inherently punishes one dimensional forwards". There is actually nothing 'punishable' about being a one-dimensional forward. Plenty of Stanley Cup teams are filled with one-dimensional role players who play their role to perfection. My problem with current plus/minus is that it concludes that a guy with 58 goals on a great team who is overall +73 is "officially" minus 5. The same goes for Mario in 1985-86, as one example of a high-skilled forward on a so-so team: He is in reality +73 but officially minus 5. Lemieux was obviously the MVP of his team, but ends up with the 5th or 6th worst plus/minus on it. Does no one agree with me that that renders this stat meaningless?

I guess my question would be: Since the MVP of NHL teams is ending up a minus, what exactly is plus/minus measuring??

Yeah, but we already have measures for "one dimensional" offensive play. Goals and assists. I'm not certain why we'd need a another way to measure that.

Ostensibly plus/minuses purpose is to measure the net value added by a players offensive/defensive play. It's a correlation stat (though with a weak correlation due to the luck involved in low frequency events).

Let's put it this way. Suppose the Oilers' 3rd and 4th line centers both finish a season with 8 goals each. Guy A is minus 10. Guy B is minus 40. What plus minus tells you is that guy B is probably (not definitely, probably) a greater net negative than guy A due to how much more often the puck ends up in the net. There may be mitigating factors that can explain this away (usage, teammates, zone starts), but much more often than not, what's going on is what it looks like is going on.

 

[MXD]

This said, Ceteris Paribus, a player whose +/- is disproportionnaly high or low vs. rest of the team is probably deployed in a way that isn't completely appropriate.

Unless we're talking of an absolutely elite player.

 

[Black Gold Extractor]

Right, but it's not only that -- it's also the fact that Lemieux, as the #1 P.P. guy on a so-so team, probably was on the ice for 95% of the Pens' power-play time during 1985-86. The Pens in 1985-86 had 425 power-play opportunities, so let's guess'timate they spent 850 minutes, give or take, with the man advantage. That means Mario Lemieux maybe spent the equivalent of 12 full games (or, if you like, 24 games with 30 minutes' ice-time per night), on the power-play. Is it right that for those 24 games' worth of ice-time, he is ineligible to get a 'plus' rating? For 25% of his ice-time on the season -- when he doing his job most successfully -- he is ineligible.

I think the conception of plus-minus was to measure the balance between offense and defense (i.e. see whether a player was cheating for offense). On the powerplay, an attacking player is ~2.5x more likely to score per minute than at evens while defending players are ~2.5x less likely to score. If you're on the powerplay, you shouldn't be the one defending.

I can see why they decided to give shorthanded scorers a +1, but ideally they should have kept it purely an even-strength measure.

I don't think we can do that unless we know ice-time, player style, and other things. But I guess my question is: in the era of advanced stats, why should we (fans) have to work backwards to arrive at a number? Surely there's a more equitable and mathematically sound way to arrive at a number that is more meaningful than 'plus/minus'?

Sure. I'd think that an adequate measure for defense (for starters) is to look at unblocked shot attempts against per 60. Then, we can look at zone starts and such. Even now, though, advanced stats are kind of in a state of flux (like a new advanced stat, "expected goals", that was tracked starting one season ago).

Unfortunately, advanced stats go back to only 2007-08. The only reason why plus-minus is still a thing is for legacy reasons, like Canadiens1958 alluded to. It gives us a bridge to older times, back until 67-68. (Sadly, it's still not that far back if you think about it. It's only about 5 "hockey generations".)

Can we? I'm not so sure. When Gretzky was 19 he was +41 on a losing team. Can we guess that Gretzky was defensively sound?

Gretzky was the highest plus on the 80-81 Oilers, with Kurri being second with +26... but Gretzky scored 105 EV points and 7 SH points. Kurri scored 67 EV points and 0 SH points. You could probably guess that Kurri was more defensively sound than Gretzky.

Gretzky was honestly more "offense is the best defense". It's a stupid idea for the modern-day Oilers, but when it's prime Gretzky...

I know you're aware of these flaws I point out, but I just think there could very easily be a much more useful stat to make 'official' and to replace plus/minus with.

Unblocked shot attempts for and against are probably an indicator for long-term success. There's not much for old teams, but just in case you've never seen these:

Even-strength on-ice shot attempt differentials (Corsi) for:

1984 Stanley Cup Finals Game 5

1987 Stanley Cup Finals Game 7

It's fascinating to look at the vastly different results for these two games. One shows us a future dynasty team having to try to earn its place against a battle-hardened dynasty with 19 straight playoff series wins. Another shows us a two-time Stanley Cup champion with something to prove against another great (but ultimately underpowered) team.

In 84, Gretzky was the king of generating offense. As such, he had the most chances for (on-ice 13 shot attempts for), and the best shot attempt differential when he was on ice (+8, tied with linemate... Semenko). As for defense, Gretzky's 5 on-ice shot attempts against is surprisingly the second best on the team (behind only linemate... Semenko with 2). Messier's line, seeing the tough assignments with tougher zone starts, is basically mauled. Messier himself had an on-ice -15 shot attempt differential, worst on the team. The Islanders, at evens, outplayed the Oilers (32 vs 45 attempts) but lost, because everyone's time eventually has to end.

In 87, how things have changed: the best line was Messier's line (though that's probably thanks to Gretzky's concussion). Messier himself had the best on-ice shot attempt differential (+21!) followed by linemate Anderson (+20!). The Oilers outplayed the Flyers, with an even-strength shot attempt differential of 65-25. I kind of get why Hextall got the Conn Smythe.

 

[Sens Rule]

So does that suggest that coaches over-shielded ****** players to the detriment of their better players and should have been shielding the better players more?

With Priessing he was on a totally stacked team. They had 3 very good lines and a good 4th line. He only played 15:15 minutes per game. Almost all of those against weaker lines with unusually good forwards with him and an excellent partner offensively in Corvo.

You have Philips / Volchekov as your top pair. Often on with the Pizza line. Or with the best defensive line.

Redden / Meszaros as your "2nd" pairing... Used in both key offensive and defensive roles.

And the third pair of Corvo and Priessing that got to play very sheltered minutes against the least good lines of other teams... While occasionally having the Pizza line with them but also pretty high end guys like Kelly, Schaefer, Fisher, Vermette etc on with them.

The Senators had by far the best line in the league. And played them a ton versus anyone. Usually the other teams best scoring line and best checking line. Then you have them rolling their next most effective line versus the Fisher Schaeffet line. So then you end up with your worst line versus 2 decent offensive D in Corvo and Priessing and Comrie, Kelly, Vermette, Eaves... Etc.

Plus Schubert was a constant 7th D 12th forward and played defensively on D when needed over Priessing during games.

So Priessing basically always got to play versus the weakest lines of other teams with mostly players capable of being good second liners or tweeners on the ice with him.

Also in looking up the stats on Hockey Reference I see Priessing got 4.9 DPS and 8.2 overall PS for that year on hockey reference. That shows the uselessness of their point shares.

Priessing is supposedly the 3rd most valuable Senator defensively when he literally was like the 17th or 18th most valuable defensively... Maybe lower then that.

Priessing is supposedly the 5th most valuable Senator overall that year. More valuable then Phillips or Redden or Fisher. Another absurdity.

Even more absurd.... Schubert is the 7th most valuable Senator that season. He played 11:23 minutes per game and was literally both the 12th forward and 7th D all year long.

He was by definition the least valuable regular skater on the team. But apparently he was more valuable then Volchenkov, Redden, Fisher, Vermette, Kelly etc.

So this long post shows the best evidence about how the point shares are driven so much by a useless stat like plus/minus and that Hockey Reference point shares should never, ever be even considered to have any utility in any way. Love the site... Hate their Point shares. They come from a completely useless formula and thus distort reality and do not measure at all what they are supposed to.

 

[Morgoth Bauglir]

With Priessing he was on a totally stacked team. They had 3 very good lines and a good 4th line. He only played 15:15 minutes per game. Almost all of those against weaker lines with unusually good forwards with him and an excellent partner offensively in Corvo.

You have Philips / Volchekov as your top pair. Often on with the Pizza line. Or with the best defensive line.

Redden / Meszaros as your "2nd" pairing... Used in both key offensive and defensive roles.

And the third pair of Corvo and Priessing that got to play very sheltered minutes against the least good lines of other teams... While occasionally having the Pizza line with them but also pretty high end guys like Kelly, Schaefer, Fisher, Vermette etc on with them.

The Senators had by far the best line in the league. And played them a ton versus anyone. Usually the other teams best scoring line and best checking line. Then you have them rolling their next most effective line versus the Fisher Schaeffet line. So then you end up with your worst line versus 2 decent offensive D in Corvo and Priessing and Comrie, Kelly, Vermette, Eaves... Etc.

Plus Schubert was a constant 7th D 12th forward and played defensively on D when needed over Priessing during games.

So Priessing basically always got to play versus the weakest lines of other teams with mostly players capable of being good second liners or tweeners on the ice with him.

Also in looking up the stats on Hockey Reference I see Priessing got 4.9 DPS and 8.2 overall PS for that year on hockey reference. That shows the uselessness of their point shares.

Priessing is supposedly the 3rd most valuable Senator defensively when he literally was like the 17th or 18th most valuable defensively... Maybe lower then that.

Priessing is supposedly the 5th most valuable Senator overall that year. More valuable then Phillips or Redden or Fisher. Another absurdity.

Even more absurd.... Schubert is the 7th most valuable Senator that season. He played 11:23 minutes per game and was literally both the 12th forward and 7th D all year long.

He was by definition the least valuable regular skater on the team. But apparently he was more valuable then Volchenkov, Redden, Fisher, Vermette, Kelly etc.

So this long post shows the best evidence about how the point shares are driven so much by a useless stat like plus/minus and that Hockey Reference point shares should never, ever be even considered to have any utility in any way. Love the site... Hate their Point shares. They come from a completely useless formula and thus distort reality and do not measure at all what they are supposed to.

Check out Jeff Schultz (a pylon if there ever was one) in 2009-10 for the best illustration of absurdity.

 

[MS]

+/- is a measure of how effective a player is in his role 5-on-5, relative to his team.

So yes, occasionally a Priessing or Schultz who did a good job on the 3rd pairing of a good team has an excellent +/-. Doesn't mean he's a great player, just had a good season in his role at ES.

Seems like the people who complain about the stat expect it to be an all-encompassing measurement of 'good player' or 'good defensive player' and then get all offended when they find out it isn't. It's a somewhat useful stat for picking out outliers who are struggling/excelling in their role.

 

[Chili]

Nice work Sens Rule.

Interesting article here on plus minus, based on a five year study of Oilers goals for/against.

Check this part out ...

Number of goals against the Edmonton Oilers 2008-2013: 843

Minus marks awarded to position players under NHL’s official system on 843 goals against: 4,228, 100 per cent of players on the ice.

Minus marks awarded to position players (non-goalies) under Neilson system on 843 goals against: 2,083, or 49 per cent of the players on the ice.

Roger Neilson's method...

After every game, he or one of his assistant coaches would go over videotape of the game, identifying all the scoring chances (hard shots from the slot that hit the net), then giving out plus marks to those players — and only those players — who contributed to scoring chances, as well as minus marks to only those players who made mistakes on goals against.

That method looks reasonable, although still doesn't account for roles/strengths of teammates-opponents.

I do like the fact that it includes non scoring plays.

Edit: Apologies for the broken link above...the filter blocks out the 'cuss' word...don't know how to fix that.

Edit#2: If you google 'edmonton oilers study plus minus' the article I linked above should be the first item listed. (Edmonton Journal article dated 13 May 2013)

 

[Sens Rule]

+/- is a measure of how effective a player is in his role 5-on-5, relative to his team.

So yes, occasionally a Priessing or Schultz who did a good job on the 3rd pairing of a good team has an excellent +/-. Doesn't mean he's a great player, just had a good season in his role at ES.

Seems like the people who complain about the stat expect it to be an all-encompassing measurement of 'good player' or 'good defensive player' and then get all offended when they find out it isn't. It's a somewhat useful stat for picking out outliers who are struggling/excelling in their role.

Fans could more accurately judge players if plus/minus did not exist at all. It is almost an entirely useless stat. Every time I see it about players or teams I follow closely it shows how useless it is without context.

You stating this "+/- is a measure of how effective a player is in his role 5-on-5, relative to his team."

Literally shows how you and most fans totally misinterpret that stat. Because there are many roles to be done 5-5 on a team. Pahlsson in his uber-elite defensive season in 06/07 was a MINUS player. Minus compared to many forwards who were a plus on his President's trophy team. So his plus/minus does not say crap about how good he was. He played on a line that had two other defensive forwards on it that almost exclusively faced the best lines of other teams 5-5. So while he might have only been scoring 30 points or something 5-5... So we're Thornton's line and every other top line the Ducks faced.

It seems like plus/minus ALWAYS REQUIRES CONTEXT to be at all useful or to understand why great defensive players have good or bad plus minus or good two way players do or elite scorers do or 1st or 3rd or 4th liners do. Or first vs third pair defenceman do.

I can explain pretty much why every player in the Senators had whatever plus/minus they did in any season since 05/06. Because I absolutely know their roles. But other teams? Maybe some in our division like Buffalo or Montreal or Boston.

Most teams in the league... I don't know the roles the players played. I don't know nearly enough about them. So if I take plus/minus to mean anything about any of the players I don't follow closely enough I am being pretty ignorant.

Because I know plus/minus means almost NOTHING.... For every player I know intimately by watching them most every game... It literally tells less then nothing about the play of the players looking at plus minus. But so many fans assume plus minus tells you "something". But it tells nothing if you have no context.

I know I have virtually no idea about how the Phoenix Coyotes roster is played. In what roles. With what linemates. Vs which type of opposition are they employed. So the plus minus makes me less informed about them and not more! Even in just comparing the plus minus vs the other members of one team.

But so many fans think it says something about a players defensive or offensive skill. It never... Never is good at doing this.

Plus minus should really just go away. I mean not even exist as a stat.

 

[Black Gold Extractor]

But so many fans think it says something about a players defensive or offensive skill. It never... Never is good at doing this.

Plus minus should really just go away. I mean not even exist as a stat.

Most stats can be improved with some context. Kind of like when Shawn Horcoff got 50 points in 53 games in 07-08 and got himself a $5.5 million average contract from the Oilers (a big overpay). It was a good season for him, but he did it by facing secondary competition while Jarret Stoll was thrown to the wolves. The fact that Horcoff benefited from facing easier opponents at home? That's not recorded on his boxcars.

Comparing players from 07-08 onward? Yeah, I'd ignore plus-minus completely unless the difference between players was really huge and sustained. But if we're looking at players from relatively long ago, it's literally the only publicly available and consistently-recorded stat that offers any semblance of a measure of defensive play relative to offense.