Wouldn't +/- be much better if

[Overrated]

1) A player who got fouled and thus created a PP would be be awarded a + if such PP resulted in a goal even if he weren't on the ice during the PP. Nobody else would be awarded a +.

2) A player who fouled thus penalized his team with a SH play would be awarded a - if such PP against his team resulted in a goal. Nobody else would be awarded a -.

3) Empty net goals didn't count.

I think this would make the stat way more useful.

Click to read more...

 

[morehockeystats]

1) A player who got fouled and thus created a PP would be be awarded a + if such PP resulted in a goal even if he weren't on the ice during the PP. Nobody else would be awarded a +.

2) A player who fouled thus penalized his team with a SH play would be awarded a - if such PP against his team resulted in a goal. Nobody else would be awarded a -.

3) Empty net goals didn't count.

I think this would make the stat way more useful.

I tried to use 2) and 3) in a slightly more different way:
ENG produced about -0.5 spread evenly between all attackers and +0.5 between all defenders
PPG produced about +2 spread evenly between pp, and -1 spread evenly between sh, and -1 spread evenly between those of the defending team in the penalty box who were on non-matching penalties.
SHG plus was reduced for the shorthanded side so that the amount of + and - handed was 0 in total.
The usefulness was not greatly improved. The projecting power was close to none.

 

[Overrated]

I tried to use 2) and 3) in a slightly more different way:
ENG produced about -0.5 spread evenly between all attackers and +0.5 between all defenders
PPG produced about +2 spread evenly between pp, and -1 spread evenly between sh, and -1 spread evenly between those of the defending team in the penalty box who were on non-matching penalties.
SHG plus was reduced for the shorthanded side so that the amount of + and - handed was 0 in total.
The usefulness was not greatly improved. The projecting power was close to none.

It might not change things overall but many highly penalized defensmen could suffer a lot with this statistic. Also the guys who create disproportionaly more penalties by getting fouled could benefit from this statistic quite a bit. At this point these are rather invisible deeds which aren't included.

 

[MVP of West Hollywd]

I think counting the guy in the penalty box as if he's "playing" is cool. Not sure about drawing the foul though.

 

[Czech Your Math]

For one, it shouldn't matter what happened on the ensuing PP. It's also contingent on players being penalized or drawing fouls in proportion to the number they actually deserve, which given the sometimes random nature of penalties called is a dubious assumption.

 

[Chili]

Believe teams have better numbers, they are just not made public.

Breaking down the NHL's secret stats | National Post

 

[JaegerDice]

There was a Red Wings blog awhile back that presented what they called their 'Common Sense Scoring Index', and went through each game adjusting the +/- for it.

So on a goal against, every player that did their job would have the minus removed from their stats, while the players responsible would keep their minus.

Likewise for goals for, players that nothing to do with the play would have their plus removed, while guys screening the goalie or drawing coverage would keep their plus.

Don't know whatever came of the project. I imagine once corsi and xg took off, such a system wasn't really needed, since any model built on goal-scoring will be less valuable than a model that includes, attempts, shots, chances, etc.

 

[plusandminus]

I tried to use 2) and 3) in a slightly more different way:
ENG produced about -0.5 spread evenly between all attackers and +0.5 between all defenders
PPG produced about +2 spread evenly between pp, and -1 spread evenly between sh, and -1 spread evenly between those of the defending team in the penalty box who were on non-matching penalties.
SHG plus was reduced for the shorthanded side so that the amount of + and - handed was 0 in total.
The usefulness was not greatly improved. The projecting power was close to none.

Why spread evenly when in reality it is very unevenly spread? It is usually the same players, over and over again, that either benefits or the-opposite-to-benefit by EN goals. Similar with SH goals.
I very much think EN goals should be eliminated from the official +/-, and They should get rid of SH goals too.
Just count even strength goals where both teams have their goalie on the ice.

 

[morehockeystats]

Why spread evenly when in reality it is very unevenly spread? It is usually the same players, over and over again, that either benefits or the-opposite-to-benefit by EN goals. Similar with SH goals.
I very much think EN goals should be eliminated from the official +/-, and They should get rid of SH goals too.
Just count even strength goals where both teams have their goalie on the ice.

You misunderstood.

 

[Hockey Outsider]

1) A player who got fouled and thus created a PP would be be awarded a + if such PP resulted in a goal even if he weren't on the ice during the PP. Nobody else would be awarded a +.

2) A player who fouled thus penalized his team with a SH play would be awarded a - if such PP against his team resulted in a goal. Nobody else would be awarded a -.

I don't think this should be included in plus/minus (it should only show what happens when a player's on the ice), but there might be some value in tracking this separately. There should be a clearer link between penalties taken, and the number of goals the team surrenders as a result. It's crazy that for most of the 1980's and 1990's, penalty minutes were seen as a good thing.

The downside is this introduces other biases/complications - a player who takes a lot of penalties in front of good defense/goaltending could look better than a much less disciplined player, who happens to play in front of strong teammates.

3) Empty net goals didn't count.

Agreed. This biases the stat in favour of defensive players, and against top scorers (because of the likelihood of allowing an empty-net goal against). There's the same bias with shorthanded goals being included in the calculation of +/-.

Maybe there's some value in tracking +/- for EN situations, but make that a separate category and exclude it from the main statistic. (Same with shorthanded goals for/against).

 

[AHockeyFanatic]

1) A player who got fouled and thus created a PP would be be awarded a + if such PP resulted in a goal even if he weren't on the ice during the PP. Nobody else would be awarded a +.

2) A player who fouled thus penalized his team with a SH play would be awarded a - if such PP against his team resulted in a goal. Nobody else would be awarded a -.

3) Empty net goals didn't count.

I think this would make the stat way more useful.

Why wouldn’t you just have a seperate special teams stat for that?

Makes no sense to get rid of one of the truly good statistics in hockey.

 

[bossram]

No. Just make +/- even-strength only. At least it'll measure what it was intended to.

It would still be a mostly useless stat, but at least meaningfully reflect what it is trying to measure.

For player evaluation, there are just better stats (e.g. scoring chance differential, xG differential, regressed metrics, etc.).

 

[mattihp]

I do not know if it would be better, but it would be different. Would it be a more valuable stat? Perhaps.

 

[MadLuke]

For one, it shouldn't matter what happened on the ensuing PP. It's also contingent on players being penalized or drawing fouls in proportion to the number they actually deserve, which given the sometimes random nature of penalties called is a dubious assumption.

First reflex, but a player that take penalty if they have the best PK in the league or draw them if they have the best PP, there is a different value in that context if it actually have any influence and control from players in that regard.

Could easily have both, one with the league average and one with the actual team.

One really easy and with almost no controversy modification is the EV with goaltender measure that could be shown just next to the traditional one.

 

[The Panther]

1) A player who got fouled and thus created a PP would be be awarded a + if such PP resulted in a goal even if he weren't on the ice during the PP. Nobody else would be awarded a +.

2) A player who fouled thus penalized his team with a SH play would be awarded a - if such PP against his team resulted in a goal. Nobody else would be awarded a -.

3) Empty net goals didn't count.

I think this would make the stat way more useful.

1) and 2) are interesting ideas, and certainly 3) should be the case, I agree. Awarding a 'minus' for empty-nets goals against is absurd. The question, though, is: Would this type of stat still be "plus/minus", or we getting into a new stat?

A few years ago (maybe in this forum), I came up with a "new" plus/minus system that I liked. I thought it more meaningful than actual "plus/minus", which, while interesting, doesn't really tell us anything useful in itself and is frequently enormously misleading.

If I remember correctly, my new system was a combination of (i) standard plus/minus (i.e., raw ES difference), (ii) difference of total goals-for minus goals-against, and (iii) some simple formula I had for calculating each player's goal-differential relative to how his teammates did when he was off the ice (I can't remember the details). I suggested that 1/3 + 1/3 + 1/3 combination of those three measurements would result in a stat that was really quite meaningful -- it would tell us a lot about how each player was actually impacting the team's fortunes when on the ice (which, for me, plus/minus doesn't do except in 'outlier' cases).

I took a bunch of historical examples of player-seasons and applied my formula to them and they seemed to come out quite reasonably.

Anyway, this stat I suggested is certainly different from Plus/Minus, but I think it would be a much better stat to include in "official" NHL stats.

 

[PROGFAN66]

1) A player who got fouled and thus created a PP would be be awarded a + if such PP resulted in a goal even if he weren't on the ice during the PP. Nobody else would be awarded a +.

2) A player who fouled thus penalized his team with a SH play would be awarded a - if such PP against his team resulted in a goal. Nobody else would be awarded a -.

3) Empty net goals didn't count.

I think this would make the stat way more useful.

The plus +/- system in Hockey is flawed in my opinion its really a team stat.

However, the stat doesn't take into account a hot goaltender or a weak goaltender. There are games if a team plays poorly on defense and the last line in the defense the goaltender has a dominant game.

I agree with your first two points though if it was based on an individual +/-.

 

[SnowblindNYR]

+/- is supposed to track on ice performance and this adjustment doesn't do that. Plus the player drawing a penalty might have no impact on the Powerplay scoring, he might be a 4th liner that gets no PP time.

 

[Overrated]

+/- is supposed to track on ice performance and this adjustment doesn't do that. Plus the player drawing a penalty might have no impact on the Powerplay scoring, he might be a 4th liner that gets no PP time.

Well at least he drew the penalty. He did something for his team after all.

 

[supsens]

The plus +/- system in Hockey is flawed in my opinion its really a team stat.

However, the stat doesn't take into account a hot goaltender or a weak goaltender. There are games if a team plays poorly on defense and the last line in the defense the goaltender has a dominant game.

I agree with your first two points though if it was based on an individual +/-.

Every stat is flawed then, there is not any stat that isn’t less dependent on your team than any other stat
All stats are team stats.
The irony is “isolated impact” the absolute most team dependent stat is sold as the opposite to people.
That should be the first red flag against the math people selling their goods online

 

[The Panther]

I found my old post, where I suggested a different stat that would be preferable (or an alternative) to Plus/Minus. I called it 'IV', or "Individual Value". Let me explain, with historical examples:

There are two important points to keep in mind when looking at strictly 5-on-5 goal differentials:
i) Without (team and other) contexts, plus/minus is completely meaningless (i.e., it's at least 50% -- maybe 80% -- a team stat, akin to a goalie's GAA)/
ii) Plus/minus only has regular significance when it exists as a clear 'outlier' stat.

Point (i) is clear, I think. If Bobby Orr had been traded to the disaster Washington Capitals for the 1974-75 season and partnered with Bill Mikkelson, and then had gone -25 for the Capitals, for example, it wouldn't mean that Orr suddenly was a weak defensive player overnight. We all know this.

Point (ii): If a bunch of players on a bad team have plus/minus ratings of -11, -15, -18, -9, -20, those numbers are all basically meaningless for comparative purposes. We can guess that these players' team had a bad season, but we cannot distinguish individual players contribution on ice from one another if they're all comparatively poor minuses. The exception is an outlier. If one player had gone -48 with average ice-time, and nobody else on the team had gone worse than, say, -18. Then, maybe we have a legit outlier. And especially if this had re-occurred in subsequent seasons, we might fairly be able to reasonably assume that this particular player was rather poor five-on-five.

But, that's rare. I would say in 95% of cases, plus/minus is a meaningless stat; as in, it doesn't tell us anything useful.

So, what do I mean that plus/minus 'rewards' or 'punishes' certain types of players? Well, obviously, players who are power-play specialists are not going to have their offensive value (i.e., their entire value, in some cases) represented in plus/minus. The stat clearly "discriminates" against these kinds of players, even if those players' only purpose on the team is to contribute to power-plays and they're very, very good at it. To use an old example -- Tim Kerr on the 1985-86 Flyers. Kerr, for those who don't know, was a very sluggish skater (like Lucic now, but much slower) with fabulous hands, who banged in power-play goals galore for the 80s' Flyers. The 1986 club was 2nd overall in the NHL, the #1 team defensively and the 4th-best (1st in East) offensively. Kerr scored 58 goals in 76 games for Philly, and was second on the club in points. Yet, he finished -2 on the season! (His teammate, Mark Howe, was +87.) The reason, of course, is that Kerr did most of his damage on the PP, where he struck 34 times (and had 56% of his points). So, of what use is plus/minus to such a player? None.

It's also discriminatory to some very good all-defense forwards. Another historical example is Bob Gainey -- the guy who the Selke trophy was invented for. He won the Selke four seasons in a row from '78 to '81. During those four seasons with Montreal, Gainey's plus/minus rating on the team was ranked 17th, 15th, 23rd, and 11th. Yet he both played for a dominant, winning team and was judged the best defensive forward in the game every single season. So, of what value is plus/minus with such players? None whatsoever.

Plus/minus also offers meaningless results with many, many very good (Hall of Fame) players on poor or mediocre teams. It not only makes the players look worse then they were (due to ignoring their PP contribution), but also in some cases makes them look worse than they actually were in comparison to teammates.

Wayne Gretzky offers a good example to break down. Gretzky's plus/minus was utterly dominant from the early-80s to late-80s, and stayed strong his first three years in L.A. As he passed his prime years after autumn 1991, his plus/minus went down, and he finished a minus for the first time on the 1991-92 Kings. Two years later, as the Kings bottomed out with a .393 record and missed the playoffs, Gretzky finished the season with a -24, the worst-such mark on the club (further to my point, the three worst on the team were Gretzky, Robitaille, and Kurri... their three best players).

With that in mind, here are the Kings' 1993-94 top-eight scorers, ranked not by points or by standard plus/minus (i.e., even-strength) but ranked instead by total goal differential:
+50 Robitaille
+29 Zhitnik
+24 Gretzky
+20 Sandstrom
+16 Blake
+3 Donnelly
-3 Kurri
-12 Sydor

This gives us a somewhat more rounded picture of what the more offensively successful players' actual contribution, overall, was to the team.

We might parse this further, because of course some players (Blake, Kurri, Gretzky) were regular penalty-killers, while some others (Robitaille, Donnelly) were not. So, taking the same top-eight scorers on the Kings but ranking them again after removing PP-goal against:
+63 Gretzky
+60 Blake
+53 Robitaille
+48 Zhitnik
+32 Kurri
+27 Sandstrom
+17 Donnelly
+10 Sydor

In sum then, Gretzky's plus/minus was the worst on the 1994 Kings, but when you factor in his actual offence (incl. power plays) and ignore the PP-goals against while he killed penalties, he (not surprisingly) has the best result on the team.

Now, no one of these three stat-results (official plus/minus or the two others I mention) is the be-all and end-all, but five-on-five results only tell us part of the story, clearly (unless you think Coffey and Lemieux were the worst players on the 1990 Penguins, and that Lemieux was the third-worst forward on the '85 Penguins). We can see that Gretzky was being used in all situations in 1993-94 -- ES, PP, SH -- and that he was not particularly effective, overall, at ES (in the latter half of the season, when the team started throwing in the towel). So, let's take Gretzky's 1993-94 goals' results and put it into the three categories: (1) official plus/minus (ES results); (2) actual goals-for and goals-against results; (3) actual results minus PP goals against. So, his three-part stats ends up:
(1) -25 (2) +24 (3) +63

What might be interesting is to compare these three stats with his other seasons on struggling teams and with other, similarly offensive star-forwards on other teams of the same era when the teams were somewhat similarly poor.

- W. Gretzky .431 Oilers 1979-80: (1) +14 (2) +39 (3) +57
- W. Gretzky .463 Oilers 1980-81: (1) +41 (2) +91 (3) +108
- M. Lemieux .475 Pens 1985-86: (1) -8 (2) +73 (3) +75
- S. Yzerman .488 Red Wings 1986-87: (1) -2 (2) +38 (3) +52
- M. Lemieux .450 Pens 1989-90: (1) -18 (2) +17 (3) +49
- J. Sakic .325 Nordiques 1991-92: (1) +5 (2) +27 (3) +51
- W. Gretzky .393 Kings 1993-94: (1) -25 (2) +24 (3) +63

Now, I realize not every NHL player, or forward, gets power-play time, and thus giving out lots of 'pluses' to forwards on the PP might also be discriminatory in one way, since it's bound to make one-dimensional scorers look better in comparison to two-way players. I wonder if combining the results of all three of these stats might be more meaningful? Like, if we assigned a 1/3 value (divide each by three) to all three of Gretzky's results from 1993-94 and add them, he then comes out at -8.3 +8 +21 = +21. On the same team, Robitaille is -6.7 +16.7 +17.7 = +27.7. By the way, the guy who actually had the best plus/minus on the Kings that season was John Druce (!) at +16 in 55 games played. If we apply this system to Druce's season, he comes out +5.3 + 6.7 + 6.3 = +18.3, which still looks good, but respects the fact that he didn't actually tilt the ice in the Kings' favor more than Gretzky or Robitaille.

Remember Bob Gainey? He was a -1 in 1979-80 and won the Selke. In my proposed system, he comes out -0.3 -2 +5 = +3. Okay, it's not a huge improvement, but at least it's a 'plus'.

How about Mario Lemieux on the 1985-86 Pens? He was a -8, despite scoring 141 points and the team's being just four games below .500. In my proposed system, he comes out -2.7 +23.7 + 23 = +44.

____________

So, that's "Individual Value". And then, there could be another stat that was 'LIV' or "League-Individual-Value", which would be similar but which would also take into account the relative strength of each team (i.e., delete the team-strength aspect of 'plus/minus', so that even some players on the 1974-75 Capitals would rank up there with Bobby Orr, even though their 'IV' would be really low in comparison).

_____________

So, those are my ideas. The NHL should relegate current Plus/Minus to become an aficionados-only stat (like Corsi), and should instead use "Individual Value" and "League Individual Value" as official stats.

 

[Oddbob]

Why on earth would you get a plus for helping the team get a PP opportunity? The players on the ice still have to score.

 

[Overrated]

Why on earth would you get a plus for helping the team get a PP opportunity? The players on the ice still have to score.

Nah you would only get a + if your team scored. If it didn't you drawing the penalty would get you nothing. Same with the -. You'd get the minus only if your penalty resulted in a powerplay goal.

This would not change much but I think it would make it more accurate. Think about it this way. A guy who fouls a lot sometimes gets away with his rough play which in turn makes his ES stats better. We however don't see what happens when his rough play gets penalized. If that results in more goals against than the amount of goals he prevents at ES it makes sense for it to show in a stat. Likewise a guy who draws disproportionately more penalties compared to others is creating more opportunities for his team and if these opportunities result in scoring I think it makes sense for it to show.

This is just an idea I am not saying the current +/- is all bad.

 

[Czech Your Math]

I found my old post, where I suggested a different stat that would be preferable (or an alternative) to Plus/Minus. I called it 'IV', or "Individual Value". Let me explain, with historical examples:

There are two important points to keep in mind when looking at strictly 5-on-5 goal differentials:
i) Without (team and other) contexts, plus/minus is completely meaningless (i.e., it's at least 50% -- maybe 80% -- a team stat, akin to a goalie's GAA)/
ii) Plus/minus only has regular significance when it exists as a clear 'outlier' stat.

Point (i) is clear, I think. If Bobby Orr had been traded to the disaster Washington Capitals for the 1974-75 season and partnered with Bill Mikkelson, and then had gone -25 for the Capitals, for example, it wouldn't mean that Orr suddenly was a weak defensive player overnight. We all know this.

Point (ii): If a bunch of players on a bad team have plus/minus ratings of -11, -15, -18, -9, -20, those numbers are all basically meaningless for comparative purposes. We can guess that these players' team had a bad season, but we cannot distinguish individual players contribution on ice from one another if they're all comparatively poor minuses. The exception is an outlier. If one player had gone -48 with average ice-time, and nobody else on the team had gone worse than, say, -18. Then, maybe we have a legit outlier. And especially if this had re-occurred in subsequent seasons, we might fairly be able to reasonably assume that this particular player was rather poor five-on-five.

But, that's rare. I would say in 95% of cases, plus/minus is a meaningless stat; as in, it doesn't tell us anything useful.

So, what do I mean that plus/minus 'rewards' or 'punishes' certain types of players? Well, obviously, players who are power-play specialists are not going to have their offensive value (i.e., their entire value, in some cases) represented in plus/minus. The stat clearly "discriminates" against these kinds of players, even if those players' only purpose on the team is to contribute to power-plays and they're very, very good at it. To use an old example -- Tim Kerr on the 1985-86 Flyers. Kerr, for those who don't know, was a very sluggish skater (like Lucic now, but much slower) with fabulous hands, who banged in power-play goals galore for the 80s' Flyers. The 1986 club was 2nd overall in the NHL, the #1 team defensively and the 4th-best (1st in East) offensively. Kerr scored 58 goals in 76 games for Philly, and was second on the club in points. Yet, he finished -2 on the season! (His teammate, Mark Howe, was +87.) The reason, of course, is that Kerr did most of his damage on the PP, where he struck 34 times (and had 56% of his points). So, of what use is plus/minus to such a player? None.

It's also discriminatory to some very good all-defense forwards. Another historical example is Bob Gainey -- the guy who the Selke trophy was invented for. He won the Selke four seasons in a row from '78 to '81. During those four seasons with Montreal, Gainey's plus/minus rating on the team was ranked 17th, 15th, 23rd, and 11th. Yet he both played for a dominant, winning team and was judged the best defensive forward in the game every single season. So, of what value is plus/minus with such players? None whatsoever.

Plus/minus also offers meaningless results with many, many very good (Hall of Fame) players on poor or mediocre teams. It not only makes the players look worse then they were (due to ignoring their PP contribution), but also in some cases makes them look worse than they actually were in comparison to teammates.

Wayne Gretzky offers a good example to break down. Gretzky's plus/minus was utterly dominant from the early-80s to late-80s, and stayed strong his first three years in L.A. As he passed his prime years after autumn 1991, his plus/minus went down, and he finished a minus for the first time on the 1991-92 Kings. Two years later, as the Kings bottomed out with a .393 record and missed the playoffs, Gretzky finished the season with a -24, the worst-such mark on the club (further to my point, the three worst on the team were Gretzky, Robitaille, and Kurri... their three best players).

With that in mind, here are the Kings' 1993-94 top-eight scorers, ranked not by points or by standard plus/minus (i.e., even-strength) but ranked instead by total goal differential:
+50 Robitaille
+29 Zhitnik
+24 Gretzky
+20 Sandstrom
+16 Blake
+3 Donnelly
-3 Kurri
-12 Sydor

This gives us a somewhat more rounded picture of what the more offensively successful players' actual contribution, overall, was to the team.

We might parse this further, because of course some players (Blake, Kurri, Gretzky) were regular penalty-killers, while some others (Robitaille, Donnelly) were not. So, taking the same top-eight scorers on the Kings but ranking them again after removing PP-goal against:
+63 Gretzky
+60 Blake
+53 Robitaille
+48 Zhitnik
+32 Kurri
+27 Sandstrom
+17 Donnelly
+10 Sydor

In sum then, Gretzky's plus/minus was the worst on the 1994 Kings, but when you factor in his actual offence (incl. power plays) and ignore the PP-goals against while he killed penalties, he (not surprisingly) has the best result on the team.

Now, no one of these three stat-results (official plus/minus or the two others I mention) is the be-all and end-all, but five-on-five results only tell us part of the story, clearly (unless you think Coffey and Lemieux were the worst players on the 1990 Penguins, and that Lemieux was the third-worst forward on the '85 Penguins). We can see that Gretzky was being used in all situations in 1993-94 -- ES, PP, SH -- and that he was not particularly effective, overall, at ES (in the latter half of the season, when the team started throwing in the towel). So, let's take Gretzky's 1993-94 goals' results and put it into the three categories: (1) official plus/minus (ES results); (2) actual goals-for and goals-against results; (3) actual results minus PP goals against. So, his three-part stats ends up:
(1) -25 (2) +24 (3) +63

What might be interesting is to compare these three stats with his other seasons on struggling teams and with other, similarly offensive star-forwards on other teams of the same era when the teams were somewhat similarly poor.

- W. Gretzky .431 Oilers 1979-80: (1) +14 (2) +39 (3) +57
- W. Gretzky .463 Oilers 1980-81: (1) +41 (2) +91 (3) +108
- M. Lemieux .475 Pens 1985-86: (1) -8 (2) +73 (3) +75
- S. Yzerman .488 Red Wings 1986-87: (1) -2 (2) +38 (3) +52
- M. Lemieux .450 Pens 1989-90: (1) -18 (2) +17 (3) +49
- J. Sakic .325 Nordiques 1991-92: (1) +5 (2) +27 (3) +51
- W. Gretzky .393 Kings 1993-94: (1) -25 (2) +24 (3) +63

Now, I realize not every NHL player, or forward, gets power-play time, and thus giving out lots of 'pluses' to forwards on the PP might also be discriminatory in one way, since it's bound to make one-dimensional scorers look better in comparison to two-way players. I wonder if combining the results of all three of these stats might be more meaningful? Like, if we assigned a 1/3 value (divide each by three) to all three of Gretzky's results from 1993-94 and add them, he then comes out at -8.3 +8 +21 = +21. On the same team, Robitaille is -6.7 +16.7 +17.7 = +27.7. By the way, the guy who actually had the best plus/minus on the Kings that season was John Druce (!) at +16 in 55 games played. If we apply this system to Druce's season, he comes out +5.3 + 6.7 + 6.3 = +18.3, which still looks good, but respects the fact that he didn't actually tilt the ice in the Kings' favor more than Gretzky or Robitaille.

Remember Bob Gainey? He was a -1 in 1979-80 and won the Selke. In my proposed system, he comes out -0.3 -2 +5 = +3. Okay, it's not a huge improvement, but at least it's a 'plus'.

How about Mario Lemieux on the 1985-86 Pens? He was a -8, despite scoring 141 points and the team's being just four games below .500. In my proposed system, he comes out -2.7 +23.7 + 23 = +44.

____________

So, that's "Individual Value". And then, there could be another stat that was 'LIV' or "League-Individual-Value", which would be similar but which would also take into account the relative strength of each team (i.e., delete the team-strength aspect of 'plus/minus', so that even some players on the 1974-75 Capitals would rank up there with Bobby Orr, even though their 'IV' would be really low in comparison).

_____________

So, those are my ideas. The NHL should relegate current Plus/Minus to become an aficionados-only stat (like Corsi), and should instead use "Individual Value" and "League Individual Value" as official stats.

I agree the better PP players tend to be shortchanged by plus-minus.

Individual total goal differential means almost nothing, for as you said players have different roles. Deducting PGA gives players the same credit for PGF (when it's much easier to score and there's no offset category like ESGA is for ESGF) as a positive ES differential (which is obviously much, much more difficult to attain).

ES goal differential (ESGF - ESGA) is simply a metric of the ES advantage or disadvantage the team had while that player was on the ice at ES. This may also be unfair (at least in some cases) to scorers, since they may be greatly responsible for the ESGF side of the equation, but only a small part of the ESGA side.

To include PP in the equation, one has to determine what advantage or disadvantage the team had while the player was on the PP. One could compare the team's PP% to the league average or to replacement level (say, the worst team in the league in PP%). Still, this may be unfair to players that are better on the PP than the team average.

I'm slowly tinkering with some ideas on this, but it's complex. One example is adjusting individuals' ESGA to league average SV% (so a team with above average SV% would have their players' ESGA adjusted upward, and vice versa). It seems like shutdown D and checking forwards will be difficult to value using this type of system, but for other skaters I think there is good potential.

 

[Czech Your Math]

Here's a starting point for something along the lines of "Individual Value":

First, as I said before, I don't know that this will work well at all for players that derive the vast majority of their value from ES defense (shutdown d-men and checking forwards).
Second, while I think something like this could work well for comparing at least certain types of players within a single season, there would need to be further adjustments to compare players from different seasons. For one, ES/PP/SH scoring levels vary from season to season, which allows more value to be apparently created in some seasons compared to others, so one would have to adjust for scoring levels. Also, parity varies from season to season, which should affect replacement level (since I'm basing it here on the worst team in the league in that particular category). A solution might be to use an average or median for a multi-year period, to smooth out variations due to randomness and/or differing environments (e.g., expansion or otherwise unusually bad teams). Finally, another issue may be scale, such that comparing skaters to goalies (using goals saved vs. league avg. or replacement level) may not be possible, since the scale of the metrics for each may be different.

Panther used '94 Gretzky as an example, so I'm comparing '94 Gretzky to '94 Fedorov.

Even Strength Value

One starting point would be Adjusted Plus-Minus: Adjusted Even-Strength Plus-minus 1960-2017 | HFBoards - NHL Message Board and Forum for National Hockey League (mandatory.com)

Basically, we calculate ESGF & ESGA from TGF, PGF, TGA & PGA.
ESGF = TGF - (PGF + SGF) ... SGF is estimated from team SGF and the individual's ratio of PGA (individual PGA/team PGA). One difference is that if individual's SHP (SHG + SHA) were more than the estimate, I used SHP instead of the estimate (since it had to be at least that much).
ESGA = TGA - (PGA + SGA) ... SGA is estimated from team SGA and the individual's ratio of PGF (individual PGF/team PGF).

Once we have individual ESGF & ESGA, "OFF" (team's ESGF & ESGA) is easily derived from (Team - Individual) for each of ESGF & ESGA.

Overpass, the author of that study, regressed the expected plus-minus (using OFF) to average (1.0), but I am not bothering with that here (although I'm not necessarily against) for the sake of simplicity.
The regression he does favors players on above average ES teams and penalizes players on below average ES teams (again, I'm not saying that's necessarily unjustified).

Without regressing to average, the adjusted plus-minuses are: Gretzky -12.8 & Fedorov +35.4
The raw (ESGF - ESGA) are: Gretzky -25.2 & Fedorov +45.4 ... Overpass' adjusted plus-minus would be between the two numbers (i.e. Gretzky -12.8 to 25.2, Fedorov +35.4 to 45.4), so in this case not regressing to the average yields a smaller gap between the two players.

I also calculated ESV another way. I gave the player credit for ESGF based on his ES points (ESG + ESA), but penalized him for ESGA based on him being one of five skaters on the ice. Before I go through the details of the calculation, I should say that I think it would be fairest to first adjust ESGA for team SV% vs. league average, but I didn't do that here (this again favored Gretzky, since Kings were around league avg., while Wings were worse).

Gretzky had 20 ESG & 42 ESA, while he was on ice for an estimated 77.0 ESGF. So he had points on 62/77 ESGF or 80.5%. His team had 479 ESP on 183 ESGF or 2.62 points awarded per ESGF.
Dividing 80.5% by 2.62, we get 30.8% and will call this his ESGF Responsibility %. We divide Responsibility for ESGA equally among each of the five skaters (this is an assumption for simplicty, the actual number would be slightly lower), so 20% each. We do the same for Fedorov, who had points on 75.0% of ESGF.

This yields these numbers:
Gretzky: ESGF = 77.0 * .308 = 23.7 ; ESGA = 102.2 * .20 = 20.4... so +3.3
Fedorov: ESGF = 108.0 * .287 = 31.0; ESGA = 62.6* .20 = 12.5... so +18.5

Power Play Value

We're using the worst team in the league in that category as "Replacement Level." So the Kings converted 20.7% of their PPs, while the Ducks only converted 14.4% of their PPs. We multiply the difference by the Kings' PP opportunites to yield: 444 PPO * (.207 - .144) = 28.2 Goals Above Replacement Level. So how do we distribute part of those 28.2 GARL to Gretzky? He had 14 PPG & 47 PPA for 61 PP points on 78 PGF, so 61/78 = 78.2%. Dividing 78.2% by 2.89 points awarded per Kings PP goal, we get 27.0% for his PGF Responsiblity %. We multiply that by the 28.2 GARL for his team to get 18.7 GARL for PGF. We do the same for Fedorov to get 9.5 GARL for PGF.

We calculate how many SGA were saved by using the same methodology for SH goals given up (but equally divided among five or 20% each), which yields Gretkzy 2.5 & Fedorov 0.9

PPV Totals: Gretzky = 18.7 +2.5 = 21.2; Fedorov = 9.5 + 0.9 = 10.4

Short Handed Value

Using the same methodology for SGF as for PGF, we get: Gretzky 3.0 & Fedorov 4.9 for GARL while SH.

We calculate how many PGA were saved by using the same methodology as For SGA (excep tequally divided among four for 25% each) which yields: Gretzky 8.3 & Fedorov 8.8

SHV Totals: Gretzky = 3.0 + 8.3 = 11.3; Fedorov = 4.9 + 8.8 = 13.8

Summary

ESV
---Simple (ESGF - ESGA): Gretzky -25.2, Fedorov +45.4
---"Raw" Adjusted Plus-Minus: Gretzky -12.8, Fedorov +35.4
---Responsibility %: Gretzky +3.3, Fedorov +18.5

PPV: Gretzky 21.2, Fedorov 10.4

SHV: Gretzky 11.3, Fedorov 13.8

TOTALS, Using
---Simple (ESGF - ESGA): Gretzky +7.4, Fedorov +69.5
---"Raw" Adjusted Plus-Minus: Gretzky +19.8, Fedorov +59.6
---Responsibility %: Gretzky +35.8, Fedorov 42.6

The first two are rather large gaps. Another thought is that perhaps, assuming the line of three forwards is driving ES advantage/disadvantage, is to divide those numbers (Simple ESGF-ESGA & Raw Adj. PM, since Resp.% already distributes the advantage/disadvantage) by three. If we did this and made the adjustment for SV% vs. league avg., the gaps between Gretzky & Fedorov would be about 17, 10 & 8 goals, respectively. That range seems reasonable to me, but I welcome any comments or suggestions.

 

[TeddyBare]

Penalty differential isnt discussed nearly enough on Hfboards

I remember Yzerman discussing its importance in an interview