79

Stop Using Plus-Minus

I had an interaction this morning that hurt my brain and my soul a little. It was about the difference between Plus-Minus (+/-) and Goal Differential (5v5, Even Strength, and All Situations). My takeaway from the interaction was that we still have a long way to go in dispelling the awful mythology of plus-minus.

Let us begin with the basics. If you think that +/- is the stat that tells you the difference between how many goals for and against were scored with a player on the ice, you are incorrect. That is not what +/- does, even though a lot of people think that’s what it does.

+/- is actually a mix of all kinds of things with a few caveats. It keeps track of even strength scoring for and against, but it also counts shorthanded goals for and against. In addition to that, it also counts empty net goals that are scored. So what it does is mix a whole bunch of game states together. That’s where this thing really falls apart.

The difference between Plus-Minus and Goal Differential used by anyone else in the stat community is that the nerds don’t want to cross the streams. It makes more sense for us to all know what we’re looking at when we discuss numbers. The standard game state of discussion is 5v5 play since the majority of a skater’s time is played there. It is not limited to that game state though, so when others are discussed, they are cited appropriately. For example, often the PP is broken down to even more specifically the 5v4 PP and conversely the PK to the 4v5 PK.

This way we can more clearly break down if a player is really struggling all the time, 5v5, or on one of the special teams. Empty net situations are almost never discussed because the amount of time spent playing in that state is extremely limited and under some very obviously negative conditions. If your goalie is pulled at the end of the game, it isn’t because you’re already up by three and you really want to rub it in Calgary’s face. You also aren’t likely playing the 4th line with the goalie pulled either.

When a team, like Edmonton, has been bad for the better part of a decade then those teams should have more opportunities to play with the empty net. Bad teams lose more and conversely have more situations when pulling the goalie late is a viable tactic. When they do that, good players are put on the ice in an attempt to score, but without a goalie, it’s very easy to be scored against.

For reasons that don’t make any sense, the NHL will give all of those players who were on WITHOUT A GOALIE IN THE NET a minus for that goal. The longer a quality player is employed by a bad team, the more of these minuses he will get.

EXAMPLES

Oct 19, 2017; Ottawa, Ontario, CAN; New Jersey Devils left wing Taylor Hall (9) skates with the puck in the second period against the Ottawa Senators at Canadian Tire Centre. Mandatory Credit: Marc DesRosiers-USA TODAY Sports

The interaction that sparked this article was about Taylor Hall. I don’t want to turn this into an article about Hall though, but he is an interesting case study in +/- vs goal differential.

Over the course of his Oiler career, Taylor Hall was a -27 per the NHL’s stats page. Ignoring all context, being -27 is bad. To a lot of people, that suggests to them that Taylor Hall was poor defensively. Positive players are good. Negative players are bad. Makes perfect sense. Right?

However, when it comes to 5v5 goal differential, Taylor Hall was +3 as an Oiler. When it comes to All Situations goal differential, Taylor Hall was +92 as an Oiler. In his Oiler career, the club scored 426 goals for and had 334 scored against with him on the ice. The Oilers only scored 686 for and had 1074 against with him on the bench (all situations) for a -388 goal differential.

Still, +/- tells you that Hall was a negative player. It has to do with the fact that with 6 skaters on the ice (goalie pulled), the Oilers had only 21 goals for but 57 goals against during his time as an Oiler. That is a -36 being applied to a bunch of players who the coach trusted to play without a goalie because they needed a goal. If you want to know how Taylor Hall ended up with a -27 as an Oiler, this game state is the culprit.

For a non-Oiler example, let’s look at Jack Eichel. Since coming into the NHL Jack Eichel has been playing for the Sabres, who are Oilers-East but without the lottery wins. In just 2.5 seasons Eichel is a combined -45 in Buffalo. It’s ugly.

In 5v5 goal differential, Eichel is actually -22 in that time. That’s half the negative that +/- reports! In all situations, he is +25 in goal differential (thank you, Power Play). The Sabres have scored 208 goals and had 183 scored against with him on the ice. They have scored 102 goals and 210 against with him on the bench for a -108 differential.

The difference between -22 5v5 and -45 by +/- standards is still a lot. What’s going on there? Well, in just 2.5 seasons the Sabres have a -22 goal differential with the goalie pulled. We can be pretty sure that Eichel is on the ice for the majority of the time the goalie is pulled.

What about Connor? Deep down, everything is about Connor.

Since he joined the NHL, Connor McDavid is +35 (highest among Oiler players). But even he is getting short-changed! 5v5 McDavid has a +39 goal differential. In All Situations, he has a +74 goal differential. Connor McDavid is offensive potential in human form and that ridiculous empty net/shorthanded goal portion of +/- is even affecting him.

We can run this exercise for every good player on bad teams in the NHL and you’ll see the same thing over and over again. Goal differential is a very coarse measure because goals happen pretty infrequently, but if we start mixing and matching game states like +/- then we lose the ability to be certain about the things *we think* we know about players.

For me, if I want to get closest to the spirit of what most people think they’re getting from looking at +/- then I go to 5v5 goal differential. There are no special teams. The game isn’t being played under really strange conditions. It’s hockey as we know it. I am happy that we have tools available to us that let us get much more precise than +/- ever can. We can break things down really quickly and be able to say more with more confidence because of it.

So please don’t ever use plus-minus. Ever. It’s hot garbage. We can do better. It’s 2018.

*stats courtesy of Corsica.Hockey

    • BringitbacklikeSlats

      Corsi is based on counting shots and judges players on the quantity of these shots but not the quality. As a team or even line assessment tool I agree it’s got some value. But as an individual players assessment its of little use and really only slightly better than +/-.
      Hits, takeaways, passes, and player movement off the puck indirectly lead to goals not just shots and to assume only shots lead to goals is ludicrous.
      Also by only assessing shots Corsi ignores goals, or more specifically ignores defense. Perhaps a player did allow more shots against but he forced them to be wide or forced bad angle shots with a low chance of them getting by the goaltender. Corsi does not allow for the analysis of how well a player may be playing defensively in the zone despite allowing these shots.
      Good players on bad teams will suffer from this statistic simply because they will never be able to muster up enough shots while bad players on good teams will look far better than they actually are.

  • Jordan88

    Hall and Eberle were awful defensively. Eberle often changed when the puck was coming back. Hall often coughed the puck up or gave it away.

    Stop looking at numbers and watch the game please.

      • Jordan88

        I never ripped on Henderson, but if you can’t see Hall or Eberles lapses in their own end you’re blind. +3 over what 7 seasons at 5×5? And basically a 70-80 points per season.

        Yea edmonton was bad but you should be able to be better than +3.

        • fisherprice

          Nobody is saying they were perfect defensively or Selke candidates. Of course they had bad turnovers once in a while. This is also true of literally any gifted offensive player in the entire league. If you’re willing to take the risks to score goals, sometimes it’ll come back to bite you. This is the kind of idiotic thinking that leads morons to believe that Erik Karlsson is bad at defence.

          You also have to keep in context that Hall and Eberle basically ALWAYS saw the other team’s best match-ups, playing against top pairings and top lines. The fact that they came out on the positive end of the ledger is remarkable considering every team knew if they shut down the Oilers top line they were going to win the game.

          What it often comes down to is a bunch of people who can’t adjust to how the game is played nowadays. They don’t understand that good defensive play isn’t simply squaring up an opposing player or blocking shots anymore, it’s also the ability to transition the puck back the other way towards your opponents net. The only player who is better at this than Hall in the entire league is Connor.

          • Gravis82

            Every player is imperfect. If we have enough good players, everyone imperfections would be ignored because we would be winning.

            When you only have a few good players on a losing team, people think they are losing because the star player is not good enough…when really they are losing because they just don’t have enough good players.

          • The Whispererer

            Well said fisherprice. You are entirely correct. This fan base has booed far too many good players out of town; explains why we now have no reliable shooters on the wings with our top 3 centers.

          • aye

            “The only player who is better at this than Hall in the entire league is Connor”
            Really? So, you are saying Hall is the second best player behind McDavid in the entire league? And you would take Hall over players like Kucherov, Stamkos, Tavares, Crosby, etc.? Time to go play with your fisherprice Hall action figures.

          • Heavy Stick

            He’s slightly better than klef!
            Coaches us plus-minus so it means something. Really just about any fancy is a group metric. 5 guys on ice for shots for and against.

    • puckle-head

      What a silly, vacuous statement. Of course the “stats nerds” are watching the game, that’s actually how a lot of this data started being collected. They wouldn’t just watch the games but record them, and rewatch them over and over. Statistics are a valuable tool to supplement what you are seeing, and both forms of studying the game can provide context to the the other. Frankly I don’t trust people who don’t trust statistics. It may not always be apparent what knowledge you can glean from numbers, but they never lie. If you don’t want your ways of thinking scrutinized by any type of statistical analysis you are at best misguided, at worst afraid that your ways of thinking will be exposed as fundamentally wrong.

        • puckle-head

          Well of course I only trust some statistics. I’m sure we could track how many toilets are flushing in Rogers place when the Oiler’s penalty kill lets in a goal and come up with an average flush/SHG, but that would be a completely useless stat. The OP I was responding to only said, “stop looking at numbers.” It’s perfectly legitimate to debate what a particular statistic means, or if it’s meaningful at all, but to disregard that kind of analysis entirely is silly.

          • Gravis82

            There is a difference between two things being associated and one thing being the CAUSE OF the other. When you get them confused you end up making bad decisions.

      • Mc🙏

        They might ‘never lie’ but they can be looked at to prove ones bias, as much as someone using +/- to prove their point. It is this type of thinking ( i dont trust prople who font trust statistics ) that continue to fuel the back and forth debate, goving ‘conventional hockey people reasoning to ‘not trust people who only believe stats’.

    • Klef abs

      Most of the really great players in the league give the puck away alot. But look at their takeaways or the differential then they usually shine. Guys like Hall, Tavares, Burns, Karlsson are all up there in giveaways. But they are also the leaders in takeaways. Hall was always one of the best differential in that sense as well. But the idiot fans without brains only see the giveaways. Tavares has always been one of the leagues worst in that stat. I guess he sucks and is bad defensively too?

  • BringitbacklikeSlats

    It’s a truly valid point there Matt and one I won’t dispute. The fact that it includes special team play is ridiculous and skews it to a point of unreliability.

    Kind of like how Corsi as an individual player metric penalizes for blocking shots and rewards any shot on net regardless of how bad it is or where it’s taken.

    • puckle-head

      You are right, Corsi is an imperfect stat. But you’re misinterpreting it’s intent. The point of corsi isn’t to track shots because shots might lead to goals. The point is to use shot attempts To figure out which team is spending the most time with the puck in the offensive zone, and since you need to have the puck to shoot it (whether it’s blocked or not), shot attempts make a decent stand in for procession. Of course, the concern is that some coaches might “game” possession by encouraging their players to shoot from any and all angles, so it is flawed, but not for the reason you stated.

      • BringitbacklikeSlats

        Actually it’s you who misinterpreted my comments. Thank you for the explanation but I’m aware of the intent of the metric. What you failed to recognize in my statement was “as an individual player metric”.

        I can get behind it insofar as a number to contemplate in regard to line or team play. But I certainly don’t value it as an individual players statistic… and neither should anyone else

          • BringitbacklikeSlats

            Corsi is solely based on counting shots and judges players on the quantity of these shots but not the quality.
            Hits, takeaways, passes, and player movement off the puck also indirectly lead to goals not just shots and to assume only shots lead to goals is ludicrous.
            Also by only assessing shots Corsi ignores goals, or more specifically ignores defense. Perhaps a player did allow more shots against but he forced them to be wide or forced bad angle shots with a low chance of them getting by the goaltender. Corsi does not allow for the analysis of how well a player may be playing defensively in the zone despite allowing these shots.
            Good players on bad teams will suffer from this statistic simply because they will never be able to muster up enough shots while bad players on good teams will look far better than they actually are.

  • OliverPromenade

    Why don’t you write a letter to the league instead of posting articles whining about what you don’t like? What do you expect us to do about it? This is nothing more than whining

  • Mark Lesser

    I think +/- is a good stat. However, I completely agree that the NHL should not count empty net goals in the calculation. Maybe the NHL counts empty net goals because it is an attempt to reward the +/- of defensive minded players who are on the ice at the end of a game. Or maybe it is just the inability to use common sense.

  • Disappointed

    I did not know that, thank you Matty. Hall was always a good player, just not a god. One and dones ending with a muffin in the goalies chest is what I rememeber him as, oh and skating away from teamates in a scrum. Is goal differential a NHL recognized stat?

    • OilersGM

      I don’t blame Hall for walking away from scrums because when he got drilled few time (Doughty) no one came to his aid long before he walked away. Also some moron not naming names (Eakins) gave the captaincy to Ference when it should’ve been Hall that’s where IMO his frustration came to show. Another one of many idiotic moves this organization has made over the years.

  • camdog

    If one bad player is on a winning team and plays with 4 other good players they will generally have a good +/- and a good Corsi. These stats don’t really mean anything on their own.

    • Rock11

      Which is why anyone doing any real analysis of a player always uses corsi rel. This measures the individual against his own teammates in an effort to determine which player is having the bigger impact. Not to say this is perfect either. But then again one could argue goals and assists are flawed as well. Consider player X gets the puck off the opposition in his own end, skates it out of the zone, dekes 3 guys and the goalie only to hit the post. A teammate then grabs the rebound has his shot stopped and a second teammate then bounces it off the shin of a third teammate into the net. Player X gets nothing and shinpad guy gets the goal. All stats have flaws. Knowing when and how to use them is the key.

      • camdog

        3-4 years ago everybody would say that you need to build a team like the Kings, their metrics were through the roof. This season the Oilers numbers look similar to the Kings numbers of old. As the game changes the numbers don’t always mean what we thought they meant. There is a lot of which came first the chicken and egg with some of these metrics.

  • So watching the Shames v Bolts and in Game 44 in the last minute of the 1st period I see Stamkos laying out blocking a shot. A Captain of a team that is 22 games over.500. He is also is second in the league scoring. It wasn’t a cannon but it it was a chance and sent a message to his club.

    I don’t hate Hall, thought he got shafted when Ference was offered the “c” . No one hated losing more than Taylor. At the time we needed to make a trade for a d- man he had value .

    Eberle should have been sat in the play- offs . Not cause he wasn’t scoring cause he was sending a crap message. I saw he dropped the gloves in Brooklyn the other night and was shocked. No question a gifted, offensive hockey player but even if he scored 50 I think team Canada would not invite him to camp.

    • OilersGM

      I disagree with you.
      Hall never should’ve been traded. All the years of pain and you just wash it way in a blink of an eye. Teams built around the core that’s how you win not give them away case and point this year Oilers .One thing I do agree with you about is no one hated losing more than Hall.

      • I didn’t agree with trade at the time, didn’t know much about Adam but last year i thought he played well and the young players around him were getting better. This year the defence has regressed ( besides Nurse) and add we can’t score equals the most negative blog in sports.

        Relying on kids to turn this team into a real NHL team is always going to be a roller coaster. Another player I’ve liked this year is Nuge, becoming stronger and a veteran on this team. I get sick to my stomach thinking of losing him in spite of Lucic. Does so much for this team that goes unnoticed.

    • The Whispererer

      Facts are stubborn things. Not including his Junior exploits, Eberle represented Canada and played an integral role at least 5 ( might have been 6 ? ) times in the IIHC World Championships. Maybe you think he invited himself.

      • I mean no disrespect to Eberle, I watched the play offs last year and didn’t like what I was seeing. Had him in a draft and keeper league and really just watched him. I don’t understand the trade in the summer when his value was at a all time low. I hope there was communication and that was the decision.

    • camdog

      Do you mean top 2 on this Oiler team this season or legitimate number 2 man in the NHL? Two completely different animals. His game has gone downhill a bit now that he’s playing with Larsson was better earlier in season with Russell. The difference is equivalent to millions on his next contract.

      • a lg dubl dubl

        Top 2 on the Oilers. Agreed that Nurse seems to play better with Russell, than with Larsson.

        I’m hoping Nurse signs for around 3mil long term. Even though he’s been probably the Oilers best dman this year, he still has to grow his game a bit more(coaching?) and at around 3mil it could turn into a value contract in another year or two.

  • puckle-head

    If anyone wants to know a REAL terrible stat, I’ll give ya one!
    The player Peter Chiarelli bought out has more goals than the player that he traded Jordan Eberle for. But it’s okay! Because if you combine those two players goals together, they have one more goal than Eberle. Why can’t we score again?

  • Rebuilds30

    Do you realize this team could be icing Three lines coming at you such as..
    Maroon-McDavid-Draisaitl
    Eberle-Nugent Hopkins-T Hall
    Puljujarvi-Calligula-Kassian

  • aye

    @Henderson
    While I am also not a fan of +/-, I do have to disagree with your argument for it being a completely useless stat.
    First off, you are making it sound like that people are only looking at +/- stat when judging a player, which you know is not true. So why are you even arguing the usefulness of this stat in the first place? Everyone knows it’s just one of many stats and only serves as a reference.
    You are right that a (-) player does not mean he is bad defensively, but have you actually done a ranking of all the players based on their +/-, and a ranking based on 5v5 goal differential as you are advocating, and compare the 2 rankings? I’m willing to bet that you’ll probably find that the rankings reflect very similar results, rendering your argument meaningless. Just like loser points, everyone complains about them, but at the end of the day, the standings don’t really change much using a different point system. You may have a couple of wild card spots change hands in the odd years, but overall, the good teams still end up on top, and bad teams remain in the basement. Same here, regardless of which stat you use to rank the players, the elite players will be elite players, and bad players will show poorly.
    I totally agree that +/- don’t tell the whole story, but it’s true for any stat, should never be taken at face value, and should always be used in conjunction with other stats to see a more complete picture. I’m sure the stat community, team management, and fans in general are smart enough to understand that, and don’t need you to go on ranting about which stats are useless, and which stat should be used. Because the people that get it don’t need you to tell them, and the people that don’t get it aren’t knowledgeable fans, and then there are fans that really don’t care about this technical stuff and just want to enjoy the game for what it is.

  • BringtheFire 2.0

    “If your goalie is pulled at the end of the game, it isn’t because you’re already up by three and you really want to rub it in Calgary’s face.”

    Makin’ you look at Chucky and T-Bag for that.

  • Theoboxer

    Enough Taylor Hall!! Taylor Hall was a very one dimensional player on the Oiler…. Zero defense on a team that was crying for defense. Yes Taylor Hall could score goal but that was it.

    • The Rookie

      Damn it Taylor, stop scoring goals and just hang out in front of our own net to get in the way of their shots. 🙄. Why do we criticize fwds for not being Defensive and D men for not getting enough points??

  • jultz=2cups!😂😂

    So if the stat is just malarkey according to all you “experts”, why is it one of the main stats that the nhl keeps and publishes for all to see? Just to rile up guys like Matt Henderson? A player is on the ice for more goals for than against. Hockey in its simplest form.

    • The Rookie

      Probably because it was one of the original stats kept by the league and to remove it now or change it to 5×5 dif would mean the league was wrong. And that clearly never happens

  • The Rookie

    I knew short handed goals against were a – but I didn’t know that EN goals were also. That is a huge misinterpretation of number. I assume these were some of the worst case scenario to help get the point across and that mostly bad teams are affected by this, but I get the point. Player values on bad teams get perceived as being lower by fans because of these numbers. Like scientific studies…all depends on who is paying for the research.