• The Big Mistake

    by  • April 9, 2014 • Hockey • 22 Comments

    If someone asked me what I think the biggest failing of the eyeball test is, I’d respond that it’s the emphasis on the big mistake. There are gigabytes of information contained in a hockey game. So much information that I think it’s difficult for anyone to take it in and organize it rationally. The way that our brains deal with that is by focusing on the big mistake.

    What is the big mistake? The big mistake is the play that leads to a goal against. When we see a player who’s made a bunch of big mistakes in a row, we get down on him. There are two problems with this. First, big mistakes don’t end up in the net on a clockwork basis. If for, example, one big mistake in four ends up in the net, it’s not like it goes goal, no goal, no goal, no goal, goal, no goal, no goal, no goal, goal, no goal, no goal, no goal, goal, no goal, no goal, no goal, goal, no goal, no goal, no goal. There will be clusters of goals and clusters of no goal.

    Second, focusing on the big mistake ignores the context in which the mistake happens. A team with a 55% Corsi% has fewer opportunities for its players to make the big mistake than a team with a 45% Corsi%.

    Steve Simmons went on TSN 1050 this morning and fell into this trap. He was asked about the Leafs lousy numbers and had this to say:

    If the current administration is in charge, they pay next to no attention to, whatever you want to call the analytics people, to that realm of hockey. There are people who watched last night’s game, I’ll give you an example, and say ‘See? It’s all because of puck possession.’

    That seems to be the new term for everybody that really doesn’t know the game. They’re great on talking about puck possession. What you get into though, look at last night’s game. They get a goal scored because the defenceman backs in and a goalie who can’t stop anything high glove. So it’s 1-0.

    The Leafs had three power plays prior to that and didn’t score on any of them. What in God’s name that has to do with puck possession or Corsi or Fenwick or anything…their power play is lousy right now. That’s got nothing to do with any other statistic other than power play percentage.

    Then they take a penalty and get scored on seven seconds in and essentially the game’s over. So you’ve had a mistake on a d-man backing in, you’ve taken a penalty and got scored on seven seconds in to your power play, where does your puck possession that everyone’s obsessing about come in to this?

    Some will say sample size. That’s the other great term they use. Well, sample size is 82 games. If you break down each game individually, you’ll see puck possession plays a factor sometimes. So does defensive zone coverage, so does turning the puck over, so does penalty kill, so does power play, so does goaltending.

    You can right now take a laundry list of what’s wrong with the Maple Leafs and you can go through many, many things but to continually point to these, for my mind, statistics that have predicted this doom and gloom and that everybody’s gleeful about, well the same statistics predicted for the Colorado Avalanche and they’re going to the playoffs and they’re home to Chicago. So, explain that to me.

    The Colorado Avalanche, as I’ve discussed previously, are piling up points in an unsustainable way. They will either get better and earn their points, or they will probably get theirs next season, just like the Leafs did this year. I expect that Chicago will wipe them out in the playoffs but, small sample size, anything can happen.

    Two portions of what Simmons said caught my eye:

    If you break down each game individually, you’ll see puck possession plays a factor sometimes. So does defensive zone coverage, so does turning the puck over, so does penalty kill, so does power play, so does goaltending.

    Two of those things – defensive zone coverage and turning the puck over – are directly related to puck possession. The better you are at those things, the better your Corsi% will be, all other things being equal. The Leafs goaltending was, on the whole, excellent this year. If it cost them specific games, who cares? Goaltending costs everybody some games. It cost the Leafs fewer than most.

    The Leafs are twelfth in the league in 5v4 goal difference. They’re better than average. Are there games where the power play didn’t score? Sure. As there are for every team. So there were fewer for the Leafs than there were for most.

    The penalty kill has hurt the Leafs. Their -51 goal difference when shorthanded is 11 goals worse than the league average. That’s on par with their 5v5 play, which has been 12 goals worse than the league average. Of course the reason that people focus on the Leafs 5v5 possession is that they’re great at finishing 5v5 (8.6% S%, fourth in the NHL) and very good at stopping the puck at 5v5 (.927, tenth in the NHL). If you’re great at finishing and good at stopping the puck but overall you’re bad at 5v5…well, there’s one thing left.

    Which brings me back to the big mistake.

    What you get into though, look at last night’s game. They get a goal scored because the defenceman backs in and a goalie who can’t stop anything high glove. So it’s 1-0.

    There’s a reason that “Hockey is a game of mistakes” is a hoary cliche. It’s because it’s true. The more opportunities you have to make mistakes, the more mistakes you will make. No team in the NHL has more opportunities to make mistakes in the defensive zone than the Maple Leafs. That’s the underlying cause of this rot.

    The Leafs can attack the underlying issue and try to become a 50% Corsi% team on their way to being one of the league’s elite. Or they can try and find guys who make mistakes 14% less often than the average guy so that they can survive as a 43% Corsi% team. Either way focusing on the big mistake misses the easiest way to stop suffering from them: be in a position to make them less often.

    Email Tyler Dellow at tyler@mc79hockey.com

    About

    22 Responses to The Big Mistake

    1. Punk Anderson
      April 9, 2014 at

      Tyler,

      Great post.

      I am a consultant. I work with management personnel in corporations, who will regularly “shelve” data and concepts that would make their business more efficient and profitable. They bury the work because they did not generate it, and it might make them look incompetent for not coming up with it, or worse, look outdated and unable to keep up with a changing industry.

      Surely, Dave Nonis knows that the Leafs should “try to become a 50% Corsi% team on their way to being one of the league’s elite.” He would never admit this publicly, for fear of ridicule. If he does not have the pragmatism to incorporate what #fancystats shows in a non “gut feel” way, then Tim Leiweke should strongly encourage it or find a replacement.

      Oh, include Steve Simmons and Damien Cox, and other “dinosaur, stale MSMers” as those threatened by what hockey analytics has revealed about the game they are “experts” at through observation.

      “It is difficult to get a man to understand something, when his salary depends upon his not understanding it” – Upton Sinclair

    2. mike m
      April 9, 2014 at

      Tyker,

      one of the things we look for in tryouts is who does the possession end with? Who was the last player to have control and how often is he the guy. This does not count shots or shot attempts. looking at bad passes, giveaways, over skating, trying to go 1 on 3, losing board battles, etc. It would be interesting for someone to track last guy with the puck vs. some other possession stat(ind and team)

      • Scott
        April 11, 2014 at

        Now that take-aways are a common stat, why aren’t turnovers? I’m sure they are being kept (or I’d hope), but they are never referenced for an individual, I seem to only ever hear them for a team.

    3. Greg
      April 9, 2014 at

      Great article as always.

      I think the problem that a lot of people have with corsi, etc. is that it doesn’t point to an easy fix. If Dion Phaneuf is the problem then replace Phaneuf and your problem is solved. If it’s goaltending then replace your goaltender and problem solved. However, if the problem is that you can’t outshoot your opponent on a regular basis then the problem is a complex one (as it always is) and can only be solved by addressing everything in concert (personnel, systems, coaching, etc.). This is too overwhelming for most people (although it shouldn’t be for an organization whose entire purpose is to work towards this, namely team management) and is especially inconvenient for a sportscaster who relies on soundbites and rapid analysis for their paycheque.

      • Flips
        April 9, 2014 at

        I think that this is a driving factor in a coaching/player resistance to Corsi, and you saw it in Hall’s interview a while back. Corsi is kind of the end product of a lot of things that drive or lose possession. Isolating what the factors are that contribute or hinder to a positive Corsi is where, I think, there is a chance for more mainstream acceptance. Then you can go back to the player and say “doing X is contributing to your line consistently getting out shot”. From his interview, Hall’s big question to their stats guy, paraphrased, was “Ok, so I am getting outshot, what can I do to improve this situation?”

        I think that Tyler’s work on multiple shot shifts has that potential, as well as some of the work on Zone Entries. Counting who loses possession, relative to how many times that they are in possession, may contribute to that as well. We bemoan the Giveaways stat as useless since the people at the top of the list are the people that handle the puck most often, but comparing that to “pass completion %” or “total passes attempts” might help to filter out the high end defensemen from the turn over machines. Sadly, most of this information is hard to come by, as NHL game states lag greatly behind soccer ones (even MLS).

        • Bruce McCurdy
          April 9, 2014 at

          Rather than bemoaning it as useless, maybe “we” should accept that Giveaways stat actually tells us who handles the puck most often. I find Giveaways useful as indicators, even though the stat itself is compiled of negative events. To me it’s one of those stats that has a way broader context than its original intent.

          • Tom Benjamin
            April 10, 2014 at

            I agree with this, Bruce. I wish there was a reasonable standard for this statistic. Another one that is apparently negative – the number of hits a player takes – is actually positive. A d-man who takes lots of hits has the puck a lot.

            I wish the league tried counting the number of times a given player touched the puck. Never mind what he did with it – just did he touch it. I think there would be positional biases, but I’d be surprised if the number did not reflect player quality – perhaps better than anything else we have.

    4. Tom Benjamin
      April 9, 2014 at

      I agree with what you write, mostly, except the false choice you present in the last paragraph. No team – whether they pay attention to analytics or not – sets out to become a corsi juggernaut. Nor does any team try to find guys who make 14% fewer mistakes than average. I don’t think anyone knows how to do either of those things.

      What teams try to do is find better players. If they do well at that, the corsi will take care of itself. So will the mistakes. If they do not do well at finding better players, they won’t.

      You make it sound like the quants and those who ignore the advanced stats have philosophical differences about the importance of shots. I don’t think so. Dave Nonis doesn’t dispute that. He just doesn’t see how they can be used to help him find better players.

      • Greg
        April 9, 2014 at

        The difference is when an organisation goes to find those “better” players whether they look for people who drive possession or people who don’t make mistakes. And when they’re evaluating the people that they get rid of to make room for the “better” players whether they evaluate them based on driving possession or making mistakes.

        Sure, every organisation may *actually* want better corsi, but if they don’t recognise that’s what they want then they may make trades like Tom Gilbert for Nick Schultz because Nick Schultz is a stay at home defenseman who ostensibly (though not actually) makes fewer mistakes.

        • Tom Benjamin
          April 10, 2014 at

          This assumes that individual Corsi% actually reflects a player’s ability to “drive possession”. If you believe that it does not, that the individual statistic is hopelessly polluted by the team context, then you haven’t gotten anywhere. That’s the real dispute between the quants and those who do not think advanced statistics are that useful.

          It’s easy to argue that Nonis is foolish because he does not think shots are important. It is easy to win that argument by proving that shots are important. That is not the problem. Nonis knows shots are important.

          The problem comes with the arbitrary division of the shot attempt among the five players on the ice or the fact that a shot taken when Steven Stamkos is on the ice is much more likely to go in than one taken when Tom Sestito is on the ice. It also matters whether a 45% Corsi is attached to a player who is out there for relatively few shots for and against or a player who is out there when lots of shots are taken. it is not as simple as Gilbert has a better Corsi so its a dumb trade.

          • Scott
            April 11, 2014 at

            I agree, it’s not that simple. Corsi percentage is rough when looking at someone on the Sabres versus someone on the Blackhawks/Bruins. That’s where relative Corsi can come into play though.

            The thought of the number of shots taken is a little convoluted too, but I don’t think it’s a deal breaker. The goal is to score, which means the goal is to get a shot. Depending on the system of the team (think Torterella vs. Vigneault) some teams will take more or less shots. I don’t have the numbers in front of me but I don’t imagine the relative number of shots from one line to another would vary that drastically since, in theory, they are all playing the same system. That means that relative Corsi could show similar data for a team with a 55% team Corsi just the same as a team with a 45% team Corsi. It’s relative to the system not necessarily relative to the league.

            These are thoughts, I don’t have numbers on it.

    5. April 9, 2014 at

      God, this paragraph is the perfect explanation of why you want to reduce the amount of times that the puck is flung towards your net:

      The more opportunities you have to make mistakes, the more mistakes you will make. No team in the NHL has more opportunities to make mistakes in the defensive zone than the Maple Leafs. That’s the underlying cause of this rot.

      It’s not a difficult concept and yet there are people that will argue until they are blue in the face against that logic.

    6. Jonathan
      April 9, 2014 at

      Interesting read. I agree that the more time you spend in the defensive end the more mistakes you are likely to make. But do the Leafs actually spend more minutes in their own end than everyone else or is it that they just give up relatively more shots. The latter isn’t necessarily telling if the former.

      • Kris
        April 10, 2014 at

        A leafs fan used a stop watch to time it and it lined up well with corsi. There’s also a correlation of 0.9 between zone time and corsi.

        • Scott
          April 11, 2014 at

          Kris, can you post a link to where you got those numbers? I think they are probably accurate, just would like to see them.

    7. Daniel
      April 9, 2014 at

      Man, the way you think the game, the way you see it, incredible. Every single read has substance and meaning.

      This was real good. It sounds so simple and yet not to many coaches can figure it out.

    8. Scrivy
      April 9, 2014 at

      Excellent, excellent article.

      Yes, shots are indicative of time spent in ones own zone Jonathan. More shots on your net = more opportunities for one to go in.

      Tom Gilbert, Dustin Penner. They seemed to make lots of “mistakes”, yet their corsi was good. The eyeball fails. It takes too much work for Mark Spector to check the numbers, he’d rather go off on who he “seen” make mistakes. Lazy journalism.

    9. April 10, 2014 at

      I’m still trying to pick my jaw up off the floor after seeing that Simmons sees “defensive zone coverage” and “turning the puck over” as distinct from puck possession. What does he think happens after you turn the puck over? This is a guy lecturing other people about how they must not watch the games?

    10. chelch
      April 10, 2014 at

      So good… bang on. I learned this very young. I was complaining to my uncle about Aki Berg or another player like him that seemed to make the “big mistakes” all the time, and he said, “You’re just watching the Leafs. Watch the other team’s players too, and other games, then you’ll see everyone’s mistakes.”

      I soon realized Aki Berg wasn’t that bad. Same as how I know Phaneuf isn’t as bad as many people seem to think. If the forwards would help him out sometimes, his play would look vastly improved.

      As far as Nonis goes, I don’t care if he dismisses hockey analytics, at least publicly. There are a lot of people he has to work with in/out of the organization that would alienate him if he started babbling on about Corsi. (Especially when how bad the Leafs are). I just hope the Leafs, with their bottomless pit of money, have their own mathletes gathering data and crunching numbers so Nonis and the gang can make more informed decisions.

      But to be honest, analytics tell you how you’re doing, not what to do. So how much of an impact they have on decision making is questionable.

      • Scott
        April 11, 2014 at

        I’m glad you brought up the part about forwards help. In a system/on a team where forwards don’t help in the D zone, the only people you can compare your defense to are the other defensemen stuck in the same bad situation. (It’s real tough to step up at the blue line without back-checkers, but if you have them and you can step up and prevent a zone entry, other team isn’t shooting – direct effect on corsi.) So now we only have 5 people to compare a defenseman like Phaneuf to… speaking of sample size… just not a good reference.

        In response to the idea that the analytics tell you how you’re doing, not what to do – I see the logic, but if they tell you you’re doing poorly, it tells you to make a change. If it tells you to make a change, they can tell you who has been doing well, and therefore tell you those are people you might want to target (obviously given the role you’re looking for).

    11. Pingback: James Wisniewski and the Big Mistake – Season Review | BS Hockey

    12. Pingback: Between or Beyond: The Curious Case of Grebeshkov’s Final NHL Season - The Oilers Rig - The Oilers Rig

    Leave a Reply

    Your email address will not be published. Required fields are marked *