CAPTAIN VIDEO AND FANCY STATS

Roger Neilson

Having more information to draw on when you’re building or coaching a NHL team is never a bad thing, be it in the form of grainy VHS videotape or advanced stats and analytics.

Almost 40 years ago, Roger Neilson was hailed throughout the hockey world as an innovator and a pioneer for using videotape as a coaching tool. Breaking down tape of games he’d just seen with his own eyes earned him the handle Captain Video. Cutting edge stuff, it was.

Decades later, with videotape long having long given way to the digital era and teams employing full-time “video” coaches for years now, Neilson’s “innovation” is as much a part of NHL hockey life as sharpening skates and taping sticks. Doing so isn’t news. Not using it is.

In that regard, employing advanced stats and analytics as a tool to provide a more complete picture beyond old-school boxcar numbers is today’s version of videotape. “Fancy stats” have been in use to varying degrees around the NHL for some time, but not to the point where the hiring of Kyle Dubas as an assistant GM by the Toronto Maple Leafs this week wasn’t news.

Five years from now, a hiring like this will be a sidebar.

INFORMATION AGE

UnderwoodKeyboard

I’m not a progressive guy. At 55, I’m a product of my generation. Hell, when I graduated journalism school we were using typewriters and carbon paper in the classroom, although there were bread-boxed sized laptops being used at the Pacific Coliseum and B.C. Place when I started covering the Vancouver Canucks and B.C. Lions in the early 1980s.

Advanced stats? I’d covered baseball, where statistics had long been in broader use than in hockey, with The Edmonton Journal from 1992-96, but when I was asked about advanced stats as part of an interview with the blog Oilgasm in March of 2008, I pretty much blew them off.

Q: On the Oilogosphere, many references are made to advanced hockey statistics such as EV/60 (Even-strength points per 60 minutes), PPP/60 (Powerplay points per 60 minutes), EV+/EV- (+/- after filtering out empty net goals for/against situations). Do GM’s take these numbers into account, or do they take more of a conventional “Eye-based” scouting approach when negotiating contracts and signing free agents?

A: “I’m not the least bit interested in these numbers. I know what I see and I know what I think. I’ll go with that over pages of statistics any day. As for GMs, that’s a broad question. I suspect there’s a wide range in answers for that.”

Hardly a warm embrace of advanced stats and analytics, although in the context of covering a team day-to-day as a beat writer with daily deadlines as opposed to building a team or compiling data to look more deeply than “who’s hot and who’s not,” I hadn’t delved into them.

Now, seven years removed from the daily grind of the beat, I find myself considering the merits of advanced stats far more than I did when they actually might have helped me get a clearer picture of what I was seeing and writing about. In that, I’m clearly not alone.

HERE AND NOW

The use of advanced stats has come a long way even since I was asked about them in 2008. It’s an evolving field of study with refinements being made on an almost daily basis. There’s a long way to go to sort the meaningful from the meaningless, but teams committed to doing so already and those in the process will have a leg up on stragglers who don’t.

There are a lot of people, many of them right here in Edmonton, doing the kind of work that pushed Dubas into the spotlight with the Maple Leafs this week. The Edmonton Oilers are among the growing number of teams taking advantage of that brain power. There are more hires to come.

Separating the useful from the useless will play itself out in good time, likely sooner than later. What has merit will stand, what does not will fall. There is more information to be had, and that’s a good thing. Neilson understood that 40 years ago. Some of us are just getting it now.

Listen to Robin Brownlee Wednesdays and Thursdays from 3 p.m. to 5 p.m. on the Jason Gregor Show on TSN 1260.

  • A philosopher once said “there are lies, damn lies and statistics”. If a person wants to, one can make statistics back up almost any argument one might want to put forth. We see that on this site on a daily basis.

    Statistics are a tool when used properly can be helpful, but are only tool. Like any tool you need a skilled person to get the desired result.

      • ubermiguel

        “In that regard, employing advanced stats and analytics as a tool to provide a more complete picture beyond old-school boxcar numbers is today’s version of videotape. “Fancy stats” have been in use to varying degrees around the NHL for some time, but not to the point where the hiring of Kyle Dubas as an assistant GM by the Toronto Maple Leafs this week wasn’t news. “

    • That philosopher you quote is generally recognized to be Leonard Courtney, the same Leonard Courtney who was President of the Royal Statistical Society, a professional body for statisticians.

      It was popularized by Mark Twain, who made it pretty clear that even someone like himself, who was ‘often beguiled by figures’ could tell the difference between senseless and sensible statistics with a little bit of thought.

      • I’m not trying to say that only a very few people “skilled in the art” can get some meaning out of looking at stats. Although the more complex stats become the more this will be the case.

        Rather if a person comes in with a conclusion in hand and only looks for stats that support his conclusion, he can probably find something.

        EG If Taylor is good player? No way!!!! he lead the league in give-ways. Don’t say that he was also 4th highest in take aways.

        • justDOit

          And people who try to use one stat to prove their point are very quickly exposed.

          As in your example, give-aways are tied to puck possession – you can’t give the puck away if you don’t already have it. Puck possession is generally regarded as a good thing, and you’ll find that the top ‘give-away’ artists in the league are also the top players in the game. So by digging a little, you can actually use that person’s argument against them.

          So arguing that stats can be used for whatever purpose the presenter has in mind, is ridiculous. Interpretation of all the data soon reveals the truth.

        • That’s a fair comment in a lot of ways.

          The one thing I’d add is with regard to the really complex stats. I wouldn’t trust something too complex to be explained to the average fan – if there’s value there, it should be demonstrable.

          • Serious Gord

            Then why did the leafs hire the guy? Complex stats, just like complex engineering formulas are highly beneficial in the right hands – you trust them every time you turn on the lights or start your car.

            But in both cases it is the trustworthiness of the analyst that matters – and that’s the point of the quote of Leonard Courtney.

    • Confirmation bias is a huge problem in the stats world. There’s something greasy about a writer who has a theory and then proceeds to manipulate the stats to fit the theory.

      Dellow’s new article on the Leafs and the Kings was well done though. But it is definitely not the norm.

      There’s just too much info out there to know who has it right.

      For example, on one hand you have people saying that corsi and scoring chance differential would be comparable over a long period of time, on the other you have Staples saying J.Schultz’ scoring chance+/- was the best of all the D on the team, yet his corsirel was brutal. Not sure how unbiased and objective his counts are for contributions to scoring chances, but that goes for any person tracking any stat.

      It’s nice to see that the hockey world is finally starting to use more in depth numbers though.

    • The Last Big Bear

      Oh baggedmilk, you can’t even adequately moisten a serving-sized box of Froot Loops without getting milk all over my grandma’s coffee table.

      You’re not fooling anyone about your ability to do math.

  • Jayz

    Question. Hockey is a team sport. Players individual stats are more of a result of a teams overall on ice strategy. Meaning a players corsi isn’t so much a reflection of his performance per say but the performance and execution of a superior system put
    in place by the coaches and executed properly by the players. Case and point Clarkson had a decent corsi on Nj and terrible corsi on the Leafs.

    I think fancy stats are great but when you start losing sight of hockey being a team sport driven by 5 players working together to get the tide pushed the right way we lose the meaning. Corsi is not any indication of any one players skill IMO. Corsi Rel and quality comp stats are. No?

    • ubermiguel

      With a large enough sample size and where a player plays with many different people we can start to see patterns. Willis’ amazingly titled article about Michal Handzus is a case in point. No matter who Handzus plays with the team performs worse. It’s a team sport, but the team is made up of 6 individuals on the ice and each one of them impacts results. Clarkson’s decline is likely due to age and being placed in the wrong role.

      • And isn’t that the rub, it’s not the stats themselves, but the meaning people get from the stats. And that can be subjective. With human nature being what it is people tend to see what they want to see. Some people are better at remaining objective than others.

        As in your example you have came to the conclusion that Handzus is just not good at the NHL level. But Clarkson’s decline may be due to other reasons. It’s not the stats themselves but the meaning you get out of them.

        • ubermiguel

          Stats are just one way to interpret reality and create meaning. The thing about any stat is they usually require more investigation. Willis did a great job with Handzus and I will agree with the conclusion that he is no longer an NHL calibre scorer (age 37 and low scoring were are also factors).

          For Clarkson I was just repeating a few possible theories that explain the change in Corsi. The question then becomes “how do I test my theory?” We could look at TOI, quality of competition, injuries, quality of line-mates etc.

          • Stats are just one way to interpret reality and create meaning.

            This is patently false. Properly collected statistics are not opinions. They are not relativistic interpretations. They are recorded factual observations of actual real events.

            That you may use these facts to reach an incorrect conclusion is your fault, not the fault of the statistics.

            Consider, for example a car accident insurance investigator. The investigator notices after hundreds of incidents that in 98.7% of the cases there are police cars at the scene. She then concludes that police cars cause accidents and procedes to recommend a policy of banning police cars.

            The number of police cars is not in dispute, not up for debate, and most importantly not relevant to the goal at hand.

          • ubermiguel

            You are using statistics incorrectly. Statistics and models are a simplified description of reality, designed to yield hypotheses about reality that can be tested. In your example we see this glaring statistic and you came up with a theory: police cars cause car crashes. We investigate that theory somehow (e.g.: observing a sample of police cars, interviewing car crash participants, reviewing available video) and discover that theory to be wrong and we have to come up with a new theory: police cars arrive after crashes.

            Change “police cars at accident scenes” to “Clarkson’s declining Corsi” and you might start to see the value in the diagnostic and predictive ability (as opposed to merely descriptive) of statistics in hockey.

            I’m saying that seen-it-by-eye and advanced stats both are appropriately used to understand hockey. Who understands the reality of flying to the moon better: the mathematicians that calculated the force and trajectories or Neil Armstrong? The answer is neither, they both interpret the same reality using different tools and experiences.

      • Jayz

        So my question was corsi rel/ qualcomp is the stat that best depicts a players impact on any given system, and corsi is a stat that has more to do with the coach/system?

        • ubermiguel

          Ah, I see what you’re getting at. They’re both the same in that regard. Any Corsi stat is best used to compare players within the same system. Qualcomp adjusts for the quality of opponents faced, but is still based on Corsi.

          If Willis had any clarification I would love to hear it.

  • I have no problem with alternative methods of analysis like video and statistics in sport. I DO have a problem with some blogger’s pompous narrative that all writers before them are basically dinosaurs, positioning themselves as the new guard despite having not a whiff of real-world experience in professional sport. It comes across as massively insecure, like they’re the cool kids now. As if some high school math classes, a cadre of like-minded followers and a website suddenly elevates them to credibility that was earned the hard way previously.

    But guys using advanced tools with a solid understanding of pro-level sport (beyond the pile of books they read on the subject)? Yeah I’m all over that.

  • northof51

    Cheers to RB for warming up to advanced statistics.

    I think people are missing the point a bit if they think that advanced stats are the antithesis to the old school, “seen him good” judgements. As mentioned, stats, when used properly, are just another tool for player and team evaluation. You can’t ignore other factors as well, including the intangibles.

  • The Last Big Bear

    There’s no one stat that best describes a player’s performance.

    Vollman’s player usage diagrams are a very good start for forwards. I don’t really think there’s much benefit when evaluating a forward in going beyond his box stats, his PP/PK/EV breakdown, and his Vollman diagram. Corsi Rel is good too, but loses a lot of value if the guy plays for a really good or really crap team.

    I think the best metric for judging defencemen is “Time On Ice”. This is obviously just a way of saying stats are no good for defencemen, and deferring judgement to an expert (ie an NHL head coach). I don’t think there is any advanced stat that provides information above and beyond a defenceman’s ice time breakdown, and maybe his points totals.

    I never understood people using Wins and GAA to assess goalies. EV sv% pretty much covers goalies, I think.

    • Reg Dunlop

      So, grading a tender on whether he wins or loses makes no sense to you? I admit, I always thought that winning was the purpose of sport. I STAND CORRECTED.

    • Serious Gord

      EV sv% helps but certainly isn’t perfect – the way defensemen play in front of a goalie can have a huge impact.

      And some goalies are far more clutch than others – dubnyk versus cheevers.

  • Spydyr

    You can break the game down as much as you like.Say he scores on more two on ones on a Tuesday when the moon is full than any other time.

    There is no stat that shows a guy took a puck to the face and came back to score the winning goal.

    At the the end of the day the only stat that matters is did you score more goals than the other team at the end of the game.

    • Serious Gord

      In baseball – which is essentially a group of individual athletes doing individual tasks – statistical analysis Can be highly effective at determining who is the better player. And that has been the case for decades. And as the advent of computers and more granular stats has grown, so too have the stats and the accuracy and preciseness of the analysis.

      That has now spilled over into hockey which is much more of a team sport where the actions of one or more players impacts the success or failure of other players and there are very very few repeated events as compared to baseball where eighty or ninety percent of the time a pitch is not hit into fair territory. Thus the accuracy and preciseness of the analysis will never be remotely as good as in baseball

      That said the analysis has become very good at separating good players from bad and is getting better all of the time. As a strategic tool they are already proving their worth and tactical benefits are next.

      A war analogy would be dropping regular iron bombs from a plane over a target using a bombardier and an eyesight versus using a laser pointer and a smart bomb released 60000 feet in the air fifteen miles away from that target.

      The nhl has moved into a tech war era. Teams that don’t adopt the latest statistic analysis are going to get beaten more than those that do. I have no idea where the oil are in this war but if I understand stauffer correctly, they definitely are making changes in the right direction.

      And fans who wager without the benefit of these new technologies are going to lose a lot of money.

    • ubermiguel

      On a per game basis that’s true, but over the course of the season it’s important to figure out whether you are winning based on your play or by getting breaks.

      Take the leafs last year for example. They won a bunch of games early and everything was great in terms of the standings. There were a bunch of people calling it a house of cards, but the Leafs were convinced they found the magic formula that had eluded teams like Minnesota, Colorado and Dallas in the past.

      Instead of recognizing their short comings, they continued on and they fell apart (much like the 3 previously mentioned teams).

      So yes, at the end of the game scoring more goals than the other team is what matters, but you have to be able to do that over the course of a full season.

  • Serious Gord

    Until I see results in the Oiler organization
    that can be attributed to advanced stats then I remain unconvinced. When you have incompetent and/or inexperienced coaches and management you introduce variables that can affect player outcomes.

    Case in point, certain bean counters said Hemsky was dead weight and was effectively done. He gets traded to a better team with a topline center and suddenly gets productive for another organization.

    Advanced stats in the wrong hands are the next thing to reading tea leaves and soothsaying. This is not something I’m willing to take on faith. Show me….

    • Serious Gord

      I and many others wanted hemsky gone – not because he wasn’t a good player but because his skill set was in surplus on the oil and this he was being out o. Lines and given ice time that was inappropriate/mismatched to those skills. Once traded to a team that had a need to him his play and stats blossomed (for now, I do question how long he can sustain it).

      But you do touch on an aspect where stats in hockey are not as nor can they be as definitive as they are in baseball – the impact of the play of other players on ones play is far greater in hockey and thus that has to be taken into consideration when making conclusions.

  • Quicksilver ballet

    I think we can all agree, these personal numbers assigned to each player (advanced stats) are still really team based rather than player only based. Hemsky could literally play on 30 different teams and be portrayed in 30 different slants.

    Who knew Spezza playing with Hemmer would trump 83 playing with Gagner. Nothing to see here folks……move along.

    I still feel this is a make work project, created by the Canadian Government to keep unemployed hockey fans working/off welfare. They should turn these mathletes loose on the airline industry….. help reduce that multi decade old flying through the skies while seated in a chair experience.

    *Cool fact. The lines painted on the roads heading into Toronto (hwy 400) are Oiler orange. They are obviously still not aware of the problems that have plagued Oil Country since the early 90’s. I commend them on their sticktoitiveness.