Dorkapalooza – A look at the MIT Sloan Sports Analytics Conference



The Sloan Sports Analytics Conference, affectionately dubbed Dorkapalooza by ESPN’s Bill Simmons, is an annual student-run hockey stats seminar hosted at MIT in Boston. It was founded in 2006 by Daryl Morey of the Houston Rockets and Jessica Gelman to “provide a forum to discuss the increasing role of analytics in the sports industry.”

Advertisement - Continue Reading Below

This year in was attended by about 2,200 people, including students, professors, front offices, reporters, scouts, agents and fans, of which about 300-500 attended the hockey-specific programs, among them Chris Snow of the Calgary Flames.

The Cutting Edge…Err Not Really

While you might be on the edge of your seat waiting to hear the exciting ways that the league is using incredible statistics to improve how our beloved sport is played, let me encourage you to snuggle right back into the bucket of your chair – nothing to see here.

Advertisement - Continue Reading Below

In order to have a discussion about the latest developments in hockey analytics, you’d want a panel liberally sprinkled with analysts on the cutting edge, like Gabriel Desjardins whose Behind the Net site is the sole provider of the most advanced and relevant statistics and from accessible advances like drawing penalties and zone starts to the more advanced shot-based metrics and Quality of Competition; or like Vic Ferrari, innovator of many of the aforementioned statistics, and whose site is essentially the only way to collect specific situational shot-based metrics for teams or players. Combine analysts like them with the Billy Beanes of NHL front offices, and you’d really have something exciting.

Instead we got Brian Burke and Mike Milbury.

While hockey has been notoriously slow to adopt analytics, there are still roughly a dozen NHL teams at some stage of using advanced analytics, according to Pittsburgh’s Dan MacKinnon, of which eight sent representatives to this conference.  Unfortunately Toronto is clearly not one of those twelve teams, with Burke ridiculing statistical analysis in his opening speech, and scoffing at everything even remotely analytical that accidentally arose throughout the day.

Now the official public position of most NHL teams is that statistical analysis is useless in hockey, but there’s some debate about whether they’re just saying that to avoid tipping their hands, or whether they’re genuinely stupid. Having personally worked with NHL front offices who were eager to soak up even the smallest scrap of analytic knowledge (one who even fired their coach later that week –coincidentally I trust), and knowing several analysts who have been asked for far more (there are quite a few teams who have Desjardins on speed-dial), I’ve always believed that it’s more a case of the former than the latter.

Where are the Calgary Flames on this? They recently hired Chris Snow, and he was in attendance at the conference, but that’s all we really know. We have reached out to him but to my knowledge he has yet to engage any of the established local statistical analysts. While apparently not a league leader, at least they’re off to a start.

The Edmonton Oilers seem to be a bit further ahead in this Battle of Alberta, with Snow’s approximate counterpart Nick Wilson having invested in a five-person hockey analytics team of their own, and having brought Assistant GM Ricky Olczyk with him to the conference for a closer look this year.

Advertisement - Continue Reading Below

Where To Go From Here

Back to the Conference, which fortunately had at least some of its value salvaged with the inclusion of one statistical hockey analyst, St. Lawrence University’s Michael Schuckers, who you may know from Defense-Independent Goalie Ratings (DIGR) – and he was just a last-minute addition. 

In fairness Desjardins was reportedly invited, but declined due to lack of financial sponsorship – perhaps that was the case for Ferrari and others as well. Can’t some NHL team spring a few grand for a flight and hotel and effectively double the value of the panel?

Hockey analytics obviously draws its inspiration from baseball, and Moneyball was essentially about how you could use analytics to get your teams small advantages that ultimately add up to more and more wins. It works by quantifying value for various game events, finding which ones have the biggest impact, and moving teams in those directions. Just as baseball teams improved by finding players who could get on base, today’s hockey teams could get an edge by finding players who tilt the ice in their favour possession-wise, in a variety of ways.

For example, Schuckers presented his studies on winning faceoffs, which get you 1 goal for every 100 face-off wins.  According to the 3-1-1 rule (3 goals gets you roughly 1 point in the standings and is worth roughly $1 million), a marginal player like Jeff Halpern can earn much of his salary at the faceoff dot alone.

Unfortunately most of the conference was derailed with nonsense discussions, like how a player’s weight is allegedly the most important analytic factor in determining whether a draft pick will be a bust (are you listening Iain Fyffe and Corey Pronman?) and how much plus/minus sucks – a debate that’s only worth engaging in if you have a DeLorean. While it’s interesting that, according to Peter Chiarelli, Boston’s using a new version of plus/minus that incorporates scoring chances, that’s something that would only be useful to Roger Nielsen if you sent it back in time at least 40 years. 

Advertisement - Continue Reading Below

Sadly Schuckers could be right that it’ll be at least ten more years before advanced hockey analytics are adopted in hockey.  In fairness, at least the sport is off to a start, and there’s some real long-term potential. The Calgary Flames would be wise to join Edmonton as an early adopter, and build an analytics team of their own.

  • RexLibris

    Great read Robert, thanks for the perspective on the conference.

    I had already read some of Burke’s comments about analytics and I am inclined to believe that he is both a.) that stupid and b.) trying to be that cunning. In fact, when I saw the picture of Max Smart and read of Burke I pictured him standing at a news conference telling reporters that they missed the playoffs “by THAT much”.

    I have read recently (might have been Staples’ Cult of Hockey) about the Oilers investing heavily in hockey analytics as part of the new direction under Katz. The strategy being “spend a few hundred thousand here and turn that into a few million somewhere else” which is the kind of entrepeneurial approach that the Oilers have had very rarely in their existence. It made sense to me, a few years back now, that with the massive online community that offers up it’s time to breaking down the mathematics of their favourite teams, and with a higher-education institution readily available in town it would seem negligent not to take advantage of those resources.

    It’s a shame that there didn’t seem to be any discussion of significance at the conference, however I wonder if perhaps a conference would have to be help without the explicit involvement of any NHL teams for anyone to really open up about what they are studying. With the parity and competitiveness of the league it’s almost like expecting a conference of national security advisers to open up about how they assess and gather date from other countries. Not going to happen.

    With the Flames having a windfall of roster positions available this summer relative to previous seasons it would behoove Feaster to use every means at his disposal to determine which players are best re-signed and which FAs are best pursued in a complementary role.

    Were I a Flames fan I would be worried about his spending too much time hunting whales when it’s really just a modest fish dinner that the fans would like.

  • Maybe next year they will invite Don Cherry to speak?


    • RexLibris

      Whoa, Kent. Did you use some sort of psychic channeling mechanism there? That was creepy. I could actually see the jacket, high collar, goatee and thumbs up all while reading that.

      I know this is a little off topic, but I will give Cherry credit for one thing: a few years back, maybe five, he started harping on about elbow and shoulder pads being made like medieval armour and comparing them to the leather pads from earlier eras. His argument was that, basically, if you want to keep a player’s head on his shoulders switch to softer pads.

      Back to the topic at hand: if they invited Cherry to speak at that convention it would almost be worth the price of admission. Do they sell popcorn at those things because it sounds like it’d be some great entertainment.

      • Cherry occasionally speaks sense. I’ve liked his points on equipment, icing and the placement of the benches in the past.

        As for his spectacle, I’ve actually gotten to the point where even that doesn’t hold any value for me anymore. I don’t bother to watch coaches corner, even out of morbid curiosity. It simply doesn’t interest me.

  • RexLibris

    I attended one of the season ticket holder luncheons today and asked king about chris snow’s position and influence. fwiw, king claims the flames are top-5 in the league in terms of implementing advanced statistical analysis.

    • loudogYYC

      I suspect if someone asked Ken King where the Flames were, quality of team wise, in comparison to the other teams in the league, he would come up with a long, convoluted and opaque explanation how the Flames were “top-5 in the league”.

      Then Feaster would threaten trade all the veterans if the team didn’t shape up.

      And then I would wake up to hear “I got you Babe” playing on my alarm clock again.

      • RexLibris

        Okay, maybe I’m just being reactionary here but that statement, in my opinion, is complete hogwash. From everything I have ever read about every team’s approach to and communications about, their advanced statistical analysis team, they wouldn’t tell you if they measured the temperature. So how on earth is King supposed to know where the Flames rank in order of advanced stats out of 30 teams in such a secretive fraternity?

        I love Jeff Lebowski’s idea of putting a grid pattern on the ice. With a new rink in Edmonton it would be a very interesting idea to set up a series of cameras to get a bird’s eye view that essentially broke the ice surface down into a grid pattern so that, at least at home, the Oilers could break down the areas of play through video as well to add to the depth of analysis, or help facilitate it.

  • Jeff Lebowski

    I think the technology has to improve so that truly interesting questions can be asked.

    If for example, the ice surface was put in a grid so that you could determine (with some kind of gps or motion tracking):
    -where the shot came from
    -what was in between the puck and the goalies (screens)

    This would qualify the goaltending stats. Right now they suck because not all goalies face the same kind of shots. How many goals are scored of deflections? A deflection is a shot on net, same as if it came from centre ice in the current stat perspective.

    When Kipper’s facing an odd man rush, what’s his saver percentage when there is no passes and just a shot? What happens when there is 1+ passes and he has to be athletic to make the save?
    How does that compare to other goalies? Also important, what is performance like in context of game situation? Key saves at crucial moments? Does some dude put up big numbers that are kind of meaningless goals ie 6th goal in 7-2 blow out or do some guys score when the team needs it most (tied? late in game? after the other team has scored? after his team has been massively outplayed and outshot? etc)

    That’s how the Mike Smith, Ilya Bryzgalov, Brian Elliots get qualified. Not by GAA and SV%

    How many more goals are scored off the rush compared to dump and chase? How much better is a team’s +/- if they dump and chase more than a puck possession team? Maybe who we think are puck possession teams actually dump and chase more to score goals.

    The science needs to innovate so that more intersting data is collected. It SEEMS that old stats are just being formulated in novel ways rather than calculating new stats.

    This has been done in baseball most effectively because there are many discrete events (fixed aspects of the game) rather than the fluidity of hockey.

    It’s definitely the way of the future but once they are able to quantify and qualify all the info with better technology it will give more focused info. IMO.