Q and A with Micah Blake McCurdy

    Micah Blake McCurdy is a Halifax-based mathematician and the proprietor of one of my favourite hockey websites, HockeyViz. He’s also well known on hockey Twitter for his account, @IneffectiveMath, where he showcases new visualizations describing the goings-on of the NHL, including standings projections, player and team performance, shot locations, game flow, and even sadness.

    I’ve followed McCurdy for a long time now, and over the last few years he’s developed valuable, unique tools that can help us to better understand many aspects of the game of hockey. HockeyViz is supported via Patreon and I highly recommend contributing if you like his work. His site provides intuitive visualizations that provide a lot of insight far beyond “the Leafs controlled possession last night.” We can use these visualizations to see how a team has performed over time, how the coach likes to deploy the players, how lines/pairings have changed throughout the season, and much more. I use pictures like the one below to cheer me up when I’m having a bad day:

    Micah was kind enough to answer some questions for me about his views on the game of hockey and his interesting approach to explaining it.

    Who are your favourite players to watch right now and why?

    Mike Hoffman, Josh Ho-Sang, and Vladimir Tarasenko — for unpredictability. Constantly being surprised while they’re on the ice is a real pleasure.

    You have a bit of an unusual job – how would you explain it to someone you’ve just met?

    I tell strangers that I make pretty pictures that help people understand hockey. That normally lets me explain how you can get context more easily from graphs and plots than from numbers, and how that helps you when every interesting thing needs to be compared to something else in hockey to make sense of it.

    Unlike a lot of hockey analysts and writers, you’ve got a very extensive background in pure mathematics. How does this influence your approach to analyzing and visualizing the game of hockey?

    Mostly it gives me tools that let me attack communication problems. Having quite a bit of data-smoothing algorithms in your back pocket, knowing what problems can be solved with computers and which ones can never be, having your imagination be enlarged by your past work — that’s where my background feels valuable.

    Compared to most hockey websites, HockeyViz really doesn’t have many numbers. What’s your motivation behind this approach?

    Numbers are great for precision but what hockey (and every social science) needs is comparisons, and humans make comparisons easily when they’re presented visually. That speed is what makes pictures my preferred way to present information.

    You’ve presented your hockey work at several conferences, and most recently at VANHAC 2017 (Vancouver Hockey Analytics Conference). What were your biggest takeaways from that conference?

    I was surprised to see so many new faces, young faces. My biggest takeaway is that even as the “old guard” finds work that takes them out of the public research sphere, more and more people are joining up.

    If you could add one thing to the list of stats that the NHL tracks in each game, what would it be and how would you use it?

    I would like to see the location of the players and the puck, ideally measures several times a second. Everybody would love this but the specific detail that I would like is head orientation — I’d like to know who is looking in what direction all the time. I could make some beautiful things with that.

    It seems like every week we get a new WAR/GAR type stat from hockey twitter – what’s your stance on catch-all stats?

    In a very broad sense I approve. I think that they take a lot of stupidity out of a lot of decision making. This player is worth 12 goals a year, this player is worth 1 or 2, that tells you that the first player is a wiser choice and you can move on with your short and fragile life without nonsense. On the other hand, for more subtle decisions (how shall we adjust our lines, who should matchup against whom, which of these two 4-5 GAR players should we qualify and who should we let walk, etc.) you’ll need to know the pieces that make up the GAR. A catch-all stat is good though, since it keeps you honest — it saves you from making bad mistakes like preferring a defensive player to an offensive one when the weaknesses of the one don’t come close to outweighing the weaknesses of the other.

    Which area in hockey analytics do you think needs the most exploration?

    I’d like to see more work with non-linear models — finding out which combinations of players are more (or less) than the sum of their parts, for instance. The same sort of work would be good in optimizing rosters instead of just optimizing contracts, into optimizing whole season plans instead of just game plans.

    Could you handle giving up sour candy for Lent?

    This year I’m giving up beer which is already horrific. Sour candy would be a devotion too far.