Practical Concerns: How To Replace Superstar Goal-Scoring

Photo Credit: Derek DrummondPhoto Credit: Derek Drummond

Earlier this week, I read an interesting story by Frank Seravalli on on Mike Babcock, Phil Kessel and the Toronto Maple Leafs, which is a good read for anything interested in how coaches think. For me, it also illustrate another way analytics could be employed to make a coach’s life a lot easier and take out some of the guesswork inherent in the job.

It was not too surprising to hear that Babcock was already thinking about how to get the most out of his team this season while on vacation, but I am very curious about his thought process behind how best to replace Phil Kessel. But before we start thinking about how to replace Phil Kessel (or his production in aggregate), we need to start thinking about how we are to measure a Phil Kessel’s offensive contribution.

The Method (finding a way to make good guesses)

The first way to go about it, and the way that most hockey people I know do it, is to use intuition. A coach of Babcock’s ilk has many years of hands-on experience to lean on, and he can also sit down with a few other colleagues and crowdsource a hypothesis while shooting the breeze.

There are some very important upsides to this approach. First of all, it doesn’t take a lot time to look at each player and just make an educated guess on how many goals he’ll be good for this year – intuition works fast, and if you are intelligent and have a lot of experience, you’ll usually be fairly accurate. Second, the whole ritual of sitting down with other hockey people and hashing things out generates good ideas and reinforces relationships. There’s nothing I like better than to watch a sporting event (hockey, football, tennis, whatever) with a quality thinker of the game such as my dear friend Michael Farber, and the same must hold true in the upper echelons of the game.

The second way to go about projecting offense involves, of course, extrapolating past data. Mike Babcock and his staff have some good analytics people to entrust that exercise to, and there are some useful solutions available in the public sphere (Hockey-Reference and War On Ice being two resources I like to consult on the matter).

Of course, when trying to predict the future, method matters. In the interest of public knowledge, I will share how I accounted for the departure (via graduation – congrats ladies) of three players on the McGill Martlets: Leslie Oles (RW), Katia Clement-Heydra (C) and Michelle Daigneault (D).

The Madness (trying to make good guesses)

When I start thinking about our team losing Leslie, Katia and Michelle (all of whom I believe will be very excellent additions to the CWHL’s Montreal Stars), I think about the opening chapters of the book Moneyball, when Billy Beane and his stats guys are trying to wrap their heads around the idea of losing key contributors such as Jason Giambi, Johnny Damon and Jason Isringhausen. As the book showed, it is possible to replace the overall outputs of these players, proving that you have a good understanding of what drives results in the first place. So during the summer, while the rest of the staff was hard at work figuring out coaching and teaching-related solutions to our problem, I watched a bunch of game tape and generated some spreadsheets.

Most people interested in hockey analytics dislike using goal-based metrics, and I am no exception. We don’t play a ton of games in the CIS, so even counting exhibitions and tournaments I’ll only have about 25 games’ worth of information to process. So here is what I did:

  1. I went into our video database and rewatched every single shot on goal from our past season.
  2. I then classified each shot on goal by game situation, outcome and some other contextual factors (while also accounting for set-up passes, as inspired by Ryan Stimson’s Passing Project).
  3. I weighed these factors in a certain way, and then calculated each player’s contribution to team offense in percentage.
  4. With those percentage in hand, it become relatively easy to estimate how many goals each player created throughout the season.

Now, for some very specific reasons, this study is not related to defense or possession (as measured by CF%) in any way. Because of the makeup of our team and the way that we play, I was more curious to learn more about “finishing talent” and “shot quality,” and came to some interesting findings (which I unfortunately am not at liberty to share right now).

With just about 1000 shots in my sample, I realize that the conclusions of my work cannot be assumed to be robust, but taking a quantitative approach allowed our staff to confirm some of our prior assumptions, and challenge other ones. As we make our way through the 2015-16 CIS Women’s Ice Hockey season, the technology and framework we were able to develop will help us improve our development process, and give us reliable means to track our progress.


Jack Han is the Video & Analytics Coordinator for the McGill Martlet Hockey team. He also writes occasionally about the NHL for Habs Eyes on the Prize. You can find him on Twitter or on the ice at McConnell Arena.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s