It’s been a whirlwind in Habs land lately. First there was the firing of Marc Bergevin, the hiring of Jeff Gorton and now there’s a new GM of the Montreal Canadiens, in the form of one Kent Hughes.
After watching the press conference with Hughes, Gorton, and Geoff Molson, one word that kept getting me excited was “analytics.”
Firstly we need to define the difference between statistics and analytics. In a nutshell, stats are the numbers, and analytics are what we do with them. During the 2021 off-season, I did a series on my podcast of stats-based analysis of the Habs roster by creating a model. Then, I ran Montreal’s free-agent signings through the same model. The results were interesting to say the least.
Working on running Montreal's main pickups through my model.— Dylan Waugh (Habs-statician) (@_DylanWaugh) July 30, 2021
Let's just put it this way, it seems very obvious that Montreal is one of the few teams without an analytics department... pic.twitter.com/SmiDtnXv5e
The premise of this model is simple (I don’t profess to be one of these incredible stats folks you see on Twitter). Instead of measuring the raw stats I opted to figure out where a player fell in relation to the rest of their team. This was meant to eliminate bias if a player was on a good or a bad team.
So if player X played on a bad team, instead of saying player X had a bad Corsi-for percentage (his share of the shot attempts), I would say player X had the third-best Corsi of the 16 forwards ranked in their team.
It’s not a perfect system, but this hopefully makes players from one team more comparable to players on another team. Figuring out “fit” and “system” of course throws a monkey wrench in the whole equation, but this is a good starting point.
In rating forwards I used Corsi-for percentage, expected goals per 60 minutes, on-ice goals per 60, and individual points per 60; all five-on-five numbers. After knowing how the player ranked in relation to their team, I assigned points and then weighted those points for what I consider more or less valuable attributes.
That’s my hap-dash approach to creating analytics based on statistics. Is it perfect? Absolutely not, but it is a starting point. Before I run through Montreal’s forward acquisitions I just want to mention that the amount of points a player can get is finite. This means having less than 50% doesn’t necessarily mean a fail it means that a player is in the lower half of their team.
According to my model, Mike Hoffman had 64.5/156 points for last year’s playoffs and regular season. In the regular season he was just plain bad. He was bottom-three on his team in every stat that I measured. He had a decent uptick in the playoffs, but overall didn’t have a great year.
Mathieu Perreault had 91/195 He was pretty consistently mediocre compared to his team in both playoffs and regular season. Most of his stats were in the bottom half of his team but he was never too close to the bottom. He was pretty much what we would expect from a fourth-line guy.
David Savard was measured a little bit differently because he’s a defenceman. For the defence, I measured Corsi, expected goals against, On-ice goals against, points per 60, and time on ice per game. These are mostly defensive stats, so in theory a defensive defencemen like Savard should perform well.
Once again, that wasn’t the case. He had 55.5/119 points; yet another instance of being in the lower half of his team. The trouble is that this model was specifically designed to measure defensive defencemen. It gives extra marks for time on ice, and for lowering expected goals against. This should have been a cake walk for a guy with Savard’s reputation.
For a long time, teams that invested into analytics had a leg up on the opposition. They could identify a guy who was a good player but just wasn’t putting the puck in the net or had some other form of bad luck. Now in this day and age, almost every team has an analytics department. This certainly doesn’t prevent teams from making dumb mistakes, but the fact that all three of Montreal’s top free-agent signings were bad statistically is not a good sign.
Stats and our ability to read and define them are certainly not the be-all and end-all of creating a good hockey team. And to be clear I like all three of the players I talked about in general, but they’re a luxury that Montreal couldn’t afford, and not the solutions to the roster problems.
Look at how well Zach Bogosian did on the Toronto Maple Leafs and now for the Tampa Bay Lightning. Here’s a guy that’s had bad advanced stats his whole career, but clearly brings something good to a team, He’s a good addition to an already solid team putting up good numbers. So maybe you don’t build around a guy like Bogosian, but if you plug him into a situation where the team’s analytics are already good, then he can be a great complementary piece.
Marc Bergevin made a lot of really smart moves. I don’t even hate the signings I mentioned above. But I’m here for the new era of analytics that brings a method to roster-building, and I can’t wait to see what this team does next.