Granlund’s Hat Trick among the best performances this year

I am way late to the party on this one, but I’ve been working on an analysis of Granlund’s hat trick last week and I got buried with job interviews and meetings, and never finished the post. So, I wanted to make sure I put something out there to share what I worked on. We have been looking more and more at game score this year, and have been posting them after games on our blog. I have explained why I like the stat previously, but I think it’s important to continue the discussion of what the stat can do for us and look at it in new ways.

One of the things I’ve heard about game scores is that, “Well, I like the stat but it basically just confirms the eye test so I’m not really sure what value it adds.” And for the most part, this is true…on a per-game basis, we generally know from watching the game which players contributed the most (usually the top scorers), but it’s important to be able to quantify those contributions, particularly over the long run. Good research is all about good measurement, and once we have an established metric (even if it’s not perfect,) we have the power to answer lots more questions than we could have before. Think about diagnostics information for a car or a computer…you can measure things like engine temperature, CPU usage, etc. and when you look at them over time, a pattern emerges. As Derek emphasized on the podcast, this stat allows us to start conversations, not end them. To that end, here’s a conversation I wanted to start:

How does Mikael Granlund’s hat trick stack up against the rest of the hat tricks in the NHL this year?

To answer this, I collected game score information for all NHL hat tricks in the 16-17 season from I made my own GS calculator in Excel so I made sure it was working properly before I saved my output data. If you need a refresher, check out this post on game score. The main thing to remember when looking at the tables below are that individual stats are ‘all situations,’ and possession (Corsi and goals) are 5v5 only.

There have been 40 hat tricks as of 2/13/17, and obviously, we’re splitting hairs to a certain extent when we compare them…all these games are ‘good games,’ but the question is which of them stood out as the best of the best. I decided to add a layer to our score breakdown, and I categorized the component stats into “Primary points” (G+A1), “Offense” (A2+SOG), “Grit” (Pens Drawn, Pens taken, Blocks), “Faceoffs”, and “On-ice” (CF, CA, GF, GA). This way, we could look at the overall game scores, but also understand what kind of game the player actually had.

First, let’s look at the bottom-ten hat tricks of the year (click the image for a larger view):

I know there’s a lot in that table, and a lot of color–I try to add some kind of visual aspect whenever I do large tables because otherwise they can be almost uninterpretable. The categories in the middle columns are color-coded for sake of ease, and the scores and components on the right are formatted automatically on a ‘stop light’ scale of Green-Yellow-Red.

In theory, the ‘basic hat trick’ would be 3 goals on 3 shots, for a game score of 2.48. We see that Puempel and Green both scored lower than that by nature of the on-ice portion (-7 and -6 Corsi, respectively.) Puempel actually got dinged slightly because all of his goals were on the Power Play, and he had a goal diff of 0 at even strength. Overall, these ten games are typically 3 goals, no other assists, 3-5 shots, and a small contribution one way or the other for faceoffs (Schenn, Berglund) or other offense (Kreider with 6 SOG).

Now, let’s look at the top-ten, including Granlund’s effort which was good for seventh-best this year (click the image for a larger view):

Pacioretty leads the field this year with 4 goals and 1 primary assist on 8 shots. Add in a +16 corsi rating and +4 goals at even strength, and you’ve got a massive 5.55 score. It would have been higher if he hadn’t taken a penalty, which none of the other in the top-ten did in their games. Matthews’ opening night gem ranks second, while Marleau’s 4-goal game is matched by Rust and Perron, who both had 4 primary points and a penalty drawn to round out the top five.

Granlund’s game is easily among the best hat tricks this year, and his 3.75 score comes mostly from his four primary points, but six SOG certainly help, and while he was -2 in Corsi, he was +3 in goals. He gets goose eggs for face-offs and grit, so we see that this game was driven by strong offensive play and solid defense (0 goals allowed when on ice). Granlund’s offensive contribution has been insane this year, and it was a bit unfair that he scored his first career hat trick on the road. It would have been a lot of fun to see him celebrate that in front of a home crowd with the Wild hats raining down.

The categories I created were intended to help me wrap my head around the stats as I worked through this piece, but I suppose they can be helpful to look under the hood before I wrap this up. As I was working through this data, I realized that by looking at game score components, the bulk of each score is comprised of primary points, on-ice possession, and other offense (shots and secondary assists).

This makes sense for the games by players that all scored a lot of goals, but I got curious what this breakdown would look like for a single team game. Let’s look at the same chart for one team during a game. Here’s the Wild from the 2/4/17 game where Granlund got his hat trick:

We still see mostly primary point production, on-ice possession, and other offense making up most players’ scores, but certainly to different degrees than the hat tricks above. Some people might say that grit and faceoffs should be worth more, but research spanning almost a decade now shows that they just aren’t that impactful to winning games.

Ultimately, game scores provide a certain amount of value in being able to more or less confirm the eye test each night, but it’s important to understand how having a consistent metric can allow us to answer more questions that we couldn’t with just our eyes. The power of stats like these is not necessarily in the scores each night, but the creativity we use to think about how to use them practically and ask different questions.

Thanks for reading, make sure to follow me on Twitter @BobaFenwick and @HTHpod !