Which Devils are Grittiest? (Results)

Last week, I posted this piece in which I asked you all for ratings (1-5) on the “grittiness” of each of 28 Devils — it would become 30 after I added Seney and Gryba by request). I’m happy to say that 292 of you responded and that means I can get us some results in this promised piece.

In broad scope, there was a lot of disagreement. Every player had multiple votes of every grit level. It’s impossible for me to know if there were some people just trolling the poll, but assuming that’s not the case, it was a really highly varied dataset.

So, without further ado .. The 2019 New Jersey Devils Grittiness Rankings!

Grit Poll Results

You can see another, more comprehensive version of the data at my tableau. In this chart, the location of the circle on the x-axis is the average grit score, the size of the circle is the number of voters recorded, and the width of the line is the spread of the responses (+/- 1 standard deviation).

There is a pretty clear top tier of gritty players in Kurtis Gabriel, Blake Coleman, and Miles Wood. Almost everyone labeled each of these three players as a 4 or 5 on a 1-5 scale when it comes to grit.

This is likely unsurprising to most — they are the 3 team leaders in penalty minutes after all. Kurtis Gabriel’s primary role on the team is to amp up teammates with his physical game and willingness to drop the gloves. Miles Wood is no stranger to the ring himself, having been in 9 fights at the NHL level. Blake Coleman is the shiest of the trio with regards to formal fisticuffs, but likely earned is way up to the top as a runaway team leader in hit totals, a force on the forecheck and penalty kill, and an overall reputation for hustle.

The two least gritty players are perceived to be Jesper Bratt and Will Butcher. They weren’t quite as unanimously labeled as the gritty guys — they averaged about a 2.

Jesper Bratt is the tied for being the lightest Devil on the team and is the shortest among those with whom he is tied (Nico Hischier and Nick Lappin). He has yet to be in a fight, he has only 6 PIMs (least on team of any player with 500+ minutes), and he’s generally perceived as a bit of a liability defensively — which Hynes has essentially confirmed. Will Butcher is tied with Bratt and Vatanen as the shortest player on the team, he has just 16(!) PIMs in 2673 minutes played, and is really most effectively used in transition (exits, in particular) where his finesse game and passing is best put on display.

Does Grit Matter?

Something that you might be curious about at this point, is if being gritty is good. After all, Blake Coleman has been one of the Devils best players this year, and has been THE best over some stretches — he’s garnered two AAtJ “Devil of the Month” honors (November, January), and probably has been best player thus far in March as well. However, Jesper Bratt has scored more Pts/60 than anyone on the team not named Taylor Hall, and Will Butcher is one of the most efficient defenders in the league, ranking 24th of 139 NHL defenders with 1000+ minutes in GAR/60 — right between Torey Krug and Ryan Ellis. Some of the best players on the team are considered either the grittiest, and the wimpiest.

So, is Grit good? For one passive attempt at answering this question, I’ve decided to graph the grit vote of each player against their GAR rate this season.

It’s worth mentioning that penalties seem to be really important in gauging grit, and they are a fairly important negative component to GAR. That said, when I removed the penalty component, the graph didn’t look any more substantively correlated. I feel fairly confident in saying that gritter ≠ better. However, it also doesn’t mean worse.

So what does it mean, then? What is a grit?

Modeling Grit

I ran a 3-fold OLS regression to create a model that would be predictive of what type of player would receive a high “grit” score. I ran a few different models, trying all sorts of combinations of TOI by situation, possession metrics, GAR values, peripherals, and standard stats. Rates, Rels, and totals were all considered (Totals seemed most predictive). The only things that ended up being significant were goals, penalty minutes, time on ice, and blocks. Hits were relevant in earlier models, but it turned out that most of the predictive value of hits was contained in the penalty minute variable. Which actually does makes sense. Look at the top 3 PIM guys — its the top 3 grit guys. They have a lot of hits too, but from the models perspective, “Hits” was just a noisier “PIMs” when it comes to predicting grit.

The model had an adjusted r-squared of 0.732, all p-values were <0.01, and an RMSE of 0.293. The Expected Grit (xGrit) for each of the players was shown in the graph below as white circles.

The model seems to agree that there is a top 3, although it has Palmieri in a tier of his own above the rest of the pack. Also Bratt comes back up to the group while Butcher falls even further into last for the title of “softest Devil.” Interestingly, Zajac plummets in these rankings — the model doesn’t get why everyone thinks he’s gritty which is interesting because I think his perceived “grit” is a bit more complicated. He takes faceoffs, kills penalties, and is a veteran that wears the “A” and continues to find ways to be useful. His resourcefulness and leadership may be things people think of as “gritty” that don’t show up in the model.

Now that we have a model that (mostly?) works, I decided to give everyone a bonus graph:

This is what happens if we apply our model to the past 11 years of Devils rosters — because that’s how far back hockey-reference’s stats necessary for this model goes.

If I’m doing a self-diagnosis on how this model was extrapolated, I’d say that 1) it was clearly not prepared for David Clarkson and his 183 PIMs 2) the evaluation for defenders likely needs work, and 3) I don’t actually have a 3, but I already wrote “3)” because it felt natural and if you think I’m going back and editing an article about the statistical analysis of grit, then you’re a fool.

Summary and Your Thoughts

The Devils have 3 grit leaders in Wood, Gabriel, and Coleman that people seem to be in broad agreement on. Whether or not perceived “grit” is in any way valuable seems dubious at best. Whether it actively harms a player’s value also seems fairly difficult to prove, given Palmieri’s and Coleman’s positions among the top 5. Exactly what grit is measuring seems to be a bit of an open question, but it certainly has something to do with physicality (normally the illegal kind). The major thing’s it misses seem to me intangibles that a player like Travis Zajac typically exemplifies. And David Clarkson is too much man for my model.

What are your thoughts? Are you surprised at anyone’s results? What do you think of the predicted results and the what the model says? What does “grit” mean to you? Is it good, bad, or neither for a player to be gritty? Did you like this two-week article format and would you like to see more like it?

Thanks, as always, for reading, and leave your thoughts below!

Citation: hockey-reference.com and evolving-hockey.com were used for statistics. The idea for this article came from Dom Luzzzzzzzzzzzzzyn’s piece on The Athletic (paywall).