During the 2015-16 season, Ryan Stimson and his team of volunteers undertook the tremendous work of tracking passes in as many NHL games as possible. In all, Stimson and his team collected data for 602 skaters, ranging from a low of ~100 minutes of tracked time on ice at even-strength (5-on-5) to a high of 1361 minutes (Chicago Blackhawks defenseman Niklas Hjalmarsson). The hard work involved is difficult to put into perspective.
But, why should we care? Couldn’t we just look at assists totals to figure out who possesses good passing skill and who doesn’t?
First, some background. In one way, the answer to the question above is “sure.” Assists do help to show us which players are skilled passers. In 2015-16, Erik Karlsson led the league in assists and Thornton, Letang, Getzlaf, Backstrom, and Crosby joined him in the top ten. We expect to see these players included in any list of great passers.
But there are problems with assist-centric analysis. Eric Tulsky, now with the Carolina Hurricanes, showed years ago that goal scoring at even-strength (something generally regarded as an obvious talent) has year-over-year repeatability.
*please click on the image of Tulsky’s graph above to check his original work on the subject.
*also, please note that Tulsky restricted his group to players that played at least 60 games
The correlation is expressed like this: r^2=0.22. This means that about 22 percent of a player’s goal total in a given year is explained by how many goals that player scored the year before. Primary assists (i.e. the pass immediately preceding a goal) has a lower but somewhat similar repeatability.
*from the same original work by Tulsky. Click the image or here to dig in further.
So, like before, about 18 percent of a player’s primary assist total in a season is explained by the number of primary assists they had the season prior.
That’s something but not a lot.
For secondary assists, the repeatability is around 5 percent. That means that those assists are approximately 95 percent based on factors other than a player’s talent. They’re “noise.”
Perhaps the most important takeaway here, and Tulsky nods to this, is that goals — something we consider a skill that a player possesses — has a year-over-year repeatability of only 22 percent.
That leaves us to find other explanations for the other 78 percent. Factors like a change in role or teammates, injuries, and lots of luck (among other factors) significantly influence a player’s goal total. Assists (primary and especially secondary) have an even greater percentage of other factors going on.
Okay, back to Stimson’s work on shot assists.
Performing the same checks on year-over-year repeatability shows that passing to players who then generate a Corsi is a clear skill that a player possesses – much clearer than the ability to create goals or assists. Stimson took the time to assemble these graphs:
For forwards (>100 minutes played), their year-to-year shot assists repeatability has an r^2 of 0.4179 – about 42 percent of their shot assists total is explained by the amount of shot assists they posted the year before.
For defensemen, the r^2 is about 0.35 – 35 percent. Compared to the repeatability of assists (and even goals) from Tulsky’s work, it’s very safe to call shot assists a true player skill.
Logically, this makes sense. A player can control their own ability to set up a teammate for an open shot. Will that player fire the puck wide of the net? Will that player delay, lose their opening, and shoot the puck into Kris Russell’s shin pads? Is that player Alex Ovechkin, who blasts the puck into the top corner from the left faceoff dot past a screened Carey Price, scoring a game winning goal?
One of these outcomes is a Corsi event (missed shot). The other is another kind of Corsi event (block by Russell). The final outcome is a shot on goal and results in an assist for our player. Our player doesn’t control their teammate’s accuracy, or even who their teammates are, but our player can control delivering the puck to a linemate in order to set up the shot.
So, we’ve established that shot assists are a clear talent. Now, the fun part. Which players have this talent? Let’s dig in.
*click here to interact with the graphic.
First, some broad strokes. NHL skaters run a wide range in their ability to create shots through passes. Henrik Sedin led all skaters, posting nearly 22 shot assists per 60 minutes (ShA60). At the other end of the spectrum, New Jersey’s Devante Smith-Pelly managed less than 1 shot assist per hour.
How does it look if we separate forwards from defensemen?
*click here for full view.
The league’s best shot assist rates belong to Henrik “Passy” Sedin, future all-star Joe Thornton, NLA-bound Kris Versteeg, Brendan Gallagher, Ryan O’Reilly, and Evgeni Malkin. Each player posted more than 18 ShA60, well beyond average for forwards (~9.5 ShA60).
Arizona, Columbus, Colorado, Detroit, New Jersey lacked above average shot assist distributors. Dallas, L.A., Pittsburgh, San Jose, and even Toronto boasted a pair of high-end performers each. Though neither would be described as assist-oriented players, Rick Nash and Tyler Ennis surprised by finishing at the bottom of their teams in shot assist rates.
*click the image or follow the link for the full view here.
On average, defensemen provide fewer shot assists, which makes sense. Defensemen are mostly charged with creating pre-offensive zone passes to allow forwards to get to work in the offensive zone.
Ottawa’s Erik Karlsson dominates in this regard. Here’s one extreme example of why he’s so good.
Karlsson’s stickhandling and vision are sublime, which helped him post just under 12 ShA60. That mark is on par with Eric Staal, Tyler Toffoli, John Tavares, and Mark Scheifele — all centermen in important offensive roles with their teams.
Karlsson is joined by the offensively gifted blue liners you might guess – Doughty, Letang, Hedman, and Klingberg all hover around the 10 ShA60 mark. There are some interesting names that pop up in strange places though.
TJ Brodie’s great passing places him in the second tier of shot assist rates, around 9.0 ShA60. Jake Muzzin, Thomas Hickey, and Jake Gardiner all slot in at a similar rate.
What do Klas Dahlbeck, Mike Weber, Nate Prosser, Mark Borowiecki, and Adam Larsson have in common? All posted between 3.5-4.0 ShA60, below league average for defensemen. Fortunately for the first four in the group, only one of these players was traded for Taylor Hall.
A couple of notes that warrant further study:
- Anaheim, Carolina, Dallas, Los Angeles, and the Islanders’ defense groups were strong as a group. Is there something about their systems (i.e. activating their defensemen to join rushes) that helps to explain their elevated totals? This links to Stimson’s discussion of Lindy Ruff’s neutral zone systems in Dallas.
- Centers seemed to dominate the top-end of shot assist rates among the forwards. How does a positional breakdown look?
- How does the shot assists breakdown look team-by-team? Does the team-level focus tell us much about which players may be poised to show point total improvements in 2016-17?
- Did any teams act to address needs in shot assist provision? (Hint: answer is yes!)
In the next piece, we’ll look deeper into some of these notes and see if we can uncover any answers. For now, enjoy another example why Karlsson is so good at shot assists.