Last week, Ryan Stimson published an exciting new article that uses clustering to identify the playing styles of individual players. The data used in the model comes from Ryan and the volunteers who contribute to his passing project, as well as Corey Sznaijder who is tracking passing data and zone exit and entry data for all 30 teams this season.
I highly recommend reading Ryan’s article. The more technical concepts are presented in an accessible way. But to summarize, he used a modeling technique to assess how players tend to cluster across several different statistics. He then used those clusters to define different types of players.
Before we dive into the different types of players, let’s quickly define the metrics used for the clustering:
Shots – Any shot (corsi)
Shot Assists – The first pass before a shot
Shot Contributions – Shots + Shot assists
Build-Up – The second and third pass before a shot
Shot Contribution Influence – Shots + shot assists + build-up
Passing – Shot assists + build-up
Transition – Zone exits that lead lead to zone entries
The Forwards
For forwards, the types are:
Playmakers – Forwards who excel in all areas.
Balanced – Forwards who perform well in all areas but not quite to the level of the playmakers and who don’t excel in one specific metric.
Shooters – Forwards who are primarily shooters and aren’t as effective in passing or contributing to zone entries and exits.
Dependent – Forwards who aren’t particularly strong in any area and rely on their linemates to drive play.
With those definitions as our foundation, let’s see what we can learn about the Lightning roster. The dashboard below shows each player on the Lightning grouped by the player type according to Ryan’s work. The bar graphs show how each player rates in each statistic used in the clustering process. The dotted lines indicate the 25th, 50th, and 75th percentiles, which roughly equate to the typical four forward line system.
From this view, we can see what about each player’s performance led them to be identified as their given player type. For example, Nikita Kucherov is an elite player by every statistic and therefore is a playmaker. JT Brown by contrast isn’t nearly as well-rounded but excels most in individual shots.
You’ll probably be surprised to see Brayden Point in the dependent category. That is due to the lack of data collected on him thus far. Corey hasn’t tracked much of the Lightning season yet and most of the games he has tracked are from earlier in the year. Point has emerged as a great player in the last couple of months and I expect that if we revisit this assessment with complete data at the end of next season he’ll be grouped with the playmakers or the balanced forwards.
The Defenders
While the data Ryan used mostly focuses on offensive impact and thus is better for assessing forward play, he did also go through the same exercise for defenders. Naturally, the clusters occurred differently and were defined differently. For defenders, the types are:
All-Around – Defenders who are strong in all areas
Volume Shooter – Defenders who are primarily shooters when they get into the offensive zone
Puck Mover – Defenders who shoot less and mostly contribute to getting the puck out of the defensive zone to start breakouts
Defense First – Defenders who don’t contribute much offensively
The Lightning have just one all-around defender but what a defender he is. Victor Hedman grades out among the best defenders by all seven metrics and is clearly one of the best in the league. The Lightning defense is full of puck movers and doesn’t have a single player who fell into the defense-first cluster. I suspect that Jake Dotchin might land there when we get more data on him but he also might end up being a volume shooter. Slater Koekkoek didn’t do much to contribute in the offensive zone during his time in Tampa this year but he was one of the better players in contributing to buildup play and transition plays.
Putting It All Together
The player types alone are interesting but Ryan takes the next step in this exercise by identifying which player types are most successful in combination. In a vacuum, it’s easy to say that a team should have twelve playmakers at forward. But in reality, that’s not possible due to the salary cap and drafting.
So the real challenge becomes figuring out how to structure lines and pairs to get the optimal results. Fortunately, Petbugs used Ryan’s work to create an Excel workbook where anyone can play with all 30 teams’ rosters to identify optimal lineups. Unfortunately for us, the Lightning have made so many roster changes this year that the tool isn’t particularly effective for the Bolts. However, we can go through the process manually making some assumptions based on what we know so far.
For the forwards, I’m going to assume that Brayden Point is a playmaker since he certainly seems to be headed on that trajectory. If that’s the case, an ideal top six based on Ryan’s methodology would be:
Vladislav Namestnikov – Steven Stamkos- Nikita Kucherov
Ondrej Palat – Brayden Point – Jonathan Drouin
The only controversial point here is probably Vladdy on the top line and Palat on the second. The logic is that this creates more balance. Kucherov and Stamkos are so great that putting another playmaker with them is almost a waste. With a balanced player like Namestnikov, the first line will still be dominant and that allows Palat to slide down to the second line where he can provide more of a boost.
In this scenario, the second line is probably better than some teams’ first lines. If Brayden Point does grade out as a playmaker, this would be a line of three playmakers, which is the best combination according to Ryan’s work.
The third line would also be off to a great start built around Tyler Johnson and Alex Killorn. The right wing position on that line is open to competition heading into next season. Ryan Callahan, Yanni Gourde, J.T. Brown, and maybe even Adam Erne will all likely make a run at earning that spot. The fourth line will likely be some combination of the players above with Cedric Paquette and Gabriel Dumont also in the mix. So hypothetically, let’s say the bottom six shakes out as:
Alex Killorn – Tyler Johnson – Yanni Gourde
Adam Erne – Cedric Paquette – J.T. Brown
If we project Gourde (or whoever fills that 3rd line spot) as a balanced player and Erne (or whoever fills that fourth line spot) as a dependent player, the Lightning would project as a 54.8% expected goal share team. That would rank 3rd in the NHL this season using the Corsica method of calculating xG, which is different than Ryan’s.
The defense is a more difficult exercise because Jake Dotchin is a bit of a wildcard on the top pairing and I’m unsure of how the organization perceives Slater Koekkoek. But using what we currently know, let’s project next year’s pairs as:
Victor Hedman – Jake Dotchin
Anton Stralman – Slater Koekkoek
Braydon Coburn – Andrej Sustr
This scenario is open to lots of changes. Dotchin might not be on the first pair. Koekkoek and Sustr might not be in the lineup at all. I’ve assumed that Garrison won’t be back but that’s far from a guarantee as well. But if we assume Dotchin will be defense-first, these pairs would project to a 52.4% expected goal share. While not as good as the forwards, that’s still a solid number. And combined with the forwards, that would be a strong team heading into next season.
Wrap-Up
Ryan’s work in this area is critical to taking the next steps in hockey analytics. Identifying which players are most likely to be successful together is a huge part of player evaluation. Using this type of approach will lead to better decisions in free agency and better day to day decision making in lineup construction. I encourage you to dig into Ryan’s work on your own and play with the Petbugs’ lineup optimizer to see how these different player types interact.
