Using analytics to build the optimal lineup
We’ve got numbers and charts!
Among all the hockey accounts that I have followed since joining Twitter, my favorites have been those in the hockey analytics community. I personally find that visualizations of advanced stats enhance my enjoyment of the game, and help me to understand what each player contributes, beyond what I can see from watching games.
Today I want to share some recent research by Ryan Stimson. Back in April, he wrote an article for Hockey-Graphs.com that identified 4 playing styles for forwards, and what offensive results you could expect from different forward combinations.
I encourage you to read the entire article for a full explanation and a description of his methodology. In summary, the four styles of forwards are Playmaker, Shooter, Balanced, and Dependent. Playmakers are elite at everything, while Balanced are above average at everything. Shooters have elite shooting numbers (obviously) but are not as strong at passing. Dependents don’t contribute positively on offense. Most teams average just 4 playmakers/shooters, so the rest of the lineup consists of balanced and dependents.
The article contains this chart, which shows what expected goal percentage would result from each possible combination. Note that this does not take positions into account. For example, a playmaker-balanced-balanced lineup would produce the same whether the playmaker was a center or a wing.
Another great Twitter follow is Sean Tierney. This summer, he took Stimson’s work and created optimal lineups for each team. Here is his attempt for Columbus:
CBJ - optimal lines— Sean Tierney (@ChartingHockey) August 9, 2017
If things shake out like this, L2 and L1 are kind of interchangeable. Panarin-Jenner-Atkinson would have 55.1 xGF%. pic.twitter.com/LrIer4nUsz
This lineup looks very reasonable, but I think we can improve upon it. To do so requires answering two questions:
Can a player’s style change? In particular, Josh Anderson and Oliver BJORKSTRAND.
What role will rookies play? The safe assumption is that they will be Dependent, but if you look at BJORKSTRAND’s numbers so far, he’s already a Balanced, so if we’re being optimistic, we can give them a higher ceiling.
Let’s tackle the first question first: Anderson’s numbers say he’s a Dependent, but I think he was weighed down by his most common linemates, Matt Calvert and William Karlsson. If he gets to play with Balanced players, I could see him evolving into a player similar to Boone Jenner. (You can visit this site to look up this graph for any NHL player). Anderson just needs a slight improvement in his transition play and passing.
In the same way, I see BJORKSTRAND as a similar player to Cam Atkinson:
His shot volume and passing numbers are close. He just needs to improve his transition play and get in position for more high danger shots.
As for the rookies, PLD has been compared to Brandon Dubinsky, so I think his ceiling is as a Balanced player. The other rookie I want to see make an impact is Vitaly Abramov. In the QMJHL and in the Traverse City tournament this weekend, he has proven to be a dynamic offensive player. I hope he’ll have a strong showing in camp and earn a spot on the team. His ceiling would be as another Atkinson or BJORKSTRAND type player.
With that in mind, these are the lines I propose:
Line 1: Panarin (Playmaker) - Wennberg (Balanced) - BJORKSTRAND (Balanced/Shooter)
xGF%: 52.2 (BJORKSTRAND as Balanced) or 55 (BJORKSTRAND as Shooter)
It appears that Panarin and Wennberg will play together. I think any forward would benefit from being on a line with such dynamic forwards. Why not BJORKSTRAND?
Line 2: Foligno (Balanced) - Dubinsky (Balanced) - Anderson (Dependent/Shooter)
xGF%: 45.4 (Anderson as Dependent) or 50.9 (Anderson as Shooter)
The team clearly values Anderson, so let’s see how he does with two veterans who are equally adept at passing and shooting. Also, this line would hurt to play against.
Line 3: Atkinson (Shooter) - Dubois (Dependent/Balanced) - Abramov (Dependent/Shooter)
xGF%: 44.7 (both rookies as Dependent) or 48.5 (Dubois as Balanced) or 52.1 (Abramov as Shooter) or 54.7 (Dubois as Balanced AND Abramov as Shooter)
Dubois is a big, physical center who would be between two small, speedy wings.
Line 4: Calvert or Motte (Dependent) - Sedlak (Dependent) - Jenner (Shooter)
OK, Boone draws the short straw here, but he could earn additional minutes on special teams. I’ve been saying for months that I would drop Calvert from the lineup entirely. I think Sedlak has room to grow, with the right linemates. He put up good possession numbers getting sheltered minutes between the veterans Hartnell and Gagner.
What say you? Using this data, how would you set up the Columbus forward lines?