Lindy Ruff's Dallas Stars Line Blender: A Look at the Hits and Misses of 2015-2016
Lindy Ruff's line blender tends to get criticized by fans and media, but finding chemistry is not as simple as adding a bunch of elite Corsi values together.
Group dynamics can be a highly counterintuitive process. Take this experiment from Berkeley's Haas School of Business documented in Christopher Chabris and Daniel Simons' wonderful book, The Invisible Gorilla. Four total strangers were asked to solve math questions from the GMAT. For 94 Percent of the problems, the final answers depended not on the wisdom of Sherlock Holmes, or comparing and contrasting each other's estimates, but on who spoke first.
Affection trumped ability. Confidence over calculation.
The point of this psychological introduction is that group dynamics in hockey can seem equally counterintuitive. Just replace confidence with chemistry, and corsi with calculation. Jamie Benn, Tyler Seguin, and Ales Hemsky are wonderful hockey players. Why not stick them together? Surely all of that positive Corsi will add up? Nope. When together, they post a Corsi For of 41.5 Percent, a Fenwick percentage that is even lower, and aren't expected to generate goals, but are expected to allow plenty.
By contrast, the 4th line of Colton Sceviour-Vernon Fiddler-Patrick Eaves have superior fancy stats over BennGuinSky. Obviously, ice time with each other skewers things. And these numbers need context, but it's fun to look at where Lindy Ruff's line blender hits, and where it misses. For the purpose of clarity, I've ignored lines that didn't stick, failed to get quality time, or just aren't worth documenting.
For the bulk of their tenure, Jamie Benn and Tyler Seguin have formed the first line per Corsica. The question that has kept Ruff and fans up at night has always been who compliments them best. Who is the Martin Short to Benn's Chevy Chase and Seguin's Steve Martin?
The results are sort of surprising. Basically, When Patrick Eaves gets to play the role of Porthos in this merry band of puck muskateers, Benn and Seguin generate more offense and scoring chances. They also suppress more offense. Small sample size? Sort of. Sebeaves, said trio (I'll run out of naming convention real estate shortly, so bear with me), has been on ice together for 94 minutes compared to BennGuin+Sharp's 480. The other takeaway is that despite the offense Cody Eakin, Jamie Benn, and Patrick Eaves have been generating lately, they aren't winning the shot generation/suppression battle. Small sample size? They've only had 32 minutes together, so perhaps it's only a matter of time until Harry meets Sally.
If there's one constant in Lindy Ruff's blenderglobe it's that a duo of wingers will follow Jason Spezza's puck ministry with religious zeal. The takeaway?
Spezza with Mattias Janmark and Ales Hemsky are good, Spezza with Nichushkin and Seguin are not. Other than that there's nothing to see here except the expected 'everyone likes Spezza' giggle index.
So a real quick #fancystat refresher for those who are either skeptics, just plain lazy, or just want to have fun and talk about hockey without having to remember what it was like to fail your first math test: Fenwick calculates unblocked shots. Why distinguish itself from Corsi, which calculates shots taken on and toward the goal? Because a blocked shot doesn't approximate a scoring chance. Think of Fenwick as the velocity to Corsi's volume. Why digress?
Because the Hemousska line is better at getting chances while the Hemoussark line is better at getting volume. There's a real debate there about which is better. Against a stout goaltender, you probably want a line that is better at creating volume to play the odds. It just depends. Whatever the case, it's quite clear that Radek Faksa and Mattias Janmark are better 3C's than Cody Eakin.
Alas, we get to the droogs of the Dallas Stars. The 4th line tends to cycle in and out. Like a Highlander sequel, there should be only none. Initially I wanted to ignore the data on this one, but the data dispels the myth of how good the "Kid Line" (Devin Shore with Valeri Nichushkin and Radek Faksa) was early in the season:
They only had 30 minutes together but that's plenty of time to avoid being a complete mess. The caveat here, however, is that the trio was only ever together on the road. So no matter what Ruff wanted, he couldn't protect them against the home team outside of not playing them at all. If there's a takeaway here it's that Vernon Fiddler and Colton Sceviour should be attached at the hip.
Expected Offense vs. Expected Defense
The last thing worth mentioning is what kind of offense versus what kind of defense we should expect from the selected trio of forwards. In addition to looking at shot generation and shot suppression I decided to peak at each trio's xG for and against per 60. xG, is a model combining shot quality with shot volume to tab expected goals.
For us plebes who get to chew popcorn on the sidelines, it's a wonderfully violent debate among the analytics community. I'll ignore that for the purpose of this post. Green = good, red = rotten.
- A lot of interesting "will they or won't they" young adult projections here. For Line 1, it's the "Baking" (Cody with Jamie and Tyler) line that projects to score more goals and by a fairly wide margin. They also don't project to get scored on much. That's good for 17th in the league in expected goals for differential among all trios with at least 50 minutes of ice time together at even strength.
- The caveat to observation one is that I didn't include Spezza with Benn and Seguin. Their xG Per 60 is 4.3, but they have a higher xGA Per 60 of 3.1. That line won't stick (except for close encounters[/cocksshotgun]), which is why I wasn't interested in documenting it except as a footnote here.
- Basically any line Patrick Eaves and his handleczar mustache are on is automatically made better defensively. Or I should say, projects to allow less goals. The majority of line combinations with Spezza don't project to defend well and likewise with Benn and Seguin.
- Roussel with Janmark and Hemsky projects to score more, and get scored on less. It's so good that their xG differential (or xGF%) is actually 14th in the league among elite trios, just above the Crosby-Kunitz-Hornqvist trio, and right below Couturier-Simmonds-Voracek. To their credit, Faksa with Roussel and Hemsky are just fine together. /
Of course, this is all surface level analysis. Looking at a trio of forwards ignores the pair of defensemen that help influence these stats. A lot of these current trios have been playing with John Klingberg and Jason Demers nursing injuries. Nonetheless, it's a fun peak into who gels and who doesn't. In summation, Ruffling ain't easy.