Watching The Spreadsheets: Dallas' possession metrics are problematic. But how important is shot share in an xG world?
Perhaps more critically: can a team succeed in spite of itself?
Fifteen years ago, we pulled ourselves out of the primordial soup of plus/minus when a bunch of nerds figured out a way to predict team success. It turned out that winning percentage and goal ratio were actually less predictive of future wins and future goal differential than...the difference in shot attempts (referred to as Corsi) in tied situations. Whoa!
Of course, you would never know this discovery because your average hockey broadcast has to make time for replays of random hits that didn’t make a difference, a meaningless fight, or a faceoff that wasn’t actually that important — or least not as important as say, the goal, the shooter, the possession that led to that faceoff in the zone to begin with, etc. But it’s true. And unlike most of the analytics work done during the boom before they all got picked up in the summer of 2014, that particular research still survives.
However, things change. What felt like research back then is simply a common measure. The biggest change was the introduction of shot quality models (referred to as expected goals, or xG). Since then, the fancy stats have gotten fancier and fancier. A big part of that — looking at the work of Louis Boulet, Jack Fraser, and Micah McCurdy in particular — is owed to Corey Sznajder, whose tracking project has given a lot of model creators more access to data that never otherwise existed, and still doesn’t: passing data, for example, entry and exit success, rush versus cycle information, and a whole lot more to fill in the blanks of performance instead of merely impact.
But today we’re not getting into the weeds of anything. We’re getting back to basics. Why? Because Dallas has a shot share problem. Of course, I’ve stopped talking about it in my game recaps because nobody wants to hear the same criticism over and over. But also because I don’t want to jump the gun and sound the alarm. So let’s talk about that alarm here.
The Stars are 30th in shot share (or percentage of shot attempts at even-strength). Shot attempts are 1550 to 1295 in their opponent’s favor. That’s an incredibly wide margin. It’s an alarming margin. But does it tell the whole story? Does it tell a story at all? Perhaps it’s more important than we assume? Perhaps it won’t matter? (This will seem doomy at first, but I promise there’s light at the end of the tunnel. Mostly.)
