Yesterday, I wrote about how in hockey, the team that shoots more—at least in recent seasons—loses ever-so-slightly more than they win, and how on a game-by-game basis, the number of shots a team takes doesn’t have a very strong correlation with how many goals they score.
On a season, the story’s different.
To be clear, the story isn’t necessarily different when it comes to winning and losing. That, I haven’t looked at yet. And as I mentioned yesterday, if this all seems rather basic—it is. I’m writing about very high-level, basic numbers as I try to improve my own understanding of the numbers behind hockey, all as part of our effort to build out our NHL model in time for the beginning of the playoffs in a few weeks.
But the story. It’s different.
Over the course of a season, teams that shoot more score more.
This is five regular seasons worth of data, meaning it’s 153 data points (the Knights weren’t in the league for the first two of these seasons). And the correlation, while not strong, is definitely present, with our r-squared clocking in at 0.374.
The implication here—the implication of yesterday’s correlation and today’s correlation—is that because shots and goals don’t appear very linked in individual games, but shots and goals do appear linked over the course of the season, some teams are likely playing in higher-shot-volume games than other teams. Just as some basketball teams speed up the pace, some hockey teams, this would imply, create more scoring chances on both ends. Is this true? I don’t know. It’s the next thing to look into: Whether one team’s shots correlate with their opponents’ shots, or with their opponents’ goals. And, from there, what exactly that means.
Regardless of whether that implication turns out to be correct or incorrect, figuring this out is good for our model. Here’s why:
One of our desires with the model is to give individual game projections that we can compare against Vegas betting lines. We don’t expect our projections to make us money (and we certainly don’t expect that right away). But we do like having the ability to measure them against the most efficient predictors of sporting performance we know: betting markets. If shot volume is a piece of the equation, this gives us a way to project totals in games (also called over/unders), and may help us with the puck line (a spread that’s almost always set at 1½, like the run line in baseball). It also may help us in determining team strength: Should shot volume be a piece of that equation, or those equations? We still don’t know. But knowing this correlation will help us figure it out.
