We’ve been looking at various relationships between shots and goals in the NHL as we work on building a model (aiming to roll it out before the playoffs start) based mainly on those two variables. So far, we’ve found that the team that shoots more in an individual game doesn’t really score more, teams that shoot more over the course of a season do score more, and that over the course of the season a team that shoots more doesn’t necessarily see their opponents shoot more.
Today, we looked at whether the number of shots a team takes or goals a team scores in a single game has any relationship with the number of shots their opponent takes or goals their opponent scores. The sample, as its been for individual game relationships, is every game played between the start of the 2015-16 season and the end of the 2019-20 season, including playoff games (we’ve cut playoffs out for season-long regressions). Here’s what we’ve got:
| Relationship | R-Squared |
| Shots v Goals | 0.013486 |
| Shots v Opponent Shots | 0.01477 |
| Shots v Opponent Goals | 0.004919 |
| Goals v Opponent Goals | 0.004277 |
Nothing. No relationship. Very independent.
It’s possible that the playoffs throw this off, but given how weak any regular season relationships between these variables were, it does seem like we shouldn’t expect shots to be some magic key to show us how many goals a team will score. We’ll still experiment with incorporating shots into our model and see if we can build a better team rating formula (or better team rating formulas, if we end up splitting offense and defense or adding a shot volume metric after all) with them or without them, but it does seem like we may end up just going with goals—which would be simpler anyway.
More to come. Hopefully this weekend.
