Goalies Don’t Steal Games
Goalie Memo #1
Introduction
In August, I wrote an article arguing Tomàš Vokoun’s case for Hall of Fame Induction (embedded directly below).
Towards the end of that article, I deviated away from my original subject to talk about the irrelevance of team success in goalie analysis, using Vokoun’s career as a prime example — including stretches of games where his Florida Panthers couldn’t win despite his extraordinary play.
I wish to now restate this argument, but apply to a much more micro scale and correct a notion that has long subsisted in hockey discourse: “stealing games”.
Let’s refresh ourselves on some of those first principles (click here to read that section of the article or continue reading to see the summary)
First Principles of Goaltender Influence
The win conditions of a hockey game are to out-score the other team
Goalies only prevent goals. Despite their disproportionate influence in goal prevention, they do not directly influence their team’s goal-scoring.
A shutout — the best goalie outcome — is a guaranteed non-loss and nothing more. Goalies cannot win without their team scoring.
Linards Feldbergs and His Quarter-Final Performance
One of my Christmas gifts this year was tickets to a quarter-final game at the 2025 World Junior Championship. When I received my gift on December 25th, 2024, the teams were both TBD, but by the day of the game, the matchup had been decided: Sweden versus Latvia.
Latvia had been the talk of the tournament ever since they had defeated the host Canadians in a shootout in the round robin. But even still, they finished 4th in their group and were going up against the only undefeated team in the tournament.
Sweden got an early lead off a play that frankly should have been at least an interference penalty — an effective pick — allowing goal scorer Forsfjall to streak down Feldbergs’s blocker side and rip one just over the glove pad. Wahlberg wired one in off a faceoff just over an in-game minute later. The first period ended 2-0. 19 shots against, 17 saves. Here’s a scenario where I’d love to apply my “Win probability maintained” statistic (see below for the original article) since it is best applied in seeing a goalie’s impact on a single-game, but I lack the components of xGoal data and win probability.
The second period experienced a wild change in trajectory. It started off with an early powerplay goal from Edstrom making it 3-0. Later in the period, Sweden had initially scored at the tail-end of a 5-on-3 powerplay, but Latvia shrewdly issued a coach’s challenge on the goal for an offside that occurred over a minute prior. The goal was overturned, and Latvia got a de-facto time-out for their trouble, but were down to three players once again. Upon the second penalty expiring for the second time, Sieradzkis sent Eriks Mateiko a stretch pass coming out of the box, which the latter put past Swedish goalie Melker Theilen to make it 3-1. Mateiko added another in the final 3 minutes of the period to make it 3-2: from the cusp of 4-0 to 3-2 — Latvia was back in the game.
The third period saw no goals, as Sweden used a sound strategy of “they can’t score without the puck” for the first 15 minutes thereof. Feldbergs kept Latvia in striking distance for an equalizer that never came. 20 shots — many of them high quality — all of them turned away and all for nought. By the end of the game, Linards Feldbergs — who was on the bench for the last 1:37 of the third — had saved 47 of 50 shots from the best U20 players of one of the premier hockey nations keeping the game closer than it had any right of being. The result was still a loss.
Indeed, looking at the volume of shots in the third period, one would have expected Latvia to be winning. But they weren’t, so they lost. Had they been winning, this would be a prototypical “stolen game”. But Latvia didn’t win — independent of Feldbergs’s play, they didn’t win. They just didn’t have the puck enough to score. The shots on goal were 50-13. We don’t have the shot attempt data from this game, but using the 4 games of NHL hockey for which we do have shot attempt data (i.e. 4 out of all games since 2007-08) where one team recorded more than 50 Shots on goal, and the other fewer than 15, we get a median Corsi For % of 76.4 — that means more than 3 out of every 4 shot attempts thrown at either net were by one team. We don’t have a way of knowing this, but this number does seem reasonable given that we do know that Sweden took 79.3% of all shots on goal.
Semantics(?)
So, what’s the point of me saying this? People have been saying this for years and my ‘um, actually’-ing won’t stop that or really have an effect on the discourse. It might appear innocuous and, in isolation, describing a goalie as having “stolen a game”, is. It’s no more damaging than any hyperbole — like kids describing their trip to the amusement park as “the best day ever!” Well, little Timmy, that assertion is plausible for you considering you’re 3 years old; however, I do recall your dropping the last of your ice cream being pretty catastrophic in the moment. Harmless, right? It’s not supposed to be accurate because it captures a bigger truth about the feeling of the moment.
What becomes damaging is when it becomes a characteristic: “a goalie who can steal games for you”. The danger is in the connection between “stealing games” and Wins. See, a stolen game, as evidenced by the Feldbergs example above, must definitionally be a win. And goalies, as mentioned in First Principles, above, can’t win without their team scoring.
But Latvia didn’t win — independent of Feldbergs’s play, they didn’t win
It’s damaging because it valourizes team results over goalie-specific outcomes and processes. Look at it this way: if you had to choose the best goalie of all time and the goalie you’d want starting for your team in game 7 of the cup final, the idea of “stealing games” would tempt you to consider having a different answer for each. In reality, the correct answer should be the same: whether the sample is one game or their entire career.
And this is the crux: it is best to judge performance on what gives the best chance of success, be it on a save-level, game-level, or season-level rather that reacting to composite results1 only. Thus, goalies cannot steal games, but rather put their teams in the best position to be successful.
Since all goalie statistics are result-based at this point in time, I differentiate “composite results”, like Wins (can’t control own team or other team) and GAA (can’t control defence’s ability to supress shot attempts or other team’s ability to hit the net). I talk about Save Percentage a lot, and that’s because saves are goalie action, whereas GAA is based on minutes played, which is not up to the goalie.