Player evaluation in hockey can be a complicated process, and perhaps never more than when it comes to evaluating goaltenders, an aspect of the game that has tended to confound both the eye test and data analysts. Scott Cullen takes a look at goaltenders using expected goals against.
For years, goals against average was a popular method to evaluate goaltenders, as if the number of shots allowed was something under the goaltenders control. So, to move beyond that, save percentage seemed to be a better indicator of goaltending proficiency. That’s an improvement, but there is still plenty of room for more details to be part of the evaluation process.
The focus now seems to be on finding a way to include shot quality as part of the evaluation. As Mike Smith could show you, or Tuukka Rask could show you, teams don’t necessarily allow the same quantity or quality of shots against.
One way to try to incorporate shot quality into goaltender measurement would be to include expected goals against and compare it to the actual goals against. Expected goal models are not perfect, and are still being refined, but they do attempt to address differences related to shot quality from one goaltender to another.
Using data from Corsica Hockey, over the past four seasons, here are the goaltenders that have played at least 50 games, sorted by total expected goals against (based on shot types and location) minus total actual goals against per 60 minutes.
xGA-60-GA/60 LEADERS, 2013-2017, MINIMUM 50 GP
|7||Henrik Lundqvist||N.Y. Rangers||232||.918||0.36|
|8||Jaroslav Halak||N.Y. Islanders||175||.917||0.32|
|11||Jonathan Quick||Los Angeles||206||.917||0.30|
|14||Martin Jones||San Jose||164||.916||0.26|
|15||Cory Schneider||New Jersey||231||.920||0.25|
|17||Thomas Greiss||N.Y. Islanders||137||.917||0.24|
|20||Ondrej Pavelec||N.Y. Rangers||148||.907||0.20|
|35||Peter Budaj||Tampa Bay||85||.913||0.08|
|36||Jake Allen||St. Louis||145||.916||0.08|
|54||Keith Kinkaid||New Jersey||68||.911||-0.05|
|60||Carter Hutton||St. Louis||105||.911||-0.10|
|63||Andrei Vasilevskiy||Tampa Bay||90||.915||-0.19|
|65||Darcy Kuemper||Los Angeles||96||.909||-0.25|
Seeing Carey Price at the top of the list should come as no surprise, he also has the best save percentage among goaltenders to play at least 50 games over the past four years.
While Price’s position is as expected, some goaltenders are a little more surprising.
Philipp Grubauer has been a top-drawer backup in Washington, and seems to be next in line among second-stringers who are worthy of a starting opportunity.
Andrew Hammond had a miracle run with the Ottawa Senators in 2014-2015, and that weighs heavily on his results, it also seems strange that he’s been banished to the AHL (where he’s admittedly struggled in super small samples over the past couple of years).
Scott Darling looks like a pretty good bet by the Carolina Hurricanes. Darling performed well in limited action as Chicago’s backup.
As poor as Ondrej Pavelec’s reputation was in Winnipeg, based largely on his poor save percentage, the fact that he has saved more than the expected goals against does offer some reason to be optimistic about results in a different situation.
Interestingly, a lot of goaltenders that moved on the goaltending carousel this summer fall in a similar range. That includes Eddie Lack, Anders Nilsson, Chad Johnson, Ben Bishop, mike Smith, Steve Mason, Jonathan Bernier and Ryan Miller, all slotting between .10 and .17 xGA-GA/60. That looks like a lot of interchangeable parts.
Nashville’s Pekka Rinne, coming off a big postseason run for the Predators, doesn’t look so good by this measure. Some relatively new starters, or those on the verge of taking that job, have also had some struggles. That includes Tampa Bay’s Andrei Vasilevskiy, Vancouver’s Jacob Markstrom, and Buffalo’s Robin Lehner.
Even a goaltender like Minnesota’s Devan Dubnyk, who was a Vezina candidate part way through last season and finished fifth in the voting, and finished third in the 2014-2015 Vezina voting, is just barely treading water above the expected goals against over the past four seasons.
This isn't supposed to be a new perfect measure for goaltenders, but it is at least worth investigating cases in which shot quality may contribute to variance in a goaltender’s perceived performance. As more information becomes available, it's possible to come up with a more complete picture of a goaltender's contribution.
Scott Cullen can be reached at email@example.com