View Single Post
12-20-2012, 11:41 AM
Registered User
Join Date: Feb 2007
Country: Canada
Posts: 581
vCash: 500
Originally Posted by Mike Farkas View Post
100% correct. Game is too fast, center red line is out, trapezoid is in, you can't fully expect a team to allow 14 shots a game...shot quality (something we really don't have a metric for) is drastically reduced by teams like Boston, St. Louis, Detroit, etc. to help insulate their goalies. See: Thomas' stats before Chara/Julien, but still the same NHL team. See: Brian Elliott before arriving in St. Louis.

Save percentage becomes even more more valuable than GAA in today's game...those that dismiss one as a team stat, would be incorrect to fully endorse the other as an "individual" stat...
As seventieslord already pointed out, it is by definition false to claim that save percentage is no more valuable than GAA. Even if team effects were so massive that the same goalie would have a .950 on one team and .850 on the other, that would still do absolutely nothing at all to make GAA more valuable compared to save percentage, because all team shot quality effects impact both stats equally, again by definition.

GAA = (1 - save percentage) x (shots against per 60 minutes)

Since save percentage is a component of GAA, every team impact on save percentage has the exact same impact on GAA, and then there is the additional and separate factor of shots against that is by far mostly impacted by the rest of the team.

Everyone wants to talk about Brian Elliott's .940 as if that somehow invalidates the usefulness of save percentage, but Elliott's lead in GAA was actually far greater than his lead in save percentage:

Save Percentage Leaders:
1. Elliott, .940
2. Schneider, .937
3. Lundqvist, .930

GAA Leaders:
1. Elliott, 1.56
2. Quick, 1.95
3. Schneider, 1.96

The only way to make the argument that GAA and save percentage are equally useful is to claim that goalies are just as responsible for the team's rate of shots against as the rest of the team is, which would be frankly absurd.

Now, it could be argued that GAA is the best way to compare goalies on the same team since they should have similar shot prevention in front of them and it might capture some of the non-save skills that goalies have that can create or prevent shots against. But when comparing across teams, GAA adds essentially no value.

If you want to argue that team effects are important, fine. It's correct that the rest of the team affects goaltending stats, the only disagreement we have is on the magnitude of that effect, particularly in the current NHL where the overall evidence (i.e. not just cherry-picking a few outliers) indicates that shot quality does not have a major impact in most cases. Contrast that to the original six and post-expansion eras where goalies often had very different save percentages when moving from team to team, suggesting much larger team effects.

Either way that doesn't mean save percentage is useless, and it certainly doesn't mean we should put any stock at all in GAA numbers. The best method is to use backup numbers or other estimates to adjust save percentage up or down to where you think it should be, rather than throwing the baby out with the bathwater.

(MOD: Edited out flaming - this is in response to C1958):

You're always looking at things at a team level or a league level, and just assuming that various factors must have had a big impact on goaltenders, when often there is little evidence that those factors had any impact on goaltending specifically. It is not enough to show that there was a rule change, or that a player was added or traded away, or a new coach came in, or there was a specific scheduling setup for some teams around the league. You have to relate that specifically to the goaltender and how that affected their performance, and if we're looking at save percentage then it has to specifically affect the difficulty of chances allowed, not just shot prevention or the team's overall performance.

Let's take your argument above that Harry Lumley's below-average save percentage numbers in 1952-53 were caused by injuries at the center position for the Leafs. The evidence actually goes in the exact opposite direction of your claim, as Harry Lumley was at his worst with both Ted Kennedy and Max Bentley playing and at his best with one or both of them out of the lineup.

Lumley had a terrible first eight games (4+ GA in five of the first eight starts) and a pretty mediocre first third (.884), and after that he played very well (.927 over the rest of the season). Ted Kennedy was injured on New Year's Day and came back for the last few games of the regular season. Max Bentley got injured in mid-November and was in and out of the lineup all season with a bad back. Assuming Kennedy missed 27 straight games after his injury and played the rest of the season, here's Lumley with and without Kennedy in the lineup:

With Kennedy: 1016 SA, 100 GA, .902
Without Kennedy: 880 SA, 67 GA, .924

Not to mention that the claim that Max Bentley's injury-related absence was a big part of the difference between the 1952-53 and 1953-54 teams makes no sense given Bentley was traded for cash over the offseason and replaced by players already on the Leafs' roster.

Note that I'm not saying that injuries to centers for the Leafs wasn't significant in terms of that team's season. It probably cost the team a lot of goals and likely played a big role in the team missing the playoffs. Newspaper articles during that season mention it a lot, I'd go as far as to say that you can't write the complete story of the 1952-53 Maple Leafs without discussing it. But there's no evidence that it had a big impact on goaltending specifically (unless you think I'm knocking Lumley because of his W/L record or the fact his team missed the playoffs, which I absolutely am not). And at the end of the day, the effect on the goaltender is really all we care about for the purposes of this project.

Last edited by Trebek: 12-20-2012 at 11:46 AM. Reason: Removed flaming quote
ContrarianGoaltender is offline   Reply With Quote