Your whole article is based one huge incorrect assumption: that Sv% is independent of team quality. A better defensive team will not only allow fewer number of shots against, but also lesser quality shots against, which will leave the same quality goalkeeper with not only a lower GAA but also a higher Sv%.
If you watch going forward, and as you have seen since the trade, Ryan Miller will have a higher Sv% in addition to lower GAA.
March 29, 2014 12:43 pm
I have no facts to show that SV% is tied with quality of team, but I've always worked on the assumption that teams that allow MORE shots have goalies with HIGHER SV%.
I've employed Quick and Bernier on my fantasy roster this season and it has worked out extremely effectively for both GAA and SV% ... Quick has an excellent GAA, with average SV%, while Bernier has an excellent SV% with average GAA. Quick faces few shots with LA, while Bernier faces a ton with the Leafs.
Few people would argue that Bernier is a better goalie than Quick based on SV%, but for some reason, it seems than Bernier's got such a high SV% because of the fact he faces so many shots.
You may want to incorporate Scoring Chances Against into your formula above to increase its accuracy.
Bottom line is, hockey is a team sport, and just about every player stat -- especially goalie stats -- are a reflection of team play.
March 29, 2014 7:12 pm
Shot quality is statistically irrelevant in this context IMO. Basically, I feel like any large sample will pretty much equal any other equal-sized sample sample across the league, with respect to shot quality. So it stands to reason that a team that allows 10 fewer shots per game will allow a smaller number of "quality" shots per game, but it has very little statistical effect on save percentage. Even though it'll take them longer to build a large sample, the percentages will be largely similar at the end of the day. Just an opinion based on observation. Feel free to have a numbers guy scrape the data and prove it one way or the other. That's an article I would love to read.
April 22, 2014 4:52 pm
Hockey is still in dire need of more advanced statistics but even without them I don't think you can treat all shots as being equal. There's a distribution of difficulty there, or a baseline probability of scoring for each shot against. A goalie's SV% is going to depend on his talent and the type of shot distribution he faces. Unlike Jamie, I doubt that distribution is the same for every team in the NHL. It's a shame data like that isn't readily available, but maybe there's something similar that'd be a useful correction factor. I like the suggestion by Mike of allowed scoring chances. I'm a bit worried it'd be a little too subjective and smallish a sample size but still worth a try.
I think the larger point is that all things being equal, less shots against equals a lower GAA. That's an important if simple/straightforward concept of course. I just don't like making the assumption that the SV% will remain equal. If you're serious about predicting GAA shifts, examine how you might be able to predict SV% changes as well. As you show, a 0.1 change in SV% is about 0.25 GAA, or equivalent to the change seen by 3-4 shots against. Small changes in the difficulty distribution for those shots is going to impact both SV% and GAA.
As mentioned above, the numbers needed to do this sort of thing probably don't exist or at least aren't readily available. I am really curious about this now though, so maybe I can find a few specific situations that could be examined to determine if the distribution plays a role or averages out.
May 18, 2014 11:41 pm
Miller's SV% has gone down since the trade, not up. He was .924 in Buffalo and is .918 in St. Louis.
What evidence can you offer that better defensive teams allow lesser quality shots and that this leads to meaningful rises in SV%? Most studies that look at a team's ability to suppress shot quality reveal that the effect is negligible.
May 18, 2014 11:41 pm
Miller's SV% has gone down since the trade, not up. He was .924 in Buffalo and is .918 in St. Louis. And I invite you to check out the GAA and SV% of Jaroslav Halak who went from a low shots against team to a high shots against team
What evidence can you offer that better defensive teams allow lesser quality shots and that this leads to meaningful rises in SV%? Most studies that look at a team's ability to suppress shot quality reveal that the effect is negligible.
May 18, 2014 11:41 pm
Miller's SV% has gone down since the trade, not up. He was .924 in Buffalo and is .918 in St. Louis. And I invite you to check out the GAA and SV% of Jaroslav Halak who went from a low shots against team to a high shots against team. The results are in complete disagreement to your argument.
What evidence can you offer that better defensive teams allow lesser quality shots and that this leads to meaningful rises in SV%? Most studies that look at a team's ability to suppress shot quality reveal that the effect is negligible.
May 18, 2014 11:41 pm
The formula I posted for GAA can't get any more accurate because... it is the formula for computing GAA.
Just to follow up - http://somekindofninja.com/nhl/ tracks shot locations and an example application: http://www.dailyfaceoff.com/39366/adjusted-save-percentage-measuring-the-impact-defense-has-on-goaltending-statistics
It looks like some of the needed data is starting to become available. Shot location isn't the only factor in save difficulty, but I'd bet it is at least among the largest correlated contributors.
Leave a comment
In order to leave a comment, you must be a registered user at our website and logged in.
Your whole article is based one huge incorrect assumption: that Sv% is independent of team quality. A better defensive team will not only allow fewer number of shots against, but also lesser quality shots against, which will leave the same quality goalkeeper with not only a lower GAA but also a higher Sv%. If you watch going forward, and as you have seen since the trade, Ryan Miller will have a higher Sv% in addition to lower GAA.
I have no facts to show that SV% is tied with quality of team, but I've always worked on the assumption that teams that allow MORE shots have goalies with HIGHER SV%. I've employed Quick and Bernier on my fantasy roster this season and it has worked out extremely effectively for both GAA and SV% ... Quick has an excellent GAA, with average SV%, while Bernier has an excellent SV% with average GAA. Quick faces few shots with LA, while Bernier faces a ton with the Leafs. Few people would argue that Bernier is a better goalie than Quick based on SV%, but for some reason, it seems than Bernier's got such a high SV% because of the fact he faces so many shots. You may want to incorporate Scoring Chances Against into your formula above to increase its accuracy. Bottom line is, hockey is a team sport, and just about every player stat -- especially goalie stats -- are a reflection of team play.
Shot quality is statistically irrelevant in this context IMO. Basically, I feel like any large sample will pretty much equal any other equal-sized sample sample across the league, with respect to shot quality. So it stands to reason that a team that allows 10 fewer shots per game will allow a smaller number of "quality" shots per game, but it has very little statistical effect on save percentage. Even though it'll take them longer to build a large sample, the percentages will be largely similar at the end of the day. Just an opinion based on observation. Feel free to have a numbers guy scrape the data and prove it one way or the other. That's an article I would love to read.
Hockey is still in dire need of more advanced statistics but even without them I don't think you can treat all shots as being equal. There's a distribution of difficulty there, or a baseline probability of scoring for each shot against. A goalie's SV% is going to depend on his talent and the type of shot distribution he faces. Unlike Jamie, I doubt that distribution is the same for every team in the NHL. It's a shame data like that isn't readily available, but maybe there's something similar that'd be a useful correction factor. I like the suggestion by Mike of allowed scoring chances. I'm a bit worried it'd be a little too subjective and smallish a sample size but still worth a try. I think the larger point is that all things being equal, less shots against equals a lower GAA. That's an important if simple/straightforward concept of course. I just don't like making the assumption that the SV% will remain equal. If you're serious about predicting GAA shifts, examine how you might be able to predict SV% changes as well. As you show, a 0.1 change in SV% is about 0.25 GAA, or equivalent to the change seen by 3-4 shots against. Small changes in the difficulty distribution for those shots is going to impact both SV% and GAA. As mentioned above, the numbers needed to do this sort of thing probably don't exist or at least aren't readily available. I am really curious about this now though, so maybe I can find a few specific situations that could be examined to determine if the distribution plays a role or averages out.
Miller's SV% has gone down since the trade, not up. He was .924 in Buffalo and is .918 in St. Louis.
What evidence can you offer that better defensive teams allow lesser quality shots and that this leads to meaningful rises in SV%? Most studies that look at a team's ability to suppress shot quality reveal that the effect is negligible.
Miller's SV% has gone down since the trade, not up. He was .924 in Buffalo and is .918 in St. Louis. And I invite you to check out the GAA and SV% of Jaroslav Halak who went from a low shots against team to a high shots against team
What evidence can you offer that better defensive teams allow lesser quality shots and that this leads to meaningful rises in SV%? Most studies that look at a team's ability to suppress shot quality reveal that the effect is negligible.
Miller's SV% has gone down since the trade, not up. He was .924 in Buffalo and is .918 in St. Louis. And I invite you to check out the GAA and SV% of Jaroslav Halak who went from a low shots against team to a high shots against team. The results are in complete disagreement to your argument.
What evidence can you offer that better defensive teams allow lesser quality shots and that this leads to meaningful rises in SV%? Most studies that look at a team's ability to suppress shot quality reveal that the effect is negligible.
The formula I posted for GAA can't get any more accurate because... it is the formula for computing GAA.
Just to follow up - http://somekindofninja.com/nhl/ tracks shot locations and an example application: http://www.dailyfaceoff.com/39366/adjusted-save-percentage-measuring-the-impact-defense-has-on-goaltending-statistics It looks like some of the needed data is starting to become available. Shot location isn't the only factor in save difficulty, but I'd bet it is at least among the largest correlated contributors.