HFBoards

Go Back   HFBoards > General Hockey Discussion > By The Numbers
Mobile Hockey's Future Become a Sponsor Site Rules Support Forum vBookie Page 2
By The Numbers Hockey Analytics... the Final Frontier. Explore strange new worlds, to seek out new algorithms, to boldly go where no one has gone before.

Excellent New Goalie Stat (Save Percentage Above League Average)

Reply
 
Thread Tools
Old
10-10-2013, 11:51 AM
  #26
cheap77
Go Kings!
 
cheap77's Avatar
 
Join Date: Mar 2011
Country: United States
Posts: 2,143
vCash: 500
I don't get it. Math is hard. There is really only 1 stat I care about. Wins.

cheap77 is offline   Reply With Quote
Old
10-10-2013, 11:53 AM
  #27
Trebek
Mod Supervisor
 
Trebek's Avatar
 
Join Date: Sep 2005
Posts: 2,756
vCash: 500
Quote:
Originally Posted by cheap77 View Post
There is really only 1 stat I care about. Wins.
That's like a CEO coming in at her first board meeting and saying that "the only thing that I care about is increasing our stock price."

Well, that's great, but how do you do that?

Yes, goaltenders care about wins, first and foremost. How do they contribute to that goal? (Primarily), by stopping the shots that they face.

Trebek is offline   Reply With Quote
Old
10-10-2013, 11:53 AM
  #28
Bmessy
Registered User
 
Bmessy's Avatar
 
Join Date: Nov 2007
Location: Allston, MA
Posts: 1,487
vCash: 500
Would it be more accurate or informative to filter out the goalies who only play 10 games? 20 games?
If you want to compare the tops in the league, goalies who play all the time, why have the crappy save %s from goalies who rarely play lowering the mean.

So the 129%SV+ for BOBROVSKY!! would probably go down some, but still it's all relative.
I would like to see the numbers for just the top 50 goalies, or even the top goalies who faced a certain number of shots(say 500) or games played.

edit: I now see that Hockeys Reference says "Min. 4 shots faced per team game needed to qualify", which filters out some, I suppose.


Last edited by Bmessy: 10-10-2013 at 12:01 PM.
Bmessy is offline   Reply With Quote
Old
10-10-2013, 11:59 AM
  #29
Follower of Yzlam
OH YEEEEAH
 
Follower of Yzlam's Avatar
 
Join Date: Jun 2011
Country: United States
Posts: 16,169
vCash: 134
Wouldn't it be useful to have it against the median as well? A few exceptionally good (or bad) goalies in a season could skew the results a bit. Maybe make a minimum games played (~25) as the base for the criteria.

Follower of Yzlam is offline   Reply With Quote
Old
10-10-2013, 12:01 PM
  #30
StoneHands
Registered User
 
Join Date: Feb 2013
Posts: 1,369
vCash: 2425
I know it's somewhat off-topic but is there anything for goalies like the shooting charts they have in basketball. I think it would be really interesting to see a breakdown of where shots are taken. It wouldnt be that hard and I find it hard to believe teams don't use that kind of data themselves. It would be nice to have a dot on a map of the ice showing where a shot came from and it could be, say, a blue dot for a regular shot, a red dot for a rebound, and a blue dot with a line connecting to a green dot for a shot and a deflection. Most NHL calibre goalies are going to have a save percentage of .99% on shots from outside of the circles that are not defelected, it would be interesting to see how a particular goalie does within the faceoff dots or around the crease. I know there are shot charts but it doesnt really tell the whole story. Showing a shot, rebound, and deflection, would tell much more and save% breakdowns for each would be amazing.

Is there anything like this out there in stat-land?


Last edited by StoneHands: 10-10-2013 at 12:07 PM.
StoneHands is offline   Reply With Quote
Old
10-10-2013, 12:05 PM
  #31
zeke
#freewilly
 
zeke's Avatar
 
Join Date: Mar 2005
Posts: 30,648
vCash: 500
Quote:
Originally Posted by Bmessy View Post
Would it be more accurate or informative to filter out the goalies who only play 10 games? 20 games?
If you want to compare the tops in the league, goalies who play all the time, why have the crappy save %s from goalies who rarely play lowering the mean.

So the 129%SV+ for BOBROVSKY!! would probably go down some, but still it's all relative.
I would like to see the numbers for just the top 50 goalies, or even the top goalies who faced a certain number of shots(say 500) or games played.

edit: I now see that Hockeys Reference says "Min. 4 shots faced per team game needed to qualify", which filters out some, I suppose.
You can filter the numbers any which way you want using their "Play Index" up a the top left of the page.

zeke is offline   Reply With Quote
Old
10-10-2013, 12:06 PM
  #32
zeke
#freewilly
 
zeke's Avatar
 
Join Date: Mar 2005
Posts: 30,648
vCash: 500
Quote:
Originally Posted by Taco MacArthur View Post
I've been doing exactly this for quite some time now. For instance:

http://hockeygoalies.org/bio/nhl/toronto.html
(scroll down to where save percentage has been calculated)

Z-Score tells how many standard deviations above (or below) league average a goaltender's performance was during the season.

GD tells how many goals a goaltender prevented during the season, above and beyond the league average (looking at the link above, it appears that H-R is doing this now, calling it GSAA).

GAR is the same as above, but comparing against a "replacement-level" goaltender (since GD would assign a value of zero to a league-average goaltender, but there's definite value in being average).

SNW% gives what the goaltender's winning percentage would be if he were on a team that (a) faces a league-average number of shots, and (b) scores a league-average number of goals. (SNW and SNL are then a goaltender's support-neutral wins and losses).

For more recent seasons, I also calculate a goaltender's game-to-game variance (as a measure of consistency) as well as the strength of schedule faced by the goaltender.

I describe these all in a bit more detail here: http://hockeygoalies.org/stats/glossary.html
nice work. those are all great stats too.

zeke is offline   Reply With Quote
Old
10-10-2013, 02:00 PM
  #33
Bmessy
Registered User
 
Bmessy's Avatar
 
Join Date: Nov 2007
Location: Allston, MA
Posts: 1,487
vCash: 500
Quote:
Originally Posted by zeke View Post
You can filter the numbers any which way you want using their "Play Index" up a the top left of the page.
Good looks! That thing is awesome, I could have some fun with that.

So I sorted out the goalie stats from last year, removing any goaltenders with less than 500 saves.
The original SV%+ for Bobrovsky with all goaltenders in the league was 129.
After my removal of those goaltenders it brought it down to 118.

Bmessy is offline   Reply With Quote
Old
10-10-2013, 02:14 PM
  #34
TheStranger
Registered User
 
TheStranger's Avatar
 
Join Date: Jan 2010
Location: Ottawa, Ontario
Country: Canada
Posts: 17,648
vCash: 500
Quote:
Originally Posted by 5 Minute Major View Post
I wish there was a stat kept for quality scoring chance save percentage. Maybe there is and I don't know that?
I remember someone bringing up a stat on Varlamov in his first year with Colorado, saying he a .960sv% on shots with 15 feet or something. That would be cool information too if someone had it.

TheStranger is offline   Reply With Quote
Old
10-10-2013, 06:16 PM
  #35
invictus
Registered User
 
Join Date: Mar 2007
Posts: 16,639
vCash: 500
Quote:
Originally Posted by Taco MacArthur View Post
Bringing a few things to the table to help show why save percentage translations are important.

Here's a list of all Vezina Trophy winners:
http://hockeygoalies.org/awards/vezina.html

Over time, you can see the raw (unadjusted) save percentages fluctuate, as periods of offense dominated, and periods of defense dominated. Did all of the goaltenders in the 1980s suck? Maybe, but unlikely.

If you look at the adjusted metrics (Z-Score, Goal Differential, and Goals Above Replacement), you can still see some outliers - this is purportedly the list of each year's best goalie, after all - but the pattern evens out a bit.

It's also easier to pick out the truly remarkable performances. Note Dominik Hasek's 93% (raw) save percentage in 1993-94 - if you look at his z-score, you can see that his performance was 4.6 standard deviations above the league average. What "z score" measures is the likelihood that an average goaltender would put together a season like the one in question (since we've all seen average goaltenders put together stretches of great play). A z-score of 4.6 means that an average goaltender would reproduce Hasek's 1993-94 season about once every 475,000 years.

If you click through (on the above page), you can see the individual goaltenders' careers evolve (under REGULAR SEASON STATISTICS or POSTSEASON STATISTICS). It helps to compare (for instance) Patrick Roy's career with Dominik Hasek's career with Martin Brodeur's career (and you can see that each was remarkable in his own way).
That is insane! Is there a way to see how many hundreds of thousands of years it will take someone to be as bad as Fleury was in the playoffs against the Flyers 2 seasons ago?

I love the Penguins, but seeing this brand new metric showing how MAF sucks makes me chuckle. He is only 22nd! Haha

invictus is offline   Reply With Quote
Old
10-10-2013, 06:32 PM
  #36
Trebek
Mod Supervisor
 
Trebek's Avatar
 
Join Date: Sep 2005
Posts: 2,756
vCash: 500
Quote:
Originally Posted by invictus View Post
That is insane! Is there a way to see how many hundreds of thousands of years it will take someone to be as bad as Fleury was in the playoffs against the Flyers 2 seasons ago?
Good question - taking from here:
http://hockeygoalies.org/bio/fleury.html

(under POSTSEASON STATISTICS)

Fleury's 83.4% save percentage was 4.2 standard deviations lower than the average performance in the 2012 Stanley Cup playoffs, as a league-average save percentage in the 2012 playoffs was a beefy 92.1%. A goaltender with a long-term 92.1% save percentage will put together a stretch like Fleury's about every 75,000 seasons.

However...

It's also important to consider the opponents that Fleury was facing (it's less important in the regular season, when teams essentially all play each other, and no one plays a single opponent an inordinate amount of times). Fleury faced only the Flyers, and if you look all the way to the right on Fleury's 2012 postseason line, you can see that his opponent-weighted expected save percentage was 0.900. Stated differently, the Flyers had a 10% shooting percentage last year (other than empty-net goals), which is quite high for this era.

So it's perhaps not fair to expect Fleury to average 92.1% in his playoffs year, but a more modest 90.0%. Against that benchmark, he was still about 2.7 standard deviations below average, which would put his performance at about once in 288 postseasons (still bad but not horrendous).

Trebek is offline   Reply With Quote
Old
10-10-2013, 06:41 PM
  #37
invictus
Registered User
 
Join Date: Mar 2007
Posts: 16,639
vCash: 500
Quote:
Originally Posted by Taco MacArthur View Post
Good question - taking from here:
http://hockeygoalies.org/bio/fleury.html

(under POSTSEASON STATISTICS)

Fleury's 83.4% save percentage was 4.2 standard deviations lower than the average performance in the 2012 Stanley Cup playoffs, as a league-average save percentage in the 2012 playoffs was a beefy 92.1%. A goaltender with a long-term 92.1% save percentage will put together a stretch like Fleury's about every 75,000 seasons.

However...

It's also important to consider the opponents that Fleury was facing (it's less important in the regular season, when teams essentially all play each other, and no one plays a single opponent an inordinate amount of times). Fleury faced only the Flyers, and if you look all the way to the right on Fleury's 2012 postseason line, you can see that his opponent-weighted expected save percentage was 0.900. Stated differently, the Flyers had a 10% shooting percentage last year (other than empty-net goals), which is quite high for this era.

So it's perhaps not fair to expect Fleury to average 92.1% in his playoffs year, but a more modest 90.0%. Against that benchmark, he was still about 2.7 standard deviations below average, which would put his performance at about once in 288 postseasons (still bad but not horrendous).
Hmm, that is actually pretty interesting. A lot of "new" stats seem to be a lot of smoke and mirrors, but this one seems potentially quality. I look forward to paying attention to it as this season progresses.

invictus is offline   Reply With Quote
Old
10-10-2013, 06:44 PM
  #38
Trebek
Mod Supervisor
 
Trebek's Avatar
 
Join Date: Sep 2005
Posts: 2,756
vCash: 500
Thanks! Hopefully it does what it's intended to do - it's certainly not perfect, but it tries to view the performance in a different lens.

Trebek is offline   Reply With Quote
Old
10-11-2013, 04:13 PM
  #39
Trebek
Mod Supervisor
 
Trebek's Avatar
 
Join Date: Sep 2005
Posts: 2,756
vCash: 500
This has dropped down a few pages on the NHL board, so I moved it here.

Trebek is offline   Reply With Quote
Old
10-11-2013, 11:11 PM
  #40
Laos
Registered User
 
Laos's Avatar
 
Join Date: Apr 2008
Location: Charlesbourg, Québec
Country: Hungary
Posts: 270
vCash: 500
Quote:
Originally Posted by Taco MacArthur View Post
Good question - taking from here:
http://hockeygoalies.org/bio/fleury.html

(under POSTSEASON STATISTICS)

Fleury's 83.4% save percentage was 4.2 standard deviations lower than the average performance in the 2012 Stanley Cup playoffs, as a league-average save percentage in the 2012 playoffs was a beefy 92.1%. A goaltender with a long-term 92.1% save percentage will put together a stretch like Fleury's about every 75,000 seasons.

However...

It's also important to consider the opponents that Fleury was facing (it's less important in the regular season, when teams essentially all play each other, and no one plays a single opponent an inordinate amount of times). Fleury faced only the Flyers, and if you look all the way to the right on Fleury's 2012 postseason line, you can see that his opponent-weighted expected save percentage was 0.900. Stated differently, the Flyers had a 10% shooting percentage last year (other than empty-net goals), which is quite high for this era.

So it's perhaps not fair to expect Fleury to average 92.1% in his playoffs year, but a more modest 90.0%. Against that benchmark, he was still about 2.7 standard deviations below average, which would put his performance at about once in 288 postseasons (still bad but not horrendous).
Just being curious on how you get your numbers like" one every 75000 seasons" from your standard deviation?

Laos is offline   Reply With Quote
Old
10-11-2013, 11:27 PM
  #41
Trebek
Mod Supervisor
 
Trebek's Avatar
 
Join Date: Sep 2005
Posts: 2,756
vCash: 500
Quote:
Originally Posted by Laos View Post
Just being curious on how you get your numbers like" one every 75000 seasons" from your standard deviation?
Good question. If we assume that shots on net follow a binomial distribution (admittedly a stretch), then as the number of shots faced increases, the distribution can be approximated by a normal distribution.

At that point, and with a z-score, it's just a matter of measuring the area to the right (in the case of a positive z-score) or to the left (in the case of a negative z-score).

For instance, consider a goaltender with a z-score of -1.48 (since that's an image that I was able to swipe online ):



There is 6.94% of the curve to the left of the goaltender's performance, so we would expect a goaltender to do as poorly as he did (or worse) 6.94% of the time. So it would happen about every 14.4 seasons (since 1/6.94% = 14.4).

Does that make sense? Once we assume a binomial distribution (which isn't perfect, but it's not horrible), then it's just probability theory.

Trebek is offline   Reply With Quote
Old
10-12-2013, 07:55 PM
  #42
Laos
Registered User
 
Laos's Avatar
 
Join Date: Apr 2008
Location: Charlesbourg, Québec
Country: Hungary
Posts: 270
vCash: 500
Quote:
Originally Posted by Taco MacArthur View Post
Good question. If we assume that shots on net follow a binomial distribution (admittedly a stretch), then as the number of shots faced increases, the distribution can be approximated by a normal distribution.

At that point, and with a z-score, it's just a matter of measuring the area to the right (in the case of a positive z-score) or to the left (in the case of a negative z-score).

For instance, consider a goaltender with a z-score of -1.48 (since that's an image that I was able to swipe online ):



There is 6.94% of the curve to the left of the goaltender's performance, so we would expect a goaltender to do as poorly as he did (or worse) 6.94% of the time. So it would happen about every 14.4 seasons (since 1/6.94% = 14.4).

Does that make sense? Once we assume a binomial distribution (which isn't perfect, but it's not horrible), then it's just probability theory.
Makes perfect sense. Thank you so much for taking the time to answer me!

Laos is offline   Reply With Quote
Old
10-12-2013, 08:00 PM
  #43
Trebek
Mod Supervisor
 
Trebek's Avatar
 
Join Date: Sep 2005
Posts: 2,756
vCash: 500
Very happy to help!

Trebek is offline   Reply With Quote
Old
10-12-2013, 11:03 PM
  #44
Lazyking
Never Forget
 
Lazyking's Avatar
 
Join Date: Oct 2011
Location: Connecticut
Country: United States
Posts: 3,737
vCash: 500
I love goalies so I'm really glad I found your site Taco. However, I think even with all the stats, it's at times really hard to evaluate a goalie's performance with all the lucky goals and goals that the defense hangs the goalie out to dry.

Which is why I wish there was a metric to look at scoring chances, and where the goal comes from.

Lazyking is offline   Reply With Quote
Old
10-13-2013, 12:17 AM
  #45
Trebek
Mod Supervisor
 
Trebek's Avatar
 
Join Date: Sep 2005
Posts: 2,756
vCash: 500
Thanks!

I agree - if there was a standard definition for scoring chances, I'd love to see that metric. The various risk adjusted save percentages (or variations) that are out there are a nice compromise in my opinion.

Trebek is offline   Reply With Quote
Old
10-15-2013, 07:03 AM
  #46
ted1971
History Of Hockey
 
ted1971's Avatar
 
Join Date: Oct 2008
Location: south jersey
Country: United States
Posts: 2,010
vCash: 500
A lot of these " advanced stats" make no sense to me.

ted1971 is offline   Reply With Quote
Old
10-15-2013, 10:51 AM
  #47
Trebek
Mod Supervisor
 
Trebek's Avatar
 
Join Date: Sep 2005
Posts: 2,756
vCash: 500
Hi Ted (or anyone else, for that matter) - if you pick one, I'd be happy to walk you through it (please pick a goalie one, because otherwise I'm screwed ).

These are the ones that I use on my site:
http://hockeygoalies.org/stats/glossary.html

As a term, "advanced" statistics is a bit of a lark - these statistics are only called "advanced" because they aren't in the common language yet. For many years, save percentage was considered an "advanced" statistic.

Trebek is offline   Reply With Quote
Old
10-15-2013, 03:14 PM
  #48
Saved*
 
Join Date: May 2008
Location: Canada
Country: Canada
Posts: 2,830
vCash: 500
Quote:
Originally Posted by HarrisonFord View Post
So 1.15SV+ = 15% above average?
No, it merely means that the goalie was 1.15 standard deviations above the mean, as calculated with whatever metrics the statician decided to utilize. Z-Score of 1.0 is equal to 1 standard deviation above the mean, which correlates to a goaltender playing in the 84.13th percentile.

If that 1.15 is equal to a z-score, that means that the goaltender played in the 87.49th percentile.

Saved* is offline   Reply With Quote
Old
10-15-2013, 04:42 PM
  #49
MadLuke
Registered User
 
Join Date: Jan 2011
Posts: 2,389
vCash: 500
Quote:
Originally Posted by Fire Sather View Post
I'll watch the games
With 30 teams and some interresting duo, this is almost impossible to do anymore.

If you want to have a small 15 games for each goaltender..... plus follow your favorite team you jump at watching 300 regular season games....

Trophy in the 06 league were probably better and less on stat and reputation, but nowaday who watch every player on a good sample size ?

MadLuke is offline   Reply With Quote
Old
10-15-2013, 04:49 PM
  #50
Trebek
Mod Supervisor
 
Trebek's Avatar
 
Join Date: Sep 2005
Posts: 2,756
vCash: 500
Quote:
Originally Posted by Saved View Post
No, it merely means that the goalie was 1.15 standard deviations above the mean, as calculated with whatever metrics the statician decided to utilize. Z-Score of 1.0 is equal to 1 standard deviation above the mean, which correlates to a goaltender playing in the 84.13th percentile.

If that 1.15 is equal to a z-score, that means that the goaltender played in the 87.49th percentile.
That's what I do with the z-score calculation on my site (which I prefer to the SV+ metric, since it accounts for small sample sizes).

For the SV+ metric, it's basically:

(1 - league average save pct) / (1 - goalie's save pct)

Or stated differently:

(league average error rate) / (goaltender's error rate)

Using the link in the original post, league average save percentage in 2012-13 was 0.912 (I took this from my own database, but if you see Jonas Hiller's 101 SV+ and 0.913 save percentage, it's about what hockey-reference has as well).

Now look at Craig Anderson's 0.941 save percentage. His SV+ is equal to:

(1 - 0.912) / (1 - 0.941) = (0.088) / (0.059) = 1.49, or "149" SV+.

This is still a translation of save percentage, so a goaltender with three saves on three shots will look just as good under SV+ as a goaltender with 300 saves on 300 shots. H-R manages this by censoring the results for goaltenders below a threshold.

Z-Score has sample size built in - it basically says, if a goaltender is truly league average (0.912 save percentage), then how remarkable is it to have seen a 3 saves/3 shots performance? How remarkable is it to have seen a 30 saves/30 shots performance? How remarkable is it to have seen a 300 saves/300 shots performance? (I think we'll agree that the latter is far more improbable).

In the first case, I would calculate a z-score of 0.54 (the goaltender has performed about a half-standard deviation better than average). A league-average goaltender will stop three out of three shots about once in every 3.4 times.

In the second case, the goaltender has a z-score of 1.7. A league average goaltender will stop 30 of 30 shots about once in 22.5 times. (Stated differently, if a league-average goaltender faces exactly 30 shots in each game, they will have a shutout about once in 22.5 games).

In the third case, the goaltender has a z-score of 5.4, and a league average goaltender will stop 300 of 300 shots about once in 26.9 million times. (Putting this into context, Brian Boucher had a stretch of five consecutive games with a shutout. Over these five games, he faced "only" 130 shots).

The goaltender who does the latter feat (300-for-300) has given us astronomical evidence that he is not a league average goaltender, and that's what z-score measures.

Trebek is offline   Reply With Quote
Reply

Forum Jump


Bookmarks

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off



All times are GMT -5. The time now is 07:05 AM.

monitoring_string = "e4251c93e2ba248d29da988d93bf5144"
Contact Us - HFBoards - Archive - Privacy Statement - Terms of Use - Advertise - Top - AdChoices

vBulletin Copyright ©2000 - 2015, Jelsoft Enterprises Ltd.
HFBoards.com is a property of CraveOnline Media, LLC, an Evolve Media, LLC company. ©2015 All Rights Reserved.