Normalizing shutouts

[kmad]

Has there ever been any effort to normalize shutouts?

I look at the all-time shutout leaders and it's riddled with goalies who played before the forward pass. John Ross Roach, Lorne Chabot, George Hainsworth, etc. Clearly this is disproportionate to their relative skill level as goalies.

And it's the opposite for the 80s. Guys like Fuhr, Smith, Barrasso, etc should probably have more shutouts than they currently do.

As someone unfamiliar with data manipulation, I need to ask - what's the best way to go about normalizing shutouts? Is it even possible?

Click to read more...

 

[Infinite Vision*]

As far as I can see it would be like normalizing hat tricks for players, fairly hard to do IMO.

 

[seventieslord]

i think any normalization would have something to do with assuming goals happen at a random, laplace distribution. You'd then need to look at the average shots per game and sv% by season to determine the probabilities of getting shutouts. (But compare that to the actual shutout totals and if it's too far off, then there is much more at play, such as playing to the score, for example)

 

[Iain Fyffe]

If anyone were to have attempted it, it would have been someone like me. But I find shutouts to be a garbage stat, worthless in terms of information, so I have no interest in working with it. I know a big deal is made about shutout records (Sawchuk), but as a stat it's not worth your time.

 

[TheDevilMadeMe]

Shutouts aren't worthless. Every shutout is a guaranteed point in the standings.

But i agree that it isn't a useful enough stat to bother with some complicated normalization math.

 

[Iain Fyffe]

They're worthless in that they're completely one-sided. With no equivalent stat to tell you how many times the goalie allowed so many goals that his team was guaranteed not to get a point, it's meaningless.

The meaningful part of the shutout stat (allowing few goals) is already in the GAA. Shutouts add nothing to it.

 

[Canadiens1958]

And That Would Be..

They're worthless in that they're completely one-sided. With no equivalent stat to tell you how many times the goalie allowed so many goals that his team was guaranteed not to get a point, it's meaningless.

The meaningful part of the shutout stat (allowing few goals) is already in the GAA. Shutouts add nothing to it.

And that would be how many goals in a game? Given that there have been NHL games where the winning team gave up 7 goals you would be hard pressed to find a sustainable number at the high end.
So even if you find that the high end would be 7 + x it would be rather meaningless since what is important is knowing which goalie does not flinch in the 1-0 or 2-0 games. Knowing that in a 9-7 game that a goalie may flinch is a so what stat. He has flinched 7 times during the game.

The meaningful part of the shutout stat is allowing zero goals.Knowing that the goalie does not flinch especially at key times.

 

[Bear of Bad News]

The meaningful part of the shutout stat is allowing zero goals.Knowing that the goalie does not flinch especially at key times.

How do you know that, just because it's a shutout, it's a "key time"?

How does a shutout in a 5-0 victory demonstrate that a goaltender does not flinch at key times?

How does a 14-save shutout demonstrate that a goaltender does not flinch at key times.

If you take two goaltenders who have played two games apiece, each with a goals-against average of 3.00, and one has a shutout, then that goaltender gave up 6 goals in the non-shutout game. Which goaltender has better demonstrated that he does not flinch at key times?

Iain's right - everything you need to know about shutouts is already boiled into GAA and save percentage.

 

[Canadiens1958]

Okay

How do you know that, just because it's a shutout, it's a "key time"?

How does a shutout in a 5-0 victory demonstrate that a goaltender does not flinch at key times?

How does a 14-save shutout demonstrate that a goaltender does not flinch at key times.

If you take two goaltenders who have played two games apiece, each with a goals-against average of 3.00, and one has a shutout, then that goaltender gave up 6 goals in the non-shutout game. Which goaltender has better demonstrated that he does not flinch at key times?

Iain's right - everything you need to know about shutouts is already boiled into GAA and save percentage.

Clearly stated in 1-0 and 2-0 games not 5-0.

Two game sample space usually does not reveal anything about any team, player, stat, tendency or whatever you wish to look at.

As for the bolded, clearly it is not, since teams breakdown the stats you refer to into components - low scoring games(0/1/2 goals allowed), etc for internal evaluations.

 

[Bear of Bad News]

Clearly stated in 1-0 and 2-0 games not 5-0.

It's a shame, then, that the shutout statistic doesn't split along those lines. A 2-0 shutout is the same as a 1-0 shutout is the same as a 5-0 shutout, according to the statistic we're discussing.

Two game sample space usually does not reveal anything about any team, player, stat, tendency or whatever you wish to look at.

I used a two-game period to make the math easier to follow - it should be obvious to the reader that the situation is identical in a full-season (or full-career) sample.

 

[Iain Fyffe]

It's a shame, then, that the shutout statistic doesn't split along those lines. A 2-0 shutout is the same as a 1-0 shutout is the same as a 5-0 shutout, according to the statistic we're discussing.

This is the problem. If your team scores 8 times, giving up one or two or three or four goals is just as good as giving up zero, since you'll still get your points.

A shutout by itself does not mean that the goaltender is good in key situations or whatever. Crappy goalies can get shutouts against crappy teams in the midst of crappy seasons. Trying to derive some sort of "clutch" information from basic counting stats is doomed to failure.

 

[Canadiens1958]

Statistics

It's a shame, then, that the shutout statistic doesn't split along those lines. A 2-0 shutout is the same as a 1-0 shutout is the same as a 5-0 shutout, according to the statistic we're discussing.



I used a two-game period to make the math easier to follow - it should be obvious to the reader that the situation is identical in a full-season (or full-career) sample.

The shutout statistic does not have to split along these or any lines or end at certain points. The shutout statistic, like all statistics, is simply a starting point for those who wish to inquire further, looking for meaning that may not be obvious at first glance or insufficient data.

 

[Iain Fyffe]

The shutout statistic, like all statistics, is simply a starting point for those who wish to inquire further, looking for meaning that may not be obvious at first glance or insufficient data.

So we've gone from knowledge that the goalie doesn't flinch at key times, to just a starting point for further inquiry?

 

[Canadiens1958]

No..................

So we've gone from knowledge that the goalie doesn't flinch at key times, to just a starting point for further inquiry?

No.........more like going from your view of a garbage stat to one that is multi-faceted with loads of potential.

Two shoe salesmen visiting a developing country. One sees no potential because the people do not wear shoes while the other sees great potential because the people will need shoes.

 

[Nalyd Psycho]

Iain's right - everything you need to know about shutouts is already boiled into GAA and save percentage.

And W/L record.

 

[Iain Fyffe]

No.........more like going from your view of a garbage stat to one that is multi-faceted with loads of potential.

Multi-faceted? It means one thing and one thing only, much of which is often out of the goalie's control.

 

[Canadiens1958]

Shutout Snippet

Multi-faceted? It means one thing and one thing only, much of which is often out of the goalie's control.

Let's look at a shutout snippet to illustrate the multi-faceted nature of the stat:

Glenn Hall and Jacques Plante comparison:

http://www.hockey-reference.com/players/h/hallgl01.html

Jacques Plante
http://www.hockey-reference.com/players/p/plantja01.html

Glenn Hall had more career regular season shutouts than Jacques Plante (84/82)

Playoffs, interestingly Plante has 14 shutouts to 6 for Hall. Even though Plante played 3 fewer playoff games. Another little fact is that the majority of Plante's shut outs were on the road 8-6,Effectively Plante had more road shutouts in the playoffs on the road in app half the number of games than Hall managed home and away

Regular season their team mates and opponents were relatively constant with Plante having the advantage of not having to face the O6 Canadiens. Playoffs same team mates, sufficient sample space > 110 games yet one performed much better than the other. You may be pointed in the right direction by GAA but it would be far from conclusive.

 

[Bear of Bad News]

No.........more like going from your view of a garbage stat to one that is multi-faceted with loads of potential.

You've still demonstrated zero facets and no potential.

 

[Iain Fyffe]

Regular season their team mates and opponents were relatively constant with Plante having the advantage of not having to face the O6 Canadiens. Playoffs same team mates, sufficient sample space > 110 games yet one performed much better than the other. You may be pointed in the right direction by GAA but it would be far from conclusive.

So Plante is better in career GAA, better in career regular-season shutouts per game (using raw totals in player comparisons is much less relevant than rates), and much better in post-season shutout rates. He also played with better teams. Plante also had the better career save percentage, .920 to .917.

(As an aside, the playoff games are not a sufficient sample size for a fair comparison. In Jim Carey's first 99 games, he looked like a good goalie, going 53-30-12 with a .908 save percentage. If 112 games is a good sample size, surely 99 is as well. But for the rest of Carey's career (73 games), he was 26-35-4 with an .884 save percentage. Those first 99 games were not really reflective of Carey's total performance.)

I'm not sure what additional information you think the shutouts give us. If anything, the playoff shutout rates exaggerate Plante's numerical advantage over Hall.

 

[Canadiens1958]

Apples to Apples

So Plante is better in career GAA, better in career regular-season shutouts per game (using raw totals in player comparisons is much less relevant than rates), and much better in post-season shutout rates. He also played with better teams. Plante also had the better career save percentage, .920 to .917.

(As an aside, the playoff games are not a sufficient sample size for a fair comparison. In Jim Carey's first 99 games, he looked like a good goalie, going 53-30-12 with a .908 save percentage. If 112 games is a good sample size, surely 99 is as well. But for the rest of Carey's career (73 games), he was 26-35-4 with an .884 save percentage. Those first 99 games were not really reflective of Carey's total performance.)

I'm not sure what additional information you think the shutouts give us. If anything, the playoff shutout rates exaggerate Plante's numerical advantage over Hall.

Apples to apples or in this case playoff games to playoff games.Given that no goalie has even played 250 playoff games.Sonce Plante and Hall rank in the top fifteen GP the data is more than sufficient for playoff comparisons.

http://www.hockey-reference.com/leaders/games_goalie_career_p.html

http://www.hockey-reference.com/leaders/shutouts_career_p.html


The shutouts give a true picture of the goalies performance against the better or playoff teams. You use the word exaggerate I would say they emphasize Plante's superiority to Hall clearly stating that what may seem close is actually far apart. Especially when you add that Plante is around 5th for playoff shutouts while Hall is around 30th.

The shutouts give a true picture of the goalies performance against the better or playoff teams. You use the word exaggerate I would say they emphasize Plante's superiority to Hall clearly stating that what may seem close is actually far apart. Which would be just one facet. Another facet when looking at goalie comparisons would be in a historic fashion. Asking how many goalies and off what caliber position themselves between the two Plante and Hall in a key area. When app 25 sneak in the door then the claim of fairly close should be revised.

The playoff shutouts give a true picture of the goalie's performance against the better or playoff teams. You use the word exaggerate I would say they emphasize Plante's superiority to Hall clearly stating that what may seem close when blended is actually far apart. Prime example being the SV% difference effectively the gap between .920 and ..917 is marginal yet if sufficient analysis using shutouts as a base then it becomes rather clear whuy Plante was the vastly superior goalie with the key honours.

 

[Iain Fyffe]

Apples to apples or in this case playoff games to playoff games.Given that no goalie has even played 250 playoff games.Sonce Plante and Hall rank in the top fifteen GP the data is more than sufficient for playoff comparisons.

Which is why playoff games are often left out of the analysis, or at least aggregated with regular season totals. For the vast majority of players, there are just too few playoff games available to be of real meaning.

The shutouts give a true picture of the goalies performance against the better or playoff teams.

No, a goalie's save percentage against those teams would give a better indication of that. Again, why would you look at shutouts when you have other, better stats? Look at career playoff GAA: Plante 2.14, Hall 2.78. Do we need shutouts to know that Plante performed better in the playoffs? It's right there in the GAA. I'm sure it's in the save percentage too.

The shutouts give a true picture of the goalies performance against the better or playoff teams.

Why would we focus only on those 14 playoff game for Plante, and 6 for Hall? Why wouldn't we look at their overall performance, using save percentage for instance?

Another facet when looking at goalie comparisons would be in a historic fashion. Asking how many goalies and off what caliber position themselves between the two Plante and Hall in a key area. When app 25 sneak in the door then the claim of fairly close should be revised.

I'm having trouble parsing this, but if I read you correctly, GAA tells you everything that you say the shutouts do: Plante performed much better in the playoffs than Hall did.

Prime example being the SV% difference effectively the gap between .920 and ..917 is marginal yet if sufficient analysis using shutouts as a base then it becomes rather clear whuy Plante was the vastly superior goalie with the key honours.

Over 27,000 shots, a .003 difference in save percentage is worth about 80 goals.

So, basically Plante has somewhat better regular-season numbers than Hall, and significantly better playoff numbers. Aggregate them and Plante has better-than-somewhat-better numbers than Hall. And we don't need shutouts to tell us any of this.

Where have the shutout totals led us anywhere that save percentage or even GAA do not? For the record, Plante's playoff save percentage is .922 and Hall's is .911.

Now, your point seems to be that Plante's playoff performance prove that he is more better than Hall than their regular-season numbers would suggest. But have you factored in the fact that the average regular-season winning percentage of Hall's playoff opponents is .577, while Plante's is .525? Hall played against better teams in the playoffs than Plante, largely because Plante did not have to face the Montreal Canadiens.

Perhaps Plante recorded so many more playoff shutouts because he was playing against weaker teams.

 

[Canadiens1958]

So What Really Matters Does Not

Which is why playoff games are often left out of the analysis, or at least aggregated with regular season totals. For the vast majority of players, there are just too few playoff games available to be of real meaning.


No, a goalie's save percentage against those teams would give a better indication of that. Again, why would you look at shutouts when you have other, better stats? Look at career playoff GAA: Plante 2.14, Hall 2.78. Do we need shutouts to know that Plante performed better in the playoffs? It's right there in the GAA. I'm sure it's in the save percentage too.


Why would we focus only on those 14 playoff game for Plante, and 6 for Hall? Why wouldn't we look at their overall performance, using save percentage for instance?


I'm having trouble parsing this, but if I read you correctly, GAA tells you everything that you say the shutouts do: Plante performed much better in the playoffs than Hall did.


Over 27,000 shots, a .003 difference in save percentage is worth about 80 goals.

So, basically Plante has somewhat better regular-season numbers than Hall, and significantly better playoff numbers. Aggregate them and Plante has better-than-somewhat-better numbers than Hall. And we don't need shutouts to tell us any of this.

Where have the shutout totals led us anywhere that save percentage or even GAA do not? For the record, Plante's playoff save percentage is .922 and Hall's is .911.

Now, your point seems to be that Plante's playoff performance prove that he is more better than Hall than their regular-season numbers would suggest. But have you factored in the fact that the average regular-season winning percentage of Hall's playoff opponents is .577, while Plante's is .525? Hall played against better teams in the playoffs than Plante, largely because Plante did not have to face the Montreal Canadiens.

Perhaps Plante recorded so many more playoff shutouts because he was playing against weaker teams.

Your first paragraph sums up where we diverge. If playoffs do not matter then there is little point to the exercise be it comparing goalies, forwards, d-men, teams, coaches or any perm and comb of the aforementioned.

No one is focusing on 14 games for Plante and 6 games for Hall A key indicator has been identified when looking at two goalies who were contemporaries playing roughly the same number of playoff games. Conversely your point about SV% is a red herring if put in the context of the debate.


You conveniently trot out the winning percentage for the opposing teams, effectively measuring degree of difficulty. Are you prepared to offer degree of difficulty numbers for SV%? Very doubtful. Yet the playoffs from the last two seasons, specifically the performance of Michael Leighton in 2010 clearly showed that SV% is far from the stand alone stat that people want it to be. So again if degree of difficulty is an important factor then it is always an important factor and should be included with SV% data.

Back to shutouts. Just like goals for a forward they are the eye catchers that direct analysis.

That you prefer a career overview that blends achievements is a minor entitlement that is academically acceptable. That others prefer to dig deeper and explore at another level is a similar minor entitlement that is academically acceptable.

Pepsi or Coca-cola, either way the appreciation of the history of hockey is advanced.

 

[Iain Fyffe]

Your first paragraph sums up where we diverge. If playoffs do not matter then there is little point to the exercise be it comparing goalies, forwards, d-men, teams, coaches or any perm and comb of the aforementioned.

It's not that playoffs "don't matter", it's that there are so few playoff games compared to regular-season games that they are generally not statistically significant by themselves. Ignoring 500 games in favour of 50 is analytically bizarre.

You conveniently trot out the winning percentage for the opposing teams, effectively measuring degree of difficulty. Are you prepared to offer degree of difficulty numbers for SV%? Very doubtful.

I'm using opponent winning percentage as a proxy for difficulty here. You overlooked the context in which the shutouts were recorded, I'm trying to add some context.

Yet the playoffs from the last two seasons, specifically the performance of Michael Leighton in 2010 clearly showed that SV% is far from the stand alone stat that people want it to be.

I assume you mean the fact that Leighton had a .916 save percentage over a 14-game span in the 2010 playoffs, and that this shows he's a better goalie than his regular-season record? I'm guessing, based on your previous posts. If this is the case, it demonstrated the problem of using playoff stats by themselves for analysis. Why would these 14 games be his true performance level, and not the 103 regular-season game, .902 save percentage preformance? His .862 save percentage in the 2011 playoffs are a hint that the .916 is fluky. I'm sure he had 14-game stretches in the regular season where his s% was at least .916.

Back to shutouts. Just like goals for a forward they are the eye catchers that direct analysis.

Again, what do shutouts do that other stats do not already do better? You can just as easily say a high playoff save percentage is en eye-catcher that directs analysis.

That you prefer a career overview that blends achievements is a minor entitlement that is academically acceptable. That others prefer to dig deeper and explore at another level is a similar minor entitlement that is academically acceptable.

Considering that I'm the only one in this thread that's actually been doing any digging, I find this humorous.

 

[Iain Fyffe]

Here's a little more context.

Plante's 14 playoff shutouts came against teams with an average winning percentage of .496.

Hall's 6 playoff shutouts came against teams with an average winning percentage of .623.

If you're trying to use shutouts to prove which goaltender truly played better when it really counts, this would be a big point in favour of Hall.

Plante picked up a good number of his shutouts against relatively weak teams, which shouldn't be surprising since he played for the Habs who were always at the top of the standings and got the easiest matchups.

Hall earned his shutouts against the best teams in the league. In 1961 he shut out the #1 Habs twice. In 1962 he shut out the #1 Habs and the #2 Leafs. In 1965 he shut out the #1 Red Wings. His last shutout was in 1968 against the Flyers, who were sub-.500 but were still the #1 team in the West.

This means that all of Hall's playoff shutouts (100%) came against teams with better records than his team. Plante? Only one of his 14 playoff shutouts (7%) came against teams with better records than his team.

Do I interpret this to mean that Hall was actually the better playoff performer? Of course not. Plante often had no choice but to record shutouts against teams with worse records, because his team was often #1. His record, when you consider all things, is slightly but significantly better than Hall's.

But it does say to me that the "more playoff shutouts = proves he was much better" line of thinking is deeply flawed.

In my opinion, the playoff shutout numbers have not added anything to our understanding of these two goalies. In fact, they have tried to mislead us. This is the essence of a garbage stat.

 

[Canadiens1958]

Analytically Bizarre

It's not that playoffs "don't matter", it's that there are so few playoff games compared to regular-season games that they are generally not statistically significant by themselves. Ignoring 500 games in favour of 50 is analytically bizarre.


I'm using opponent winning percentage as a proxy for difficulty here. You overlooked the context in which the shutouts were recorded, I'm trying to add some context.


I assume you mean the fact that Leighton had a .916 save percentage over a 14-game span in the 2010 playoffs, and that this shows he's a better goalie than his regular-season record? I'm guessing, based on your previous posts. If this is the case, it demonstrated the problem of using playoff stats by themselves for analysis. Why would these 14 games be his true performance level, and not the 103 regular-season game, .902 save percentage preformance? His .862 save percentage in the 2011 playoffs are a hint that the .916 is fluky. I'm sure he had 14-game stretches in the regular season where his s% was at least .916.


Again, what do shutouts do that other stats do not already do better? You can just as easily say a high playoff save percentage is en eye-catcher that directs analysis.


Considering that I'm the only one in this thread that's actually been doing any digging, I find this humorous.

First bolded.So all comparables of regular season performance to playoff performance are flawed for the same reason or in the framework of your analytically bizarre. Comparing a forwards PPG numbers regular season vs playoff is then analytically bizarre. But then goals and assists are no different than shutouts, after all they are just numbers to be blended. Yet a star forward being held scoreless is off prime interest because it goes directly to questions about the defense employed. Likewise a weak goalie stringing shutouts goes to defense and offense(see later)

Guessing. Rather obvious that you have not read my previous posts in other threads about the 2010 playoffs.. Keeping it simple and with respect to Michael Leighton. If the Flyers limited the opposition to a perimeter game his SV% was very high, evidenced by shutouts. In the series leading up to the finals Leighton shutout the Canadiens three times yet in the one game the Canadiens won they crashed the net and scored five goals.

Now if the Hawks were scouting the series what do they do? To read your analysis they say "Well 5 games are not sufficient or revealing, nothing to be learned here.". On the other hand if they look at the shutouts they will say" Well if we do not crash the net the Flyers can shutout an NHL team with an average AHL goalie. The Hawks rather obviously learned from the 5 game sample space and won the SC by going to the net at every opportunity. Effectively from a playing the game analysis one side asks what has to be done to minimize the goalies performance while the other side asks what has to be done to maximize the goalies performance. If you can maximize the goalies performance = shutout, in the playoffs it is impossible to lose. A .950SV% may look nice but if the team loses it is an insignificant consolation.

Pro sport is very unforgiving. Once the smallest weakness is learned it will be exploited to win. Teams breakdown all the numbers and look at every event - goal, assist, shutout to get the slightest edge. Blended stats have the opposite effect since they direct the analysis away the key elements. Like the river with an average depth of 4", that gives a false sense of security.