NHL goalies better vs high shot volumes

Discussion in 'By The Numbers' started by Mathletic, Mar 26, 2014.

1. Doctor NoRegistered User

Joined:
Oct 26, 2005
Messages:
8,098
1,636
Trophy Points:
149
Okay, then.

Pick one thing that you want us to understand as your point.

Not eleven things obfuscated over a fifteen-part post, each partially responding to various responses. Quite honestly, it's exhausting.

Pick one single thing that we can discuss, or I'm not coming back to the conversation. One point that you want us to understand. We can discuss other points after we've solved one point.

2. Hockey OutsiderRegistered User

Joined:
Jan 16, 2005
Messages:
5,353
2,156
Trophy Points:
166
Thanks for calling me out. I've already responded to all of your arguments in the previous thread. You haven't presented anything new, and you're simply re-phrasing all the arguments that have already been addressed.

To be frank, I'm confident that anybody with a basic knowledge of mathematics and hockey who read through the previous thread would understand the issues with your position that I (and others) have repeatedly pointed out.

Good luck with your preaching. I have better things to do with my time.

seventieslord and GreatGonzo like this.

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey
Here's my point. Playing on strong defensive teams that limit shots makes it difficult for goalies to inflate and maintain high SV%'s. Why? There's less of a margin of error when you face a lower amount of shots per game.

If Marty Turco played one more game in 2003 and allowed 2 goals in 24 shots, his SV% would decrease.
If Tim Thomas played one more game in 2009 and allowed 2 goals on 31 shots, his SV% would increase.

It's possible for a goalie to have a deflated higher SV% like Turco in 2003 and Kiprusoff in 2004.
It's possible for a goalie to have an inflated lower SV% like Ryan Miller in 2013 and James Reimer in 2011.

Facing a low number of shots on a regular basis does not make it easier to allow fewer than 2 goals in today's game.

I can without a shadow of a doubt tell you that goalies do have a higher cumulative SV% in their high shot volume games. The facts are the facts. My question to you is, why do you think that's the case?

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey
You did not respond to my last post before the thread was closed.

Actually if you read my last response to you before the thread was closed, you'll see new arguments.

Those issues you brought forward have nothing to do with my stance. You still have not shown that you understand my point. Hell the post you just responded to has arguments from me you have NOT seen.

Can I just ask you one question? Do you believe my information in those tables is incorrect? Do you think I just made it all up?

5. Doctor NoRegistered User

Joined:
Oct 26, 2005
Messages:
8,098
1,636
Trophy Points:
149
I'm tapping out - I didn't think this would continue to be semantics based. I've got hockey practice to get to.

Please don't tag me in this thread again.

Hockey Outsider likes this.

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey
What semantics are you talking about?

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey
Last 3 seasons

NOTE: Make sure you click on the blue links at the bottom of each table to switch back and forth between shot bins.

17-18

16-17

15-16

goeb likes this.

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey
2014-2015

2013-2014

2012-2013

goeb likes this.
9. goebRegistered User

Joined:
Oct 24, 2013
Messages:
227
104
Trophy Points:
51
Gender:
Male
Location:
Grand Rapids, Michigan

No arguing those stats.....I wonder if lower shot quantity leads to higher shot quality?

Or, maybe the goalie who faced less than 20 shots let up a few goals early in the game and the opponent decided to sit back into more of a defensive setup

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey
The theory of what happens is that the number of quality shots stays roughly the same for every team no matter how many shots they get / allow. In other words, whether a team allows 30 shots a game or 25 shots a game, both teams will still allow between 10 and 13 high danger chances per game.

NHL goalies better vs high shot volumes - Sportsnet.ca

Visually we can see where a lot of the extra shots originate from.
The slot volume on the smaller workload is higher. The home plate area is the definition for most scoring-chance projects, so the lighter workload exposes goalies to a higher percentage of scoring chances. Without the save-percentage inflators provided by shots from above the faceoff circles (which have an expected save percentage of .974) it becomes difficult to register anything over .900 when facing lighter than a 20-shot workload.

11. goebRegistered User

Joined:
Oct 24, 2013
Messages:
227
104
Trophy Points:
51
Gender:
Male
Location:
Grand Rapids, Michigan
Yea this definitely makes me rethink the whole corsi thing...and even save %. Back when I was younger (15 years ago lets say), I always thought of save % as the litmus test for defining who was and was not a good goalies. Truth is, it's quite complicated due to the amount of variables involved.

12. Doctor NoRegistered User

Joined:
Oct 26, 2005
Messages:
8,098
1,636
Trophy Points:
149
Citation needed.

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey
2011-2012

2010-2011

2009-2010

15. Doctor NoRegistered User

Joined:
Oct 26, 2005
Messages:
8,098
1,636
Trophy Points:
149
Thanks.

So since 30 is 20% more than 25, why are you surprised that the range of HDCA goes from 10 to 30% more than 10?

The range for HDCA is proportionally wider than the range for total shots.

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey
That's because there's fewer high danger chances than shots. The actual quantity is what you want to look at, and not the percentage.

17. Doctor NoRegistered User

Joined:
Oct 26, 2005
Messages:
8,098
1,636
Trophy Points:
149
Well, no. Both matter. Compare two goaltenders on opposite sides of your spectrum - the first faces 25 total shots and 10 high-danger shots. The second faces 30 total shots and 13 high-danger shots.

I'm using a 98.0% save percentage for non-high danger shots, and an 85.0% save percentage for high danger shots. Feel free to replace with whatever percentages you find reasonable.

Here's the results of this exercise:

 Low Shots High Shots Total Shots 25 30 High Danger Shots 10 13 Non High Danger Shots 15 17 HD SV% 0.85 0.85 NHD SV% 0.98 0.98 HD Saves 8.50 11.05 NHD Saves 14.70 16.66 Total Saves 23.20 27.71 Total SV% 0.928 0.924 Total GA 1.80 2.29

As you can see, in this example the high shot goaltender has a lower save percentage. It matters (and my question above matters).

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey
I never said "more shots equals higher SV%". EVER.

19. Doctor NoRegistered User

Joined:
Oct 26, 2005
Messages:
8,098
1,636
Trophy Points:
149
I never said that you did. The fact that I made a significant effort to show you a flaw in your analysis, and your only response is to deny something that I didn't actually say, is a huge part of this problem (see my second to last paragraph below for more). But the statement:

"In other words, whether a team allows 30 shots a game or 25 shots a game, both teams will still allow between 10 and 13 high danger chances per game."

Doesn't even get us to it happening more often than not, and yet it's your key piece of evidence as to why this happens.

I've connected the dots sufficiently, and you've managed to draw me back into a meaningful response to you once more.

I understand that, when you never admit to having a flaw in your theory, it's not because you're trying to mislead - you actually believe that there's no flaw in your theory, and you can't be bothered to look at any of the responses hard enough to see them. I've told you this from the beginning - you have something here, and it's interesting. If you ever want to figure out why it's interesting, I'd suggest being open to constructive feedback at least once in your life.

I've said multiple times that I'm done responding, but I'm not capable of that. What I *can* promise is that I'm done giving meaningful responses, because you aren't interested in a dialogue unless people take your word as gospel.

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey
Well fine, but let's not act as if you guys aren't guilty of the exact same thing. I've tried to show the huge flaws in your analysis several times, and it just gets glossed over.

Well it's true if we're talking averages.

Right now it's just a theory, but the phenomena itself cannot be denied.

What dots have you connected? I'm not entirely sure how your example has any relevance to SV% deflation due to a large quantity of low shot volume games.

It would help if you knew what my theory was before trying to find flaws in it. That's my frustration because I don't think you guys do. BTW, if I tried to show flaws in your theories, you react the same way.

Not true at all, and I'm extremely open minded. However if I see a problem with the counterpoints that are being presented, I will mention them. I get smug and condescending responses back from you and @Hockey Outsider.

Trust me, I'm extremely thorough with my research on everything, and if I found any flaw with it I would not bring it to the forefront. I look for all the flaws well ahead of time.

See the problem here is that I have not ignored anything you have said. I have acknowledged everything you've presented. I'm extremely open minded, but if I see problems with your counterpoints I will mention them.

On top of that, I get accused of literally just making up the tables and that the information is factually incorrect. That is something I do not appreciate.

Here's another problem. You for whatever reason can't seem to realize that it's a possibility that maybe YOU'RE the one who's wrong. Either that or I disagree. If I disagree with something you said, you make statements like the one above and say that I'm stubborn. That's not trying to engage in meaningful discussion. That's just you trying to step on someone.

21. hattersonRegistered User

Joined:
Apr 12, 2010
Messages:
22,692
2,571
Trophy Points:
156
Occupation:
Software Developer
Location:
North Tonawanda, NY
Apologies for quoting pieces of multiple posts, but they all appear to be arguing essentially the same thing so I figured I'd group them together.

This seems like a bold claim. Bold claims require significant proof. I assume the supposed proof is related to your quote below:

In regards to the literal facts of what you said that 30 or 25 shots a game still means between 10 and 13 high danger chances, that's technically true, but as Doc noted, a 20% variance in total shot volume resulting in a 30% variance in high danger chances isn't proof of anything and could easily be proof of the opposite of what you're saying.

I grabbed the data from your link. I chose 2016-2017 through 2017-2018, the last 2 full NHL seasons. If that's not sufficient, feel free to grab your own range or provide game based data if you dislike aggregates.

Here's what both Corsi Against / 60 and Shots Against / 60 vs High Dangers Chances Against /60 look like:

Obviously that's not a perfect relationship, but it does strongly imply that goalies who routinely face fewer shots (or rather teams who routinely give up fewer shots) will have a much easier time giving up 2 or fewer goals in direct contradiction to the point of yours I quoted first. The relationship holds even stronger for Corsi against vs. shots against.

To note: I also plotted High Danger Save % vs High Danger Chances Against /60 As you can see, the relationship is flooded with noise which, combined with the relationship between shots/corsi and high danger chances further undermines your point I originally quoted.

Also of note to the original premise of the thread, over this time period Shots Against / 60 and Save Percentage at 5 on 5 had essentially have no correlation with an R^2 of .0002

Doctor No likes this.

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey
It's ok. The only thing I hate is when people only quote a portion of my post and delete the rest.

I'll be fair here. I was only half right when I said this. GAA is lower in the low shot volume games. However it just doesn't drop off significantly enough to where the SV% isn't decreased.

Just to clarify, it's true if you go by Natural Stat Trick's definition of high danger chance ONLY. I do not want people to think I'm making that number up.

I have a problem with using the percentage variance because the lower the quantity of anything, the higher the percentage that one of that quantity is going to be.

The actual number variance ultimately is the factor that matters in regards to this particular effect.

There's major problems with this data and info.
• First of all, it doesn't take into effect the actual quality of the goalie. All goalies are not equal.
• Second, I'm not even trying to show a correlation. All I'm trying to show is that there is an inflating and deflating effect dependent on the ratio of low shot volume games to high shot volume games a goalie plays in.
• For example, Tim Thomas had an INFLATED .933 SV% in 10/11 while Marty Turco had a DEFLATED .932 SV% in 2002-2003.
• Basically, what I'm trying to say here is that if 2003 Marty Turco allows 2 goals in a game, his SV% goes down. If 2011 Thomas allows 2 goals in a game, his SV% goes up.
• A goalie who faces 25 shots a game can definitely have a higher SV% than a goalie who faces 30 shots a game. That goes back to my above statement that not all goalies are equal.
• Bottom line each goalie can only be compared to himself for accurate conclusions.
• Third, using per 60 stats also doesn't work because each individual game is very different.
Goalies who play in a lot of low shot volume games will have a more difficult time maintaining an elite level SV% than goalies who face average to above average shots per game.
• 1,800 shots over a 70 game season is very different than 1,800 shots over a 60 game season.
• 1 GA on 25 shots will result in a lower SV% than 1 GA on 30 shots.
Bottom line is you cannot ignore the tables I've presented. I also want to point out that eliminating the games where goalies get pulled early in the game was done so it doesn't skew the data. It's important to only show the data for only the complete games the goalies play in.

In conclusion, SV% in a vacuum is not a viable way to compare multiple goalies unless......
1. Both goalies face roughly the same number of shots per game......or
2. You adjust for the difference in shots per game if there is a difference.

23. hattersonRegistered User

Joined:
Apr 12, 2010
Messages:
22,692
2,571
Trophy Points:
156
Occupation:
Software Developer
Location:
North Tonawanda, NY
In the two seasons I mentioned there was no discernible relationship between higher shot volume allowed by a team and higher save percentage by a teams goalies at 5 on 5, so it does seem that it drops off enough to not affect save percentage. Unless you have another source of information you'd like to provide?

Given you're talking about averages over dozens of games, I don't see why percentage variation wouldn't be useful. In a single game, yes saying "they scored 300% more than last time" is fairly meaningless going from 1 goal to 3 goals, but over a sample of dozens of games saying "they are scoring 20% more goals than at the start of the year" certainly has meaning.

If the claim is "individual goalies post better save percentages when they face higher shot volume" then it stands to reason that "teams who allow more shots should be expected to have a higher team save percentage" unless you're also going to claim that somehow quality of a goalie employed by a team decreases as team shot volume allowed increases.

It shouldn't matter how good a team's goalies are, if goalies in general post better numbers when facing higher numbers of shots, teams that allow higher numbers of shots should expect to see higher save percentages. Goalie quality may muddy those waters so you may expect a lower R^2 value when plotting the relationship, but it shouldn't be expected to perfectly muddy the waters and result in an R^2 value of 0 over multiple seasons.

Except this isn't true on the team level, at least in any data I've seen. Teams that allow low shot volumes do not have goalies who have lower save percentages on average.

As I mentioned earlier in this thread, I think that attempting to "not skew" the data like this is actually causing the skewing and appears to be where most of this relationship is coming from.

If a goalie is having a bad night but his team is only allowing 7 shots a period he has a solid chance of finishing the game while allowing 4 or 5 goals (.750 - .800 save percentage)

If a goalie is having a bad night but his team is allowing 19 shots in the first period (like the Leafs did yesterday) that same goalie might give up 5 goals in the first period and is almost guaranteed to get pulled.

The simple fact is that when facing high shot volumes a goalie is much more likely to get pulled unless they're having a great day. A goalie should be expected to post a higher save percentage if he plays the entire game when his team allows 50 shots simply because if he doesn't post a high save percentage he's likely allowed 5+ goals and is significantly likely to be pulled.

Saying goalies post a high save percentage when they play a full game in which they face a high number of shots seems to me to be fairly close to saying goalies post a high save percentage when they post a high save percentage simply because if they don't post a high save percentage in a game where their team allows a large number of shots, they likely don't play the full game.

Joined:
Nov 3, 2014
Messages:
24,624
1,317
Trophy Points:
124
Gender:
Male
Location:
New Jersey
Exactly why you can't ignore goalie quality variance. You can't compare goalie to goalie. It has to be a goalie to himself comparison.

I didn't say that. I'm saying that in regards to this phenomena of goalies having lower SV% in their lower shot volume games, it's the quantity variance that plays the biggest role.

Aggregate totals are practically meaningless when we're doing comparative analysis to two goalies.

This is NOT my claim.

I'm not saying that higher shot volume equals higher SV%. @Doctor No @Hockey Outsider have both fallen victim to believing that this was my claim. It's not my claim at all.

Not all goalies are equal. Funny thing is I'm actually on record saying the gap between the 31 goalies isn't very big, and I do believe that BTW.

Anyways back on topic, despite that fact the gap between all 31 starting goalies is small, there is still truth in the statement that not all goalies are equal.

Problem is, I never said this. Goalie quality will always matter no matter what. So comparing goalie to goalie does not work. Goalie to himself is the proper way to compare.

You're not looking at the correct information. Correlative values don't come into play here. When comparing goalie to himself, you'll see that all goalies have better SV%'s when they face a higher number of shots. GAA also increases.

Goalie quality can come into play simply from the fact that one goalie could have a .930/.915 split while another goalie would have a .910/.900 split.

I also made it very clear that I never said higher shot total = higher SV% or lower shot totals = lower SV%.

If I included games where goalies played fewer than 50 minutes, it'd only skew the data in my favor. That's why I excluded it.

Correct. But the problem is, we can't just look at the aggregate totals in order to evaluate a goalie's performance. We have to look at individual games.

Yes. And in that case it would skew the data in MY favor. It is obviously bad if a goalie is getting pulled a lot, but I'm not really talking about the season SV% like what you look at on the back of a hockey card.

What do YOU consider a high shot total?

EXACTLY! The problem is we forget that seasons are made up of individual games. There is no carryover effect game to game. Everything refreshes and starts over at each game.

Now imagine if a goalie plays a lot of low shot volume games like Quick or Brodeur. You can't expect their SV%'s to be as high. Even Hasek had a sub.900 SV% when he faced fewer than 20 shots. He has a career SV% of .915 when facing fewer than 30 shots. He just didn't play in as high a percentage of low shot volume games to where it affected his SV% negatively. Brodeur faced fewer than 30 shots more than 75% of his career, and Quick faced fewer than 30 shots in about 70% of his games. Hasek on the other hand faced fewer than 30 shots in only 55% of his career.

It doesn't happen often at all where a goalie plays fewer than 50 minutes and faces a high number of shots. Barrasso has the most such games at 8. Lundqvist played in 7 such games. I get what you're trying to say, but it just doesn't happen enough to really impact any of the data.

BTW, my main point in this whole thread was more about the low end of the spectrum. I'm trying to show the negative impact for any goalie's SV% who plays on strong defensive teams. Those goalies have a smaller margin of error, and the tiny decrease in high danger chances isn't enough to make up for the lack of overall shot volume to inflate their SV%.

25. hattersonRegistered User

Joined:
Apr 12, 2010
Messages:
22,692
2,571
Trophy Points:
156
Occupation:
Software Developer
Location:
North Tonawanda, NY
I'll just pick this out for now as the response is fairly straight forward. I may respond to the rest later.

I'm not talking about games where a goalie *did* face a large number of shots and got pulled, I'm talking about games where a goalie *would have* faced a large number of shots but was having a poor/mediocre outing and thus gets pulled early. By excluding those from the counting, you're artificially removing games that would have been high shot total games.

I'd be interested to see individual goalie data where games are not broken down by shot totals, but rather by shot rates per 60 of even strength play and without excluding short games.

It may also be interesting to look at it per period of play and see if a low shot volume period has a higher chance to result in a low save percentage period.

So far, the aggregate data doesn't show a relationship at the team level and the game to game comparisons are flawed (IMO) for the reasons I've laid out.