Fenwick save percentage

[Steerpike]

Before I rode the coattails of the 2013-2014 Avalanche into being a fairweather hockey fan my favorite sport to watch was men's tennis. In men's tennis these days many of the top stars are both incredible athletes and have an insane ability to read where their opponent is going to hit the ball. Players like Nadal, Murray, Djokovic, and Ferrer are prime examples of this. What this very typically leads to is opponents unnecessarily hit the ball out way more than they normally would. These events are recorded in a stat called "unforced errors" but to a large extent these errors aren't "unforced" but rather a result of trying to make their shots too good.

You've probably realized what I'm getting at by now. It seems reasonable to extend this to goaltending and see if better goaltenders force opponents to go for "riskier" shots that might miss the net. Wouldn't it make sense to primarily look at a goalie's Fenwick save percentage due to this effect? Also, do the very best goalies tend to influence opposing teams into having a worse on-net percentage?

Click to read more...

 

[Steerpike]

okay I realized I could easily investigate this on my own so I did.

Here's a link to an excel file I made on it.
https://www.dropbox.com/s/o2fhfdi4ifowfl9/FSVpercentage.xlsx

 

[Steerpike]

Okay so I also should have circled Hiller's numbers.


Perhaps what this shows is that Quick's save percentage is lower than is real impact due to his reputation. Bernier also had a better season than his save percentage rank.

I have trouble making sense of Hiller's numbers. How is it that so few unblocked shots missed the net when the shots against distance is so low and he wasn't that amazing in net?

 

[Steerpike]

Also, FA/G or SA/G (great acronyms)(I should have written out FA/Goal and SA/Goal as FA/game is already a thing.) should clearly replace Sv%. It's just so much easier to see the impact they have on the game this way.

Here's sorting goalies by FA/G

 

[Steerpike]

Upon further review sorting by OnN% (100xSA/fenwick or OnNet%) produces a seemingly random order.

This pretty much undermines my whole premise so maybe this is pointless.

 

[Steerpike]

Actually, I think something about Hiller/the Ducks is making people miss the net! Not entirely sure what it is, but it's relatively consistent.

Note I decided to change OnNet% to SA/FA as that's more clear.

 

[Steerpike]

My imgur is down... but Rask/Bruins are consistently doing something that makes a lot of fired shots hit the net.

0.737649063
0.732311321
0.721649485
0.740857788

are his SA/FA numbers


Maybe part of this has to do with the people recording these stats. Perhaps the people recording the FA at Bruins games aren't very trigger happy while the dude for Ducks games is less trigger happy.



Quick's numbers:
0.695522388
0.681419166
0.706677266
0.681451613
consistently low.

Luongo
0.742167655
0.762203963
0.732712766
0.725696029
Relatively Consistently Medium

Thomas
0.744042728
0.755120619
0.762897914
Consistently very high

 

[Steerpike]

So It looks like goalies do significantly impact how many fenwick shots are on net. Exactly how this is happening, I have no idea and would appreciate help. I remember back in the day some people were saying that Patrick Roy liked to show his 5-hole open to temp players and then snap it shut when they shot. Perhaps something like that is happening where certain goalies tempt more shots on net while others encourage wide/high shots.

Regardless, every time a player shoots the puck it takes up a certain amount of possession and risks handing the puck over. Making the opposition miss should be seen as an asset.

Time to go to class.


Oh and I haven't looked at Corsi, but I don't think goalies really have anything to do with a shot getting blocked.

 

[NyQuil]

I appreciate your work here, keep it up!

I'll have to re-read through it again as I'm doing this on a bus, but it's always neat to see someone tackle something like this.

 

[Hivemind]

Interesting premise, but I'm not sure I buy it yet. I'd want to see more multi-year numbers for goalkeepers (is this sustainable?), and how they relate to their teams' overall numbers (do Bruins/Kings/Ducks back-ups also scare away shooters?). I'd also like to see if there's any correlation between relatively poor sv% and relatively strong sv% years for the same goaltender (are more shots scared away during a goalie's good year compared to his bad one?). PK scenarios may also be interesting, though harder to control for.

MyMaybe part of this has to do with the people recording these stats. Perhaps the people recording the FA at Bruins games aren't very trigger happy while the dude for Ducks games is trigger happy.

Scorekeeper bias is a pretty recognized issue around the hockey analytics community. Certain arenas/scorekeepers have different tendencies when it comes to recording stats. A way to mildly control for this is to use road stats only. It still provides a 41 game sample size, with hopefully a more broad distribution of scorekeeper tendencies.

 

[ARSix]

This is a good concept. I'm wondering how the opposite effect plays into it - i.e., certain goalies, based on their style of play, are often said to cause opposing teams to over-pass (i.e. "looking for the perfect play") and try to set up easier goals and open net tap ins, which would depress shot attempts against.

how they relate to their teams' overall numbers (do Bruins/Kings/Ducks back-ups also scare away shooters?).

I worry a bit about the sample size but this seems like it would be useful information over a multi-year period.

 

[Raccoon Jesus]

Interesting premise, but I'm not sure I buy it yet. I'd want to see more multi-year numbers for goalkeepers (is this sustainable?), and how they relate to their teams' overall numbers (do Bruins/Kings/Ducks back-ups also scare away shooters?). I'd also like to see if there's any correlation between relatively poor sv% and relatively strong sv% years for the same goaltender (are more shots scared away during a goalie's good year compared to his bad one?). PK scenarios may also be interesting, though harder to control for.

Scorekeeper bias is a pretty recognized issue around the hockey analytics community. Certain arenas/scorekeepers have different tendencies when it comes to recording stats. A way to mildly control for this is to use road stats only. It still provides a 41 game sample size, with hopefully a more broad distribution of scorekeeper tendencies.

Agreed on the interesting premise and love the work, great thoughts, thanks! I would argue that it's not the only conclusion but there is at least some correlation there.

 

[Raccoon Jesus]

This is a good concept. I'm wondering how the opposite effect plays into it - i.e., certain goalies, based on their style of play, are often said to cause opposing teams to over-pass (i.e. "looking for the perfect play") and try to set up easier goals and open net tap ins, which would depress shot attempts against.

I worry a bit about the sample size but this seems like it would be useful information over a multi-year period.

Agreed, and as a homer, using Quick as an example, I think it's pretty clear that it's part system and part opponents not wanting to just dump random shot volume on Quick and get rebounds, there's a lot of attempts to take advantage of his aggressiveness--so maybe it's not just rep but style as well? Butterfly goalies like Bernier would likely face different challenges than guys like Thomas/Quick. I know this is anecdotal on my part but as someone who watches them frequently opponents do quite a bit of double-clutching and overpassing (which works) on Quick.

 

[jc17]

really like this idea and will try to look into it too. I'm thinking it would be smart to compare goalies of different styles and sizes to see if limited shooting area for open net also plays into it.

 

[Bear of Bad News]

This is really interesting stuff; thanks for sharing it.

I had a similar thought to the above, that taller (larger) goaltenders would show less net and force more shots to go wide.

Of course, it's not the size but how you use it (as they say ) - so a smaller goaltender who challenges well and uses his angles could have similar numbers to a larger goaltender who hangs back.

 

[Micklebot]

So It looks like goalies do significantly impact how many fenwick shots are on net. Exactly how this is happening, I have no idea and would appreciate help. I remember back in the day some people were saying that Patrick Roy liked to show his 5-hole open to temp players and then snap it shut when they shot. Perhaps something like that is happening where certain goalies tempt more shots on net while others encourage wide/high shots.

Regardless, every time a player shoots the puck it takes up a certain amount of possession and risks handing the puck over. Making the opposition miss should be seen as an asset.

Time to go to class.


Oh and I haven't looked at Corsi, but I don't think goalies really have anything to do with a shot getting blocked.

Have you compared individual goalies results to that of thier backups to try and figure out how much is system generated, and how much goalie generated? I think to get meaningful samples, you'd have to expand to multiple years though.

 

[Steerpike]

Have you compared individual goalies results to that of thier backups to try and figure out how much is system generated, and how much goalie generated? I think to get meaningful samples, you'd have to expand to multiple years though.

Hmmmmm

I would but that would be really tedious. Maybe if the data was in the right format I could write a script to do it, but as of now? Eh.

 

[Hivemind]

Hmmmmm

I would but that would be really tedious. Maybe if the data was in the right format I could write a script to do it, but as of now? Eh.

Where are you exporting your raw stats from? Does it have a "team" column you could use? Then just sort the spreadsheet by team, and compare the starters # to the back-ups #.

 

[Steerpike]

Where are you exporting your raw stats from? Does it have a "team" column you could use? Then just sort the spreadsheet by team, and compare the starters # to the back-ups #.

lol I was Ctrl-c ing from exrtra skater.

Each goalie had to be done separately if I wanted to look at multiple years. I guess that yes comparing starters to backups would root out effects based on teams and looking only at away games would eliminate scorekeeper bias.

 

[Micklebot]

lol I was Ctrl-c ing from exrtra skater.

Each goalie had to be done separately if I wanted to look at multiple years. I guess that yes comparing starters to backups would root out effects based on teams and looking only at away games would eliminate scorekeeper bias.

Try stats.hockeyanalysis.com

http://stats.hockeyanalysis.com/rat...es&minutes=100&disp=1&sort=HARTp&sortdir=DESC

They have up to 7 years worth of data combined,

 

[Hawsse]

this was a dumb post from a child with no idea what he was talking about. ignore it.

 

[Freudian]

Since you can't get a save on a missed shot it seems a bit pointless using Fenwick over shots unless you assume goalies force shooters to miss. Most of the time they don't.

Also, Bernier and Schneider have a higher number here compared to save percentage, something that should be impossible.

 

[Doctor No]

Also, Bernier and Schneider have a higher number here compared to save percentage, something that should be impossible.

This was the first thing that stood out to me.

To your first point, while I agree that "most of the time" goalies don't force shooters to miss, I also believe that there is a difference between goaltenders' abilities in this regard. In an industry where the difference between "average" and "good" is one goal per 100 shots, it's an interesting thing to consider.

This may not be the optimal way to measure it, but it's something to think about.

 

[The Bob Cole]

By adding in shots that missed you're measuring less of the goalies abilities and more of an overall teams possession capabilities (the fenwick numbers). The only thing you're still comparing directly between goalies is the save%.

 

[hatterson]

Where did you get these numbers from? They seem very off.

First, by definition, it should be impossible to have a fenwick save percentage (or SAGS%) below your standard save percentage but yet there are several goalies (All but the top 3-4 in your list actually) that do.

Second, they seem very low even for he top guy on this list (Bernier), nevermind the lower guys. If we're doing all phases (ES, PK, PP) at once then Bernier recorded 1649 saves on 1787 shots for a save percentage of .9228. Given that he (by definition) can't have any more goals allowed when looking at fenwick save percentage, A fenwick save percentage of .9382 implies that he only faced an extra 446 fenwick events (missed shots) compared to shots on goal, which means that missed shots accounted for 20% (446/2233) of the fenwick events directed at Bernier, that seems extremely low given that for the Leafs as a whole missed shots accounted for roughly 30% of fenwick events.

In fact, ~30% is one of the highest marks in the league (I don't have the exact data on hand, but a glance at ES only data shows the Leafs 2nd top in this category), and 20% would be several percentage points below the lowest in the league. It seems extremely unlikely for there to be that kind of disparity between goalies on the same team given that it would mean Reimer (or Macintyre for those couple games) had 40+% of the shots directed against them miss the net.