Special teams adjusted goaltending statistics

[Doctor No]

Thanks to those who helped me figure out my problems in the other thread. I'm starting this thread to place things that I discover about special teams-adjusted goaltending statistics.

A lot of people tout even-strength save percentage as the gold standard for measuring netminders in a vacuum. I agree with this to a point; the sample size is there and the data are more homogeneous. However, that penalizes goaltenders who do well in the other situations.

I instead prefer to measure how well a goaltender did, given the number of shots that they faced in each situation (ES, PP, SH), compared to the league average goaltender in those situations (those of you who have been annoyed by me for awhile could probably have guessed this).

For instance, Tuukka Rask of the Bruins allowed 115 goals in the 2013-14 season. Leaguewide, the average save percentages were 0.921 (vs. even strength), 0.878 (vs. power play), and 0.895 (vs. shorthanded). Rask faced 1359 even-strength shots, 251 power-play shots, and 31 short-handed shots, so we would expect him to allow: 1359*(1-0.921) + 251*(1-0.878) + 31*(1-0.895) = 140.7 goals.

In other words, controlling for manpower situations, Rask prevented 25.7 goals beyond what an average goaltender would have prevented.

Without controlling for manpower situations (just straight leaguewide save percentage), Rask prevented 27.0 goals beyond what an average goaltender would have prevented.

Therefore, the Bruins' manpower situations made Rask look about 1.3 goals better than he actually was. Is 1.3 goals a lot? How does Rask compare to other goaltenders in this regard?

Click to read more...

 

[Doctor No]

Here is every goaltender/team combination for the 2013-14 NHL season.

GD = goals prevented beyond league average
STAGD = special-teams adjusted goals prevented beyond league average
DIFF = the difference between the two (positive means goaltender benefits from special teams)

GOALTENDER|TEAM|GD|STAGD|DIFF
Anton Khudobin |Carolina Hurricanes|12.9|10.4|2.5
Antti Niemi |San Jose Sharks|-2.1|-4.3|2.2
Roberto Luongo |Vancouver Canucks|3.8|1.8|2.0
Henrik Lundqvist |New York Rangers|12.3|10.4|1.9
Alex Stalock |San Jose Sharks|10.3|8.6|1.7
Semyon Varlamov |Colorado Avalanche|28.2|26.6|1.6
Tuukka Rask |Boston Bruins|27.0|25.7|1.3
Alvaro (Al) Montoya |Winnipeg Jets|4.6|3.3|1.3
Carter Hutton |Nashville Predators|-4.4|-5.6|1.2
Darcy Kuemper |Minnesota Wild|0.5|-0.6|1.1
Kevin Poulin |New York Islanders|-18.5|-19.5|1.0
Evgeni Nabokov |New York Islanders|-9.7|-10.7|1.0
Karri Ramo |Calgary Flames|-3.0|-3.9|0.9
Corey Crawford |Chicago Blackhawks|4.2|3.3|0.9
Thomas Greiss |Phoenix Coyotes|4.0|3.1|0.9
Eddie Lack |Vancouver Canucks|-2.4|-3.2|0.8
Steve Mason |Philadelphia Flyers|6.0|5.2|0.8
Reto Berra |Calgary Flames|-13.2|-14.0|0.8
Jonathan Bernier |Toronto Maple Leafs|16.4|15.6|0.8
Cam Ward |Carolina Hurricanes|-13.1|-13.8|0.7
Pekka Rinne |Nashville Predators|-7.5|-8.1|0.6
Nathan Lieuwen |Buffalo Sabres|-1.5|-2.1|0.6
Cameron Talbot |New York Rangers|15.4|14.9|0.5
Josh Harding |Minnesota Wild|13.6|13.1|0.5
Fredrik Andersen |Anaheim Ducks|7.5|7.0|0.5
Marek Mazanec |Nashville Predators|-8.1|-8.5|0.4
Brian Elliott |St. Louis Blues|5.7|5.3|0.4
James Reimer |Toronto Maple Leafs|-3.7|-4.1|0.4
Joey MacDonald |Calgary Flames|-6.3|-6.7|0.4
Joni Ortio |Calgary Flames|-4.4|-4.7|0.3
Justin Peters |Carolina Hurricanes|3.0|2.7|0.3
Petr Mrazek |Detroit Red Wings|2.3|2.0|0.3
Devan Dubnyk |Nashville Predators|-3.8|-4.0|0.2
Richard Bachman |Edmonton Oilers|0.2|0.0|0.2
Reto Berra |Colorado Avalanche|-4.2|-4.4|0.2
Carey Price |Montreal Canadiens|24.1|23.9|0.2
Drew MacIntyre |Toronto Maple Leafs|0.4|0.3|0.1
Martin Biron |New York Rangers|-5.7|-5.8|0.1
Niklas Svedberg |Boston Bruins|1.0|0.9|0.1
Jhonas Enroth |Buffalo Sabres|-2.2|-2.3|0.1
Jacob Markstrom |Florida Panthers|-11.5|-11.6|0.1
Peter Budaj |Montreal Canadiens|-3.0|-3.1|0.1
Cedrick Desjardins |Tampa Bay Lightning|-0.9|-1.0|0.1
Tim Thomas |Dallas Stars|-2.2|-2.3|0.1
Viktor Fasth |Anaheim Ducks|-3.7|-3.8|0.1
Kent Simpson |Chicago Blackhawks|-1.4|-1.4|0.0
Ben Scrivens |Edmonton Oilers|1.9|1.9|0.0
Ilya Bryzgalov |Edmonton Oilers|-3.4|-3.4|0.0
Nathan Lawson |Ottawa Senators|-1.1|-1.1|0.0
Mark Visentin |Phoenix Coyotes|-0.2|-0.2|0.0
Nikolai Khabibulin |Chicago Blackhawks|-7.6|-7.6|0.0
Ryan Miller |St. Louis Blues|-5.2|-5.2|0.0
Michael Hutchinson |Winnipeg Jets|2.6|2.6|0.0
Pat Conacher Jr. |Vancouver Canucks|0.0|0.0|0.0
Andrey Makarov |Buffalo Sabres|0.0|0.0|0.0
Kurtis Mucha |Edmonton Oilers|0.0|0.0|0.0
Ryan Vinz |Buffalo Sabres|0.0|0.0|0.0
Sami Aittokallio |Colorado Avalanche|-1.4|-1.4|0.0
Jacob Markstrom |Vancouver Canucks|-3.5|-3.5|0.0
Cal Heeter |Philadelphia Flyers|-1.7|-1.7|0.0
Andrew Hammond |Ottawa Senators|0.9|1.0|-0.1
Christopher Nilstorp |Dallas Stars|-1.5|-1.4|-0.1
Magnus Hellberg |Nashville Predators|-0.7|-0.6|-0.1
Philipp Grubauer |Washington Capitals|5.4|5.5|-0.1
Connor Knapp |Buffalo Sabres|-1.2|-1.1|-0.1
Kari Lehtonen |Dallas Stars|9.9|10.0|-0.1
Dan Ellis |Florida Panthers|-12.8|-12.7|-0.1
John Gibson |Anaheim Ducks|3.5|3.6|-0.1
Kristers Gudlevskis |Tampa Bay Lightning|1.3|1.5|-0.2
Ilya Bryzgalov |Minnesota Wild|-0.8|-0.6|-0.2
Ryan Miller |Buffalo Sabres|13.8|14.0|-0.2
Chad Johnson |Boston Bruins|8.1|8.3|-0.2
Dustin Tokarski |Montreal Canadiens|3.0|3.2|-0.2
Dan Ellis |Dallas Stars|-5.0|-4.8|-0.2
Scott Clemmensen |Florida Panthers|-7.9|-7.7|-0.2
John Curry |Minnesota Wild|0.9|1.2|-0.3
Craig Anderson |Ottawa Senators|-5.4|-5.1|-0.3
Jason LaBarbera |Edmonton Oilers|-6.4|-6.1|-0.3
Jonas Hiller |Anaheim Ducks|-3.9|-3.6|-0.3
Tim Thomas |Florida Panthers|-6.2|-5.9|-0.3
Marc-Andre Fleury |Pittsburgh Penguins|2.9|3.2|-0.3
Mike McKenna |Columbus Blue Jackets|-1.2|-0.9|-0.3
Jean-Sebastien Giguere |Colorado Avalanche|-0.6|-0.2|-0.4
Jaroslav Halak |Washington Capitals|6.0|6.4|-0.4
Jack Campbell |Dallas Stars|-2.0|-1.6|-0.4
Cory Schneider |New Jersey Devils|7.5|8.0|-0.5
Anders Lindback |Tampa Bay Lightning|-13.1|-12.6|-0.5
Viktor Fasth |Edmonton Oilers|0.0|0.5|-0.5
Joacim Eriksson |Vancouver Canucks|-3.3|-2.8|-0.5
Jimmy Howard |Detroit Red Wings|-5.4|-4.8|-0.6
Martin Brodeur |New Jersey Devils|-12.5|-11.9|-0.6
Matt Hackett |Buffalo Sabres|-1.4|-0.7|-0.7
Ben Scrivens |Los Angeles Kings|8.0|8.7|-0.7
Michal Neuvirth |Buffalo Sabres|3.4|4.1|-0.7
Sergei Bobrovsky |Columbus Blue Jackets|15.8|16.5|-0.7
Ray Emery |Philadelphia Flyers|-7.8|-7.1|-0.7
Curtis McElhinney |Columbus Blue Jackets|-3.5|-2.8|-0.7
Jonas Gustavsson |Detroit Red Wings|-4.7|-4.0|-0.7
Ben Bishop |Tampa Bay Lightning|18.9|19.6|-0.7
Martin Jones |Los Angeles Kings|10.2|10.9|-0.7
Ondrej Pavelec |Winnipeg Jets|-21.8|-21.0|-0.8
Devan Dubnyk |Edmonton Oilers|-18.0|-17.2|-0.8
Robin Lehner |Ottawa Senators|-0.7|0.1|-0.8
Roberto Luongo |Florida Panthers|4.2|5.1|-0.9
Braden Holtby |Washington Capitals|1.1|2.1|-1.0
Anders Nilsson |New York Islanders|-9.9|-8.9|-1.0
Jaroslav Halak |St. Louis Blues|3.4|4.5|-1.1
Niklas Backstrom |Minnesota Wild|-8.0|-6.9|-1.1
Antti Raanta |Chicago Blackhawks|-10.5|-9.4|-1.1
Jonathan Quick |Los Angeles Kings|2.0|3.4|-1.4
Jeff Zatkoff |Pittsburgh Penguins|-0.9|0.5|-1.4
Mike Smith |Phoenix Coyotes|2.3|3.9|-1.6
Michal Neuvirth |Washington Capitals|0.1|2.0|-1.9

 

[Doctor No]

So the spread, at least for the 2013-14 NHL season, is about 4.5 goals - the most-advantaged goaltender (Khudobin) gains about 2.5 goals, and the most-disadvantaged (Neuvirth in Washington) loses about 2 goals.

What does that mean in terms of obfuscation? Khudobin allowed 80 goals on 1076 shots, a save percentage of 92.6%. If he were on the other end of the coin, allowing 84.5 goals on 1076 shots, his save percentage would be 92.1%.

So it's not a huge difference. I'll be running the numbers back (ultimately) as far back as the NHL tracks this separately.

For further study - I'd like to actually measure the predictive power of even-strength save percentage, relative to overall save percentage.

 

[Caeldan]

So the spread, at least for the 2013-14 NHL season, is about 4.5 goals - the most-advantaged goaltender (Khudobin) gains about 2.5 goals, and the most-disadvantaged (Neuvirth in Washington) loses about 2 goals.

What does that mean in terms of obfuscation? Khudobin allowed 80 goals on 1076 shots, a save percentage of 92.6%. If he were on the other end of the coin, allowing 84.5 goals on 1076 shots, his save percentage would be 92.1%.

So it's not a huge difference. I'll be running the numbers back (ultimately) as far back as the NHL tracks this separately.

For further study - I'd like to actually measure the predictive power of even-strength save percentage, relative to overall save percentage.

Is it that the goalie gains an advantage from special teams? Or that the special teams gained an advantage from the goalie?

Example with Khubodin at 2.5... the other two Carolina goalies are right on 0 nearly. That's a 2 goal difference within the same team? Should they not track closer if it were special teams impacting their performance?

 

[Doctor No]

Is it that the goalie gains an advantage from special teams? Or that the special teams gained an advantage from the goalie?

Example with Khubodin at 2.5... the other two Carolina goalies are right on 0 nearly. That's a 2 goal difference within the same team? Should they not track closer if it were special teams impacting their performance?

I probably should have explained it better, but it's the former.

For instance, if we ignore the fact that some teams put their goalies in more special team situations, Khudobin was 12.9 goals better than an average goaltender over 2013-14.

However, if the average goaltender were put into the exact same situations as Khudobin (facing the same percentage of ES shots, PP shots, and SH shots), then the "average" goaltender goes better, and Khudobin was only 10.4 goals better than this average goaltender.

Therefore, Khudobin's season looked 2.5 goals better than it actually was because of special team situations.

 

[hatterson]

Could you break out the STAGD for each phase? See how many goals above/below average a goalie was at ES, PP and PK.

That seems like it would give you a good insight into if special teams differences are more controllable.

IE, if a goalies who posts high differences above average at special teams in one season are likely to do so again in another season, it would indicate that goalies really can control their performances at special teams (or at least that special teams goaltending performances can be controlled, whether that's the goalie or the team contributing more would require a bit more research). However, if the distribution generally appears random, with goalies frequently going from above average to below average, it would seem to indicate that special teams performance is more susceptible to luck, or at least to the swings within small sample sizes.

 

[Doctor No]

I like that idea a lot - I have the raw numbers that built the totals, of course.

 

[Mantha Poodoo]

So the spread, at least for the 2013-14 NHL season, is about 4.5 goals - the most-advantaged goaltender (Khudobin) gains about 2.5 goals, and the most-disadvantaged (Neuvirth in Washington) loses about 2 goals.

What does that mean in terms of obfuscation? Khudobin allowed 80 goals on 1076 shots, a save percentage of 92.6%. If he were on the other end of the coin, allowing 84.5 goals on 1076 shots, his save percentage would be 92.1%.

So it's not a huge difference. I'll be running the numbers back (ultimately) as far back as the NHL tracks this separately.

For further study - I'd like to actually measure the predictive power of even-strength save percentage, relative to overall save percentage.

This is pretty awesome, though I have to say I disagree with the bolded point. A .5% difference in save percentage in today's NHL can be the difference between being the Vezina winner and not even being a Vezina finalist. Interesting to see how much of a difference a team's respective special teams play can have on the perception of a goalie. On the other hand, this is only correlation at the moment. In which cases are these because a goaltender does better/worse on special teams as opposed to his team's special teams play being better/worse?

 

[Caeldan]

I probably should have explained it better, but it's the former.

For instance, if we ignore the fact that some teams put their goalies in more special team situations, Khudobin was 12.9 goals better than an average goaltender over 2013-14.

However, if the average goaltender were put into the exact same situations as Khudobin (facing the same percentage of ES shots, PP shots, and SH shots), then the "average" goaltender goes better, and Khudobin was only 10.4 goals better than this average goaltender.

Therefore, Khudobin's season looked 2.5 goals better than it actually was because of special team situations.

Maybe I'm still misunderstanding, or misapplying the results but what I'm saying is that if you look at Khudobins numbers vs that of his teammates - there's a good difference there.
The variable in that case is the goalie, as opposed to the special teams?
If one goalie has a benefit from special teams shouldn't all goalies on the team gain a similar benefit?

Some teams everyone has very similar numbers, others you see one goalie who stands out.

 

[hatterson]

Maybe I'm still misunderstanding, or misapplying the results but what I'm saying is that if you look at Khudobins numbers vs that of his teammates - there's a good difference there.
The variable in that case is the goalie, as opposed to the special teams?
If one goalie has a benefit from special teams shouldn't all goalies on the team gain a similar benefit?

Some teams everyone has very similar numbers, others you see one goalie who stands out.

I think it's definitely something worth investigating, however you have to consider that each goalie doesn't face the same level of PPs and PKs, especially when you're looking at samples of 5-10 games.

Keep in mind we're putting an average goalie into this goalies situation, we're not putting this goalies performance into an average situation. Mind you, it may actually be interesting to see that. Take a goalies performance in terms of save percentage at the various phases of the game and then plug those into an "average" workload.

 

[Doctor No]

That's true - there's two ways of looking at the situation, which should produce comparable results:

• Compare Goalie X's performance to what an average goaltender would do in the same situation.
• Compare an average goaltender's performance to what Goalie X's performance would be under an average situation.

The first is what I've done above, and the reason is that the second is a bit more intractable mathematically - I'll show why.

Suppose that Goalie X faced 1000 even-strength shots (with 100 ES goals), faced 50 power-play shots (with 10 PP goals), and didn't face a single short-handed shot (this is more common than you'd think for backups or reserves).

We could take his performance and "stretch" it to fit an average goaltender's situation, except that he stopped 90% of even-strength shots, 80% of power-play shots, and an undefined percentage of short-handed shots.

That last difficulty is why I chose the first approach (and that's essentially the only reason as well, since the second approach lines up more closely to how I think about things in general).

 

[TychoFan]

Interesting, even the first two columns I find. Varlamov/Rask/Price are leading while Poulin/Dubnyk/Pavelec are at the bottom. Those stats should become more common, they seem to be better than save percentage for example.

 

[hatterson]

That's true - there's two ways of looking at the situation, which should produce comparable results:

• Compare Goalie X's performance to what an average goaltender would do in the same situation.
• Compare an average goaltender's performance to what Goalie X's performance would be under an average situation.

To me, the first option more aligns with how I think of a performance in a single season. How did this goalie's perform compared to an average goalie (or you could redefine it to be replacement if you so chose)? Whereas the second lines up more with what I'd think of as a goalie's value moving forward. If I stuff him in an 'average' situation, what can he do for me?

Both are going to be very similar, it just seems to be a case of phrasing.