Wilson: The numbers behind team play vs. goalie stats

Goaltending in the NHL is one of the most important positions in all of sports.  Trying to get a better understanding of the position is something that the analytics community has been working on for years.

It is becoming more and more accepted around the mainstream hockey community that possession does indeed correlate to winning hockey games.

The question is how much does possession, specifically the shot volume against, impact a goaltender’s ability to perform for his team?

My interest in this was spawned by following the situations in both Buffalo and Minnesota this past season.  Living in Western New York gave me a front row seat of Buffalo’s efforts to secure 30th place and a guarantee of landing McEichel.  Last year’s Buffalo squad was the worst possession team in the analytics era (2007-present), yet finished as the 15th ranked team in even-strength save percentage at .924. Their goaltending was so good in some stretches that it forced General Manager Tim Murray to trade both Michael Neuvirth and Jhonas Enroth in order to keep their goal of 30th place alive. It seemed strange to me that the worst team in the analytics era could post slightly above average goaltending numbers with two goalies that aren’t considered high end stars.

On the flip side you had a team like Minnesota who was one of the better shot suppression teams in the first half of the season sinking down in the standings because of the below average goaltending they were receiving. The mid-season trade to bring in Devan Dubnyk saved their season even though their shot suppressing abilities started to decline post trade.

Using the rolling ten game averages for both Fenwick Against per 60 and even-strength save percentage we can see that there was definitely not a positive correlation between the two variables for Minnesota.

Ryan W 1


Ryan W 2

After seeing the lack of correlation between shot suppression and save percentage for both a playoff team and a bottom dweller I decided to open up this sample size to every single team from 2007-15. I wanted to see if Minnesota and Buffalo were outliers that I just happened to observe on my own or if there was something to this.

First I ran the numbers at basic even-strength play

Ryan W 3


At even-strength there appears to be no correlation between a goaltender’s ability to do his job and his team’s ability to suppress shots.

I then ran the Score-Adjusted even-strength numbers.

Ryan W 4

The results are almost identical.  In fact, the correlation was slightly less than non Score-Adjusted even-strength action.

I then decided to isolate both the best and worst teams at shot suppression during the 2007-15 time frame.

Ryan W 5

Ryan W 6

Even as we isolate teams that are on the extreme edges of shot suppression we do not see any correlation with save percentage.

Not all shots are created equal so I decided dive into the shot volume vs. shot quality dynamic.

I compared each team’s Fenwick Against Per 60 to their Scoring Chance Against per 60 numbers to see if shot quality could potentially skew the data.

Ryan W 7

There is a definite positive correlation (75%) between Scoring Chances Against Per 60 and Fenwick Against Per 60.


Perhaps it is time to curb the belief that the team in front of the goaltender is to credit/blame for a goaltender’s successes/failures.  Over a sample size of eight seasons the data shows that there is no correlation between the play in front of a goaltender and the goaltender’s ability to play well.  A goaltender can be judged on his own merit without dealing with the noise that comes from team related variables.

*All data was taken from War-On-Ice*

Leave a Reply

Your email address will not be published. Required fields are marked *