Has Matt Murray Been Performing at Starting Goaltender Caliber?

Written By: Sasank Vishnubhatla

Two Stanley Cup victories in three years. Three straight seasons of at least 45 wins. The Pittsburgh Penguins have become the league’s model of consistency and domination over the past decade. The last time the Penguins missed the playoffs, the Miami Heat beat the Dallas Mavericks in the NBA Finals without LeBron James. With franchise stalwart Sidney Crosby as captain for the past 11 years, the Penguins roster have had a robust roster since the first overall pick in the 2003 NHL Entry Draft: goalie Marc-Andre Fleury. However, when Fleury departed to the Las Vegas Golden Knights in the 2017 NHL Expansion Draft, the Penguins needed to move on to young Matt Murray.

Matt Murray is a 24 year old kid from Thunder Bay, Ontario, Canada. Drafted in the third round, 83rd overall, in the 2012 NHL Entry Draft, Murray was deemed too weak and small to become a starting goaltender in the NHL. However, his efforts while playing for the Sault Ste. Marie Greyhounds (OHL) and the Wilkes-Barre/Scranton Penguins (AHL) led to him being called up right before the playoffs in 2016. At age 21, Murray recorded 15 playoff victories and won the Stanley Cup. A year later, he became the only rookie goaltender to win two Stanley Cups, a feat Ken Dryden and Patrick Roy were unable to accomplish.

With Murray’s postseason success, the Penguins moved him into the starting goaltender role. However, after an amazing 2016-2017 regular season and Stanley Cup repeat, Murray faltered in 2017-2018. Be it due to concussions or the passing of his father, Murray struggled to stay consistent. Now, in this fledgling season, Murray’s inconsistency has been on full display.

In his first two games, Murray gave up 11 goals: 6 to Washington on 36 shots and 5 to Montreal on 29 shots. His next game, against the Toronto Maple Leafs, he had a 38 save shutout in Scotiabank Arena. The Penguins would continue their tour of Canada, and Murray would continue to show his inconsistency. Against Edmonton, he gave up 5 goals on 46 shots, but still helped the Penguins win. Two days later against Calgary, Murray gave up a fluke goal in a near perfect win. His most recent start against the New York Islanders was a terrible one; he gave up 4 goals on 9 shots and was pulled halfway through the second period. Coach Mike Sullivan does not blame Murray for his performance however. Instead, he rightfully placed the blame on the defensive core. Nevertheless, Murray has been woefully inconsistent. He currently has a 3.76 goals against average with a 89.3% save percentage. Statistically speaking, he has had only 2 good starts with 3 really bad starts.

In order to determine if Matt Murray can statistically still be a starter, I built a neural network to determine a goaltender’s likelihood of starting an arbitrary hockey game. The code for this model is on GitHub and the data used was taken from Corsica Hockey and Hockey Reference.

This neural network model is built of off five advanced goaltender metrics: goals saved above average (GSAA), delta (adjusted) save percentage (dSv%), low danger save percentage (LDSv%), mid danger save percentage (MDSv%), and high danger save percentage (HDSv%).

Low danger, mid danger, and high danger save percentages are all computed from the Fenwick shooting percentage. If the Fenwick shooting percentage is below 3%, the shot is considered a low danger shot. If the shooting percentage is between 3% and 9%, it is a mid danger shot. Finally, above 9% is a high danger shot. Looking at the league, there are some minor trends in all three. Let’s take a look at the league’s values graphically:

Looking at the league with low danger shots, there seems to be a logistic relationship for goalies who have given up at least one goal. For goalies who haven’t given up a goal low danger goal, like Carey Price and Malcolm Subban, they do not fit this logistic relationship.


Unlike low danger save percentage, mid danger save percentage does not show a visual relationship. Let’s compare this now to high danger save percentage:

Looking at high danger save percentage, there is a strong argument for believing that most goalies are on equal ground when it come to saving high danger shots. Matt Murray falls well in the average for high danger save percentage, as well as for mid danger and low danger save percentage.

 

Here are Matt Murray’s advanced stats:

 

Player GSAA dSv% LDSv% MDSv% HDSv%
Matt Murray -1.67 -0.85 98.39% 90.48% 76.47%

 

For Murray’s GSAA, we see that throughout this entire season, Murray has given up one goal which he shouldn’t have. His expected save percentage is 0.85% higher than his actual save percentage. He’s saved 98% of his low danger shots, 90% of his mid danger shots, and 76% of his high danger shots.

 

Now let’s compare him to a veteran starter Devan Dubnyk of the Minnesota Wild.

 

Player GSAA dSv% LDSv% MDSv% HDSv%
Matt Murray -1.67 -0.85 98.39% 90.48% 76.47%
Devan Dubnyk 6.73 2.11 98.59% 92.31% 85.00%

 

Unlike Murray, Dubnyk is exceeding his expectations with a whopping 6.73 goals saved above average and is performing better than his expected save percentage by 2.11%. Dubnyk is very similar to Murray in the fact that neither are extremely athletic and both rely heavily on their positioning and size. A valuable comparison, in addition to Dubnyk, is current backup of the Murray - Casey DeSmith. DeSmith is a 27 year old rookie from the University of New Hampshire. Originally from Rochester, New York, DeSmith is known for being an exceptionally athletic goaltender and was signed as an undrafted free agent by the Penguins.

 

Player GSAA dSv% LDSv% MDSv% HDSv%
Matt Murray -1.67 -0.85 98.39% 90.48% 76.47%
Devan Dubnyk 6.73 2.11 98.59% 92.31% 85.00%
Casey DeSmith 2.23 1.51 100.00% 90.00% 83.33%

 

From the data, we see DeSmith is right between Murray and Dubnyk in performance. However, when we run all three through the network, we get some interesting results:

 

Player Are they a starter? Percent chance as a starter
Matt Murray No 19.48%
Devan Dubnyk Yes 82.54%
Casey DeSmith No 31.00%

 

Though, statistically, Casey DeSmith is almost the average of Matt Murray and Devan Dubnyk, the neural network deemed that DeSmith is not actually the complete average of Murray and Dubnyk. Murray, who’s streakiness has hurt his stats, only has a 19% chance of being deemed a starter on an arbitrary night, while DeSmith has a slightly better chance of 31%. Dubnyk, a proven veteran, rightfully has an 82.5% chance. With the Penguins goaltenders being so streaky, the neural network deems that overall performance is more valuable compared to hotness. Thus, Matt Murray has not been performing at starting goaltender caliber.

Leave a Reply