Verifying The Vezina Victor

Vezina.png

Many believe four starters have separated themselves from the pack to vie for the Vezina – but are we forgetting anyone? 

The NHL awards are never without its controversy, and the 2018 version will be no different with some of the most prestigious awards being heavily contested. Although Connor McDavid seems to be locking in his second Art Ross trophy, there are at least 5 legitimate candidates vying for the Hart Trophy as the league’s most valuable player (to his team, if we are using the proper definition). A similar argument can be made for the Vezina trophy this season, awarded to the league’s most valuable goaltender.

At first glance, voting for the Vezina might seem straight-forward. Andrei Vasilevksiy, Pekka Rinne, and Connor Hellebuyck look like the clear front-runners. With a gun to their head, most fans, players, and analysts would give the 35-year-old Rinne his first Vezina trophy. As the ‘safe’ pick, this is the most likely scenario – but does Rinne actually deserve the accolades?

The position of goaltending has puzzled people in analytics for years. It is a position which statistically not only varies substantially from year to year but is also heavily influenced by the team and system in front of him. Outside of 5-10 goalies who are relatively consistent year-to-year, there are 20-30 athletes that constantly fluctuate in rankings, some of whom can take the crown in any given season. The main factor is that there are only 31 (soon to be 32) jobs available in the NHL, and until a goaltender gets a legitimate opportunity as a starter, you don’t really know what you have in a player. There are several examples to point to, from Brian Elliot to Sergei Bobrovsky to Devan Dubnyk to Ben Bishop. All four of these guys have put up Vezina-caliber numbers in the past, but none on the team to which they were originally drafted.

The most recent examples are Scott Darling and Antti Raanta, who have served as backups the last few seasons behind world-class starters Crawford and Lundqvist. Both had their first opportunity as starters this season and have earned very different results. Using almost every goalie metric from Rob Vollman’s 2017 Hockey Abstract, one would think that Raanta and Darling had equal chance to succeed with their new clubs. In fact, one might have even given the edge to Darling. Clearly, that hasn’t been the case, which is why the search for reliable future indicators continues.

Opportunity is everything in the NHL, but no position has it stats altered as much by team play as a goaltending.

The Honorable Mentions

fleury

Flower leads all goaltenders in save percentage, saving 93.1% of shots in 45 starts. (photo: reviewjournal.com)

Before we jump into the analysis of the contenders, I want to give a shout out to goalies who have had amazing seasons and could definitely be in the conversation if not for one key flaw: games played. M.A. Fleury, Antti Raanta, Corey Crawford and Roberto Luongo are goaltenders that are putting up numbers on par with all of our Vezina candidates, but none have started over 45 games this season. That’s 13 games fewer than Pekka Rinne, our contender with the least amount of starts (58), or 22% fewer starts. That is significant enough to keep them out of the race as it is difficult to sustain such gaudy numbers as a goaltender. Sooner or later, like with most statistics, all numbers will regress towards the mean.

However, that’s not to say that any of these players have not performed on a similar level as the contenders, just that they haven’t had the chance to prove they can sustain these numbers for longer periods due to injuries at some juncture. Still, I give them all a round of applause and pat on the back for their performance this season; and isn’t that just as good as a Vezina?

Re-thinking Traditional Metrics

When voting for the Vezina, four main criteria are often judged: Wins, Goals Against Average (GAA), Save Percentage (sv%) and Shutouts (SO). Below we will break down each category, and try to normalize those stats which are largely team dependent. The goal is to analyze each player in a vacuum, to see how each would perform given similar circumstances.

To do this, emerging goaltender statistics will be used, and an explanation of those that are not explained below can be found here: GOALTENDING ANALYTICS GLOSSARY. Shutouts were ignored because they are not great indicators of performance. An 18-save shutout is given the same credit as a 45-save shutout, when in fact they are vastly different performances. That’s why save percentage is the most important figure. Let’s start will the largest team dependent stat – wins.

Wins

Marty Brodeur has been quoted saying, “Wins are the most important goalie statistic because that’s what we are paid to do.” As the all-time leader in goalie wins, I think it’s easy to see why he feels this way.  I think most goaltenders would disagree, given the fact that they rarely get to decide which team they play for and most elite goaltenders like Hasek or Roy would have seen similar success on a dynasty New Jersey Devils roster that finished near the top of the standings year after year. In fact, there was once a blog dedicated to why Brodeur is a fraud.

The truth is, wins, or even winning percentage, are a poor indicator of a goaltender’s performance. A goalie, or team, win a game if they score more goals than they allow. Of course, a goaltender contributes to goal prevention, but they have no significant impact on goal production (puck-moving goalies be damned!). Here’s a comparison of the Vezina contenders, looking at their winning percentage as well as their goal support.

Table 1: Analyzing Win % using Goal Support 

Goalie

Games Started

Win %

G.F.A.

G.A.A.

G.F.A. – G.A.A

Gibson

60

52%

2.83

2.43

0.4

Rinne

58

71%

3.17

2.32

0.85

Vasilevsky

63

68%

3.39

2.58

0.81

Hellebuyck

62

68%

3.45

2.4

1.05

Bobrovsky

64

58%

2.83

2.38

0.45

Quick

61

52%

2.84

2.34

0.5

As you can see, Connor Hellebuyck has had far more goal support than most of our contenders. I have created another column to find the differential in goals-against average and goals for average in an attempt to put all of our goaltenders on an equal playing field. It’s not only important that goaltenders stop the puck, but it’s imperative that they at least out-duel their opponent. There are countless cases where a goaltender can have a great game and still give up three goals or more. The far right column gives you an idea of each goalie’s margin for error to outduel their opponent for the win. Given their win percentage under their current goal support differential, let’s look at what their records would be if they had equal goal support.

Table 2: Win % using Connor Hellebuyck’s G.F.A

Goalie

Games Started

Win %

G.F.A.

G.A.A.

G.F.A. – G.A.A.

Gibson

60

71.6%

3.45

2.43

1.02

Rinne

58

81%

3.45

2.32

1.13

Vasilevskiy

63

73%

3.45

2.58

0.87

Hellebuyck

62

68%

3.45

2.4

1.05

Bobrovsky

64

75%

3.45

2.38

1.07

Quick

61

66.1%

3.45

2.34

1.11

Winning percentage was calculated by taking the GFA & GAA differential, and multplying it by Games Started. By subtracting this Table 2 value from the same Table 1 value, it provides us with a number of added goals for each goaltender. Using their current OT/SO losses, and the established 3 goals = 1 point or 6 goals = 1 win law, a new winning percentage was calculated.

You can see that winning percentage becomes much closer if they are all given equal goal support. Rinne still has a lengthy lead given his fewer games played, but Gibson, Bobrovsky, and Quick show that they would be near the league lead in wins given the Winnipeg Jets’ goal support. Hellebuyck drops from a tie for second into fifth place on our chart, and given that wins are one of his strengths in the Vezina conversation, he seems to be slipping out of contention. On the other hand, Bobrovsky seems to have made a case to be amongst the finalists if you consider winning percentage as a crucial measure. Let’s look a little deeper.

Goals Against Average and Team Defense

The numbers above are not perfect, because a large factor involved is GAA. As mentioned, goalies cannot control the number of goals their team scores but they do contribute to their GAA. But how much does a team’s defensive system (or lack thereof) affect their numbers? Given that GAA is still very much a team-oriented stat, let’s have a look at each team’s defensive play.

Table 3: Team Defensive Metrics

Goalie

SA/60

Team Corsi For % (league rank)

Team GA/60 (league rank)

Gibson

32.8

48.46 (22nd)

2.09 (6th)

Rinne

31.7

51.52 (9th)

2.01 (t-2nd)

Vasilevskiy

32.6

51.71 (7th)

2.23 (11th)

Hellebuyck

31

51.47 (10th)

2.24 (12th)

Bobrovsky

30.4

51.76 (6th)

2.19 (8th)

Quick

30.7

49.89 (16th)

2 (1st)

All statistics from nhl.com & corsica.hockey

This sheds a little more light on the amount of work each goaltender is facing. Gibson just outranks Vasilevskiy for shots-against per 60 minutes of play, which is over two shots more per game than Jonathon Quick. The Ducks are also by far the worst possession team of the bunch, and all but Quick’s Kings are in the top-10. This means that Anaheim is consistently playing without the puck and allowing more shot attempts than any of the other teams in question. Even with their lower-than-average possession numbers, the Kings are still the best overall defensive team in the league allowing exactly two goals against per 60 minutes. Not surprisingly, Rinne’s Predators are a hair away from being a top defensive team with the Kings, while also ranking in the top 10 for possession. That, in part, explains his extremely high winning percentage.
Another important tool to use is Goals Saved Above Average. Although it is not perfect as a cumulative measure (like plus/minus) it does paint a picture of how far above the average each goalie is performing based on his quantity of shots. Here’s a look at GSAA in different situations. Some numbers vary from site-to-site, but for better or for worse here’s what Corsica’s numbers produced:
Table 4: A look at GSAA (50+ Games Played)

Goalie

GSAA All sit. (rank)

GSAA ES (rank)

GSAA 5v5 (rank)

GSAA PK (rank)

Gibson

32.51 (2nd)

14.25 (2nd)

14.09 (2nd)

15.07 (1st)

Rinne

24.5 (4th)

19.49 (2nd)

13.13 (3rd)

-0.42 (62nd)

Vasilevskiy

0.31 (13th)

2.27 (10th)

4.5 (8th)

12.6 (2nd)

Hellebuyck

16.38 (6th)

2.02 (11th)

0.66 (11th)

-5.97 (88th)

Bobrovsky

34.92 (1st)

36.36 (1st)

33.45 (1st)

-1.93 (70th)

Quick

26.5 (3rd)

11.49 (5th)

12.94 (4th)

9.34 (4th)

The first column is most important, as it’s a catch-all statistic, but the others provide some context. Bobrovsky, Gibson, Rinne, and Quick definitely show that they are providing a lot more value to their team than an average goaltender would. If we go by the same law of 3 goals = 1 point in the standings, both Bobrovsky and Gibson have provided their team with over 10 points more relative to an average goaltender.

bobrovsky_sergei6401

Sergei Bobrovsky has inserted himself into Vezina contention using advanced metrics (photo: sportsnet.ca)

On the flip side, it seems that Vasilevskiy’s numbers are heavily influenced by his team, with under a goal saved above average in all situations. In fact, Cam Talbot ranked ahead of him in this category. That means if the two swapped teams their numbers this season would look vastly different. However, we shouldn’t accept this as gospel just yet, as hockey-reference.com has very different results with Vasilevskiy’s GSAA at 16.9 ranking ninth overall in the league, well ahead of Cam Talbot’s -10.93 which has him ranked 60th overall (and aligns with common sense). Corsica’s numbers were used because of the ability to separate each GSAA situation. For what it’s worth, hockey-reference has Rinne ranked 1st, Gibson 3rd, Quick 5th, Hellebuyck 6th and Bobrovsky 8th, still all ahead of Vasilevskiy.

Save Percentage and Shot Quality

Now we come to the holy grail of traditional goaltending statistics, save percentage. It has long been regarded as the most accurate for describing goaltending performance but still has its imperfections. For example, if Goalie A stopped 36/38 shots (0.947), he has a superior save percentage than Goalie B who stopped 23/25 shots (0.920) and is thought to have had the better performance. But this isn’t fair as it does not account for shot quality or man-power situation. If both of the goals Goalie B let in were on the penalty kill and Goalie A was at even strength, the picture changes. Or if most of Goalie A’s shots were routine saves from the perimeter, and Goalie B faced far more high-danger scoring chances, then who really had the better game? The classic quality over quantity debate. That’s why Sv% has been tweaked to account for shot-location and manpower situations.

Table 5: Situational Sv%

Goalie

Sv% All sit. (rank)

Sv% ES (rank)

Sv.% 5v5 (rank)

Sv% PK (rank)

Gibson

0.926 (2nd)

0.927 (7th)

0.929 (8th)

0.920 (1st)

Rinne

0.927 (1st)

0.938 (1st)

0.937 (1st)

0.869 (17th)

Vasilevskiy

0.921(6th)

0.930 (3rd)

0.932 (4th)

0.889 (15th)

Hellebuyck

0.922 (4th)

0.928 (5th)

0.930 (7th)

0.901 (4th)

Bobrovsky

0.922 (5th)

0.935 (2nd)

0.936 (2nd)

0.845 (24th)

Quick

0.924 (3rd)

0.928 (6th)

0.931 (5th)

0.894 (6th)

Looking at the first column, it’s easy to see why these are the Vezina contenders, but it should also be easy to see why further analysis is necessary when discussing the trophy. Even though Rinne’s 0.927sv% is clearly above Vaslievskiy’s 0.921sv%, in reality, that accounts for only three more goals against for every 500 shots against – AKA it’s too frickin’ close! That’s why more context is needed. For example, John Gibson is a God on the penalty kill. Given that Rinne’s Predators are only one spot below the Ducks in the PK rankings, and that his teammate Saros out-performs him in odd-man situations, one can assume that Gibson has been much better at facing high-danger scoring chances and in fact, is the major reason why his team performs so well on the PK. With Rinne and the Predators, it’s more of an ‘in-spite-of’ situation. However, it’s hard to fault Rinne too much as his Sv% ranks first in all other categories. But how much of that is playing behind the best defense in the league and facing low-danger chances? I’m glad you asked.

Table 6: Accounting for Shot Quality

Goalie  xSv% dSv% LDsv% MDSv% HDSv%
Gibson 0.908 1.74 0.984 0.933 0.828
Rinne 0.913 1.35 0.977 0.929 0.828
Vasilevskiy 0.921 0.02 0.977 0.913 0.801
Hellebuyck 0.914 0.83 0.977 0.920 0.809
Bobrovsky 0.904 1.8 0.975 0.921 0.835
Quick 0.909 1.45 0.969 0.932 0.828

Here are images to illustrate the most dangerous scoring areas on the ice borrowed from blueshirtbanter.com

 

This might be the most telling table of the bunch, and further illustrates the findings that Hellebuyck and Vasilevskiy’s impressive numbers are more a result of their team’s play than the rest of our contenders. Expected save percentage (xSv%) tells us what one would expect a goalie’s Sv% to be accounting for the quality of each shot faced. Bobrovsky and Gibson are the class of group, far surpassing the expectations as seen by dSv%, which is the difference between their actual sv% and what is expected. Quick and Rinne also perform very well, even though they play in much better defensive systems. The table also breaks down low, medium and high danger Sv% and again, Hellebuyck and Vasilevskiy fall far below the pack.

Conclusion

After much deliberation and weighing the numbers, my Vezina Trophy nominees are:

  • John Gibson
  • Pekka Rinne
  • Sergei Bobrovsky

Just edging out Jonathon Quick. The evidence above is enough for me to dismiss Hellebuyck and Vasilevskiy, given their team statistics and save percentage numbers. This analysis has provided a lot of great insight and has separated the pack in what seemed to be a very even race. Given the team circumstances, I would award the Vezina to Gibson, but any one of the finalists are extremely deserving candidates. As it is likely Rinne’s trophy to lose, it appears that justice will be served!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s