Though it’s only halfway through the season, the race for the Vezina has begun, and there are already some early front runners. Andrei Vasilevskiy currently leads the Vezina Race, with 22 wins, a save percentage over .94, a GAA of just over 2 and a league leading six shutouts. While Vas leads the league in all these categories, is there a more qualitative way to evaluate goaltenders? Wins, while holding huge importance to the goaltenders themselves, are typically viewed as a team stat. Save percentage and goals allowed average are good stats but they miss a key contextual aspect of goaltending —not all shots are created equal.
According to ESPN (I know, but they’re the only site with a hardware tracker), Vasilevskiy leads the early Vezina race, followed by Hellebuyck, Quick, Rinne, Bobrovsky, and Crawford. I’m going to take a look at these six net minders in comparison to both each other and the league as a whole using some lesser-known statistics.
Below is a combination of advanced and traditional goalie statistics where the value is followed by the overall league rank in parenthesis:
Stats from Jan 8
The Fancy Numbers
Before we dive in, I’ll contextualize some of these less-than-self-explanatory stats.
HDSA%: High Danger Shot Attempt % – the percentage of shots a goaltender faces that are taken from the “high danger zone”
LDSA%: Low Danger Shot Attempt % – the percentage of shots a goaltender faces that are taken from the “low danger zone”
HDSv%: High Danger Save % -a goaltender’s save percentage on shot attempts from the high danger zone (Read: Goaltender ceiling, saving shots you wouldn’t necessarily expect them to)
LDSv%: Low Danger Save % – a goaltender’s save percentage on shot attempts from the low danger zone (Read: Goaltender floor, saving the shots you would expect them to save)
dSv%: Differential Save % – a goaltender’s save percentage minus their expected save percentage (xSv%) calculated based on their HDSA%, MDSA%, LDSA%, and the league average weighted save percentages. (Positive numbers indicate the goaltender is performing better than you would expect the average goaltender to given their shot quality distribution against them; negative would indicate the inverse).
GSAA: Goals Saved Above Average – total goals allowed minus league average save percentage*number of shots that goalie has faced. (Positive numbers indicate the goalie is saving more shot than you would expect the average goaltender to given the number of shot they have faced; negative would indicate the inverse)
Several years ago the guys over at War-On-Ice developed the danger level save percentage stat, which takes into account shot location to create three separate save percentages by danger level and an overall weighted save percentage based on league average numbers. If you take a look at the graphic you’ll see shots are broken down into high danger chances (blue – the crease and slot), medium danger chances (pink – inside the “house” and the high slot), and low danger chances (yellow – the points, the half wall, and below the goal line). Essentially, this method uses the weighted distribution of shot quality to create an expected save percentage based on league average numbers, or xSv%. In the table above, you’ll notice I did not include xSv%, but rather dSv%, which is the difference between a goaltenders actual save percentage and their expected save percentage based on the distribution of the quality of chances they face.
Now if you’re thinking “but…but…this doesn’t seem very scientific! A high danger scoring chance can come from lots of places on the ice”, you’re probably right. A Schmaltz to Kane one timer from the end boards or a Shea Weber slapper with a screened goalie are about as high danger an opportunity as you can get; but would be considered “low danger” scoring attempts using this shot distribution method. If you want to think about it like that then HDSv% is a more “valuable” stat because shots from the slot are almost always high danger while shots from low danger areas are not going to be truly low danger the same percentage of the time. Which is why your elite goaltenders will have good league-relative HDSv%s but not necessarily good league relative LDSv%s.
So it’s not perfect, but generally it is a better way to contextualize shots than weighing all shots as equally dangerous because generally a shot from the crease will be more dangerous than a flipper from the point. Generally. But I digress.
Save Makers and Vezina Fakers
If we only evaluate the “traditional” goaltender metrics, (Wins, Sv%, GAA) the ESPN trophy list would seem about right, but if we look deeper at the advanced number, the story is a little different. I provided the distribution of high and low danger shot percentages for each goalie as a sort of measuring stick. These stats are ones that the goaltender cannot influence, but rather provide context for the situations their defense puts them in and distribution shot quality they typically have to face.
Connor Hellebuyck has received a lot of praise recently, which has brought his name into the conversation for the Vezina, but is it deservingly so? Hellebuyck has faced the fewest high danger scoring attempts in the entire league. On top of that, Hellebuyck’s high danger save percentage is below the league average and the lowest of any of the other Vezina candidates. These numbers point to the fact that the ‘peg defense is able to shield Hellebuyck from facing high danger scoring chances being exposed. If teams can figure out how to penetrate the Jet’s D and get some quality scoring chances off, Hellebuyck will be poised to topple like a house of cards.
In fact, the story gets even worse for Hellebuyck when we look at differential save percentage and goals saved above average. Given the relative ease of the shot quality distribution Hellebuyck faces, we would actually expect the average goaltender to have a save percentage .35 points higher than Hellebuyck. Hellebuyck is the only Vezina candidate who has a negative differential save percentage, while all six others are in the league’s top ten in dSv%. Furthermore, based on Hellebuyck’s GSAA, we would expect the league’s average goaltender to have allowed 2.57 fewer goals through this point in the season based only off shot totals, not shot quality distribution. This means that given the number of randomly distributed chances (and likely a more difficult workload given Hellebuyck’s HDSA%) an average goalie would be outperforming Hellebuyck by just under three goals so far this season. I think the advanced numbers show its clear that Hellebuyck does not deserve to be in the Vezina conversation regardless of how good the Jet’s D in front of him is.
Not All Shots are Created Equal
Of the six candidates, you’ll observe that only Crawford and Bobrovsky are on teams that allow high danger shot attempts at not only a higher rate than the league average, but in the top five highest rates of HDSA in the league. Not only are Crawford and Bobrovsky facing significantly more A+ chances, but also have the two best high danger save percentages of all the Vezina candidates (second and third best in the league respectively). This tells us that these goalies are making more quality stops and bailing their team out much more frequently than any of the other contender, something any halfway Hawk fan could have told you about Crow this season.
Because of this, Crawford and Bob have two of the three highest dSv% in the league, meaning given the amount of high danger chances they face, we would expect them to make far fewer saves. However, despite only facing the 29th most high danger shot attempts, (thus raising his xSv%) Vasilevskiy has the best dSv%, boasting the second highest low danger save percentage and eighth highest high danger save percentage. Moreover, Vasilevskiy also has the highest GSAA by a wide margin, saving 15 more goals than the average goalie would be expected to in his position. These contextual numbers point to Vasilevskiy being the obvious front-runner for the Vezina with Crawford and Bobrovsky both close behind in the hunt.
Thanks for reading and I’d love to hear your thoughts below.
Dan