Picking the 2019-20 Vézina Trophy Winner Using Analytics
In the race for best goalie, two players stand above the pack.
Picking the year’s best goalie using analytics is a lot simpler than choosing a best defenceman or forward, because a goalie has one job: stopping the puck.* To find the best goalie of the year, all you have to do is find the goalie who was best at that, accounting for factors out of his control like team defence. You don’t have to carefully weigh out defensive vs. offensive play, play-driving vs. goal-driving, or any of the other stuff you have to worry about when you compare skaters.
You also don’t have to use the same broad range of stats. A lot of writers who have dipped their toe into analytics like to use a wide menu of goalie stats ranging from save percentage to shutouts to goals saved above expected to goals saved above average to quality start percentage to evaluate season performance. In my mind, this is just making things unnecessarily complicated. Many of those stats are redundant when used together - like if you used both +/- and GF% to talk about a skater - and most are just outdated and irrelevant.
I only use expected-goal based goalie metrics, and completely ignore the Sv%, GSAA, and QS% stats you can find on HockeyReference. While they used to be very useful when better data wasn’t available, those three stats are fundamentally flawed because they do not account for team defence/quality of shots against. Raw save percentage (which GSAA and QS% are built on) will inherently favour goalies whose teams do not allow great chances. The concept behind GSAA and QS% is good, but can be improved. So I’ll be using these stats alone:
Goals Saved Above Expected (GSAx): How many more goals did a goalie prevent than he was expected to based on the quality and quantity of shots that he faced?
Save Percentage Above Expected (dSv%): How much higher was a goalie’s save percentage than it was expected to be based on the quality and quantity of shots that he faced?
Quality Starts %: In what percentage of a goalie’s games did he exceed expectation (finish with a GSAx above zero)?
Ultimately, when judging a discrete season of work for the Vézina, the GSAx total is what really matters.
The tricky part is that “shot quality” is not 100% settled. Multiple models have been created - publicly and privately - to measure it, using a number of variables including location, type, timing, etc. To cover my bases, I’m going to include all three of the most prominent public ones: EvolvingHockey, MoneyPuck, and Crowd Scout each have their own GSAx models factoring different things. For example, Crowd Scout’s model adjusts in order to not reward goalies for stopping rebounds that they should have prevented in the first place. I am not factoring in Clear Sight Analytics’ rankings because they’re a black-box proprietary model; nonetheless, almost all (sorry Markstrom) of their available numbers are in line with these picks.
*You could theoretically make the case that goalies also have to move the puck, but with no way to quantify the effect that that has, I think it’s fair to exclude it from this analysis.
Honourable Mention: Carter Hart - Philadephia Flyers
The choice between Hart and #3 on my list was a really tough one. He was 3rd and 4th in total GSAx according to EvolvingHockey and MoneyPuck, and even better in terms of save percentage differential. Finally, he was very reliable on a night to night basis, ranking 6th in that category. Where I hesitate to list him in the top three is CrowdScout’s lack of bullishness on him due to his lack of rebound control. He ranked 26th among starters in terms of preventing rebounds above expected, which seems to have inflated his numbers. An impressive season for sure, and definitely in the conversation for a spot on the ballot, but just edged out.
#3: Antti Raanta - Arizona Coyotes
Look, I’ve been known to make, uh, gentle critiques of John Chayka’s tenure with the Coyotes. But one thing he’s knocked out of the park is goaltending. Since 2017-18, Antti Raanta and Darcy Kuemper have had the 3rd and 5th most total GSAx respectively, especially impressive considering their injury woes. Kuemper was a serious challenger for this trophy before getting hurt in December, and Raanta picked up right where he left off. CrowdScout has him ranked 3rd among starters both in terms of total GSAx and save percentage differential. The case against Raanta is his GP (only 33), and it’s a fair one. What he did in those games is impressive, and I think worthy of a nomination, but I am fully ready to admit that you could put someone who was slightly less strong but had a more prominent role (like Hart) in this spot as well.
#2: Tuukka Rask - Boston Bruins
I’ve heard some people say that using above-expected stats for goalies is unfair to guys who play on great defensive teams, because they’re the ones who see their numbers take a hit. Tuukka Rask is proof to the contrary. Playing behind the league’s best defence, Rask nonetheless exceeded expectations by a long-shot, finishing 1st in save percentage above expected in MoneyPuck and CrowdScout’s models. He was as steady as anyone, ranking 2nd in quality start %, 3rd on the penalty kill, and 4th in rebound control among starters. His runner-up status is due in large part to his GP; the Bruins wisely chose to run a 1A/1B tandem to rest their starter, so he played 17 fewer games than the winner. This means he didn’t have the chance to finish 1st in total GSAx. Nonetheless, he deserves full credit for one of the best seasons of his career and a strong runner-up performance.
#1: Connor Hellebuyck - Winnipeg Jets
What Hellebuyck did this season was absolutely remarkable. The Jets were one of the worst defensive teams in the league this season, and as a result he faced the most shots and expected goals against of any goalie. Despite this challenge (and coming off of a down-year), Hellebuyck almost singlehandedly dragged his team into playoff contention. Every model agrees that he led the NHL in goals saved above expected, in large part because he was able to maintain his performance while playing a league-highest 58 times (82% of Jets games). He provided his team with a remarkable 9.1 standings points above replacement - especially impressive considering that the team as a whole only finished with 24.5. He’s not only my Vézina winner, but my MVP favourite as well.