Re-Awarding the 2008-2010 Vézina Trophies Using Analytics
Part I of a series applying advanced goalie stats to past awards races.
It’s the dog days of the offseason, that gap between free agents being signed to big contracts and free agents being signed to tiny, try-out contracts when there isn’t much to talk about in the hockey world. But the content wheel must churn, and nothing oils it up like Remembering Some Guys. As fun as that is, typically Remembering Some Guys is just kind of nice and nostalgic, leaving our fury and argumentativeness atrophied and unprepared for the rigors of the regular season. So I thought it might be fun to get the best of both worlds by relitigating some long-settled NHL Awards, namely the Vézina. With the development of better and more sophisticated ways of measuring goaltender performance than Wins and Vibes in recent years, why noi look backwards and apply them to goalies from that 2007 and onward period where we have advanced metrics available but didn’t at the time.
Goalie evaluation is a tricky and very frustrating thing, not least because of how damn inconsistent they are from season to season. This isn’t meant to be an open-and-shut case, and you might well disagree with the choices here and should feel free to. As dedicated a hockey fan as I was at this time (age 12 to 14), I’m going purely off the stats we now have available and not my memory or gut feeling.
What Stats Will We Be Using?
(If you’re familiar with goals saved above expected, you can skip this section)
Goalie stats have evolved quite a bit in the past twenty years or so. We’ve gone from wins to goals against average to save percentage to goals saved above average to goals saved above expected in a pretty short amount of time. The overall point of this evolution has been to figure out better ways to isolate a goalie’s performance from that of the team in front of him. It’s easy to recognize now that wins and GAA are team stats, and that they’ll favour netminders who play on great teams. But some stats that are still widely used are also subject to those same effects.
Save percentage was meant to be the answer to the team defence problem by considering the fact that some teams allow more shots against than others. But the problem with save percentage is that, as anyone who’s gotten mad about damn Corsi knows, all shots aren’t created equal. A goalie who faces 10 point shots and saves 9 will look just as good by Sv% as one who stops 9 out of 10 one-timers in the slot. And we know that shot quality is something that has a big impact, and that goalies have almost no control over it. If you watch a team like Barry Trotz’s Islanders, you’ll notice that they keep teams to the outside, forcing a lot of perimeter chances but locking down the slot. Here’s a way to visualize that, using a heatmap from HockeyViz:
If you’re a goalie playing behind that defence, life is pretty good for you - especially when it comes to your save percentage, which will be quite high since you get to stop the easy point shots and aren’t challenged as often by grade-A chances.
This problem also applies to some other stats like goals saved above average (available on NaturalStatTrick and HockeyReference) and quality start percentage (available on HockeyReference). Both of those metrics are fundamentally pegged to save percentage - GSAA just compares a goalie’s full season goals against to what it would be if they had a league average Sv%, and QS% shows how often they were above-average Sv% on a game-by-game basis.
The solution to that problem is goals saved above expected (GSAx). By taking advantage of publicly available data on many of the most important factors that influence shot quality, we can filter out those team defence effects even better by considering shot location, timing, type, etc. While these models certainly aren’t perfect and could be improved if more detailed data on passing and player location was made available, they get us a lot closer to a goalie’s true performance than the more rudimentary stuff.
A lot of people think that GSAx and expected goals models in general work by looking at a shot and deciding if it’s “expected” to go in or not - either a shot is 1 expected goal or 0. That’s not how they work. Every unblocked shot attempt a goalie faces is assigned an expected goal (xG) value based on a calculation that estimates the % likelihood that it will go in based on past data. So those shots from the point might be worth 0.02 xG (2% chance of going in) while the shots from the slot might be worth 0.4 xG (40% chance of going in). In that case, our point-shot-stopping goalie from the example above would have a GSAx of -0.8 while our slot-shot-stopper would have a +3 GSAx.
If there’s one thing holding this stat back from wider use, it’s variety. Multiple analysts have built their own expected goal and GSAx models, including some in the public sphere (TopDownHockey, EvolvingHockey, HockeyViz, MoneyPuck, and Cole Anderson) and the private sphere (Stathletes, ClearSight, SportLogiq, and likely some teams themselves). There’s only one “save percentage, but debates even between different public models - such as rink bias, rebound adjustments, season-to-season baselines - rage on. In this instance I’ll be using the TopDownHockey model, whose goalie stats can be accessed for free here. It’s satisfactorily adjusted for recording error, performs very well, and has the bonus of including a wins above replacement component which translates those goals saved into an estimated wins added by a goaltender relative to a replacement-level player (a third-stringer) over the course of a season.
So let’s get into it.
The Real Winner: Martin Brodeur (NJ)
Runners-Up: Evgeni Nabokov (SJ), Henrik Lundqvist (NYR)
Martin Brodeur won his fourth Vézina Trophy in five seasons in 2007-08, which also happened to be the final win of his storied career. Brodeur played an insane 77 games that season, one short of his own NHL record set the previous season; coupled with 44 wins and an excellent 0.920 save percentage, it’s no surprise that he finished first (albeit in a narrow victory over Nabokov).
But all three of the nominated goalies have something in common: their teams were excellent defensively. The Devils, Sharks, and Rangers all ranked top five in the NHL in expected goals against both in all situations and at 5v5. That means they limited shot quality very well, making their goaltenders’ lives a lot easier.
Throughout his career and since, Brodeur has been widely knocked for thriving behind lockdown defensive teams, and while we have no statistical way of knowing whether that was the case in his pime, it certainly was near the end of his career. Every one of his Devils teams from 2007-08 to 2013-14 was well-above average defensively in terms of quality chances against, even as they struggled more and more to generate offence and seriously contend.
The difference that accounting for the defensive system, which limited shots from everywhere but especially the slot, makes is evident in the difference between his goals saved above average (a league-leading 21.8) and above expected (just 8.3). The same is true of Lundqvist, who hadn’t quite hit his peak and put up almost identical results, and is especially the case for Nabokov who actually slightly underperformed expectations by merely putting up a .909 behind a juggernaut two-way Sharks team.
Brodeur’s season was impressive, especially considering the extent to which the Devils relied on him to play endless numbers of games. But even using a metric that would tend to favour his high games played like Wins Above Replacement, he grades out very good but not quite on top.
Our New Winner: Cristobal Huet (MTL/WSH)
Tim Thomas (BOS), Roberto Luongo (VAN)
A surprise, to be sure. Huet doesn’t have much name recognition these days, and his career ended only two seasons after he joined the Chicago Blackhawks in the summer of 2008. But Huet’s short NHL career had some big highs: he put up a league-best 0.929 save percentage in 2005-06 with the Canadiens, especially impressive considering how high scoring was that year, and was excellent in a 52-game campaign splitting time with the Habs and Caps in 07-08.
Huet had a 0.920 save percentage, the same as Brodeur but more impressive considering that the Habs were a below-average defensive team. His Montréal results were strong - 11 goals saved above expected - but it was once he went to Washington that he went off. He stopped over 10 goals above expected in just thirteen games, a torrid streak in which he went 11-2 while allowing only 21 goals. He got a few Vézina votes, but it’s unsurprising that in a league with six goalies playing over 70 games (something no goalie has done since 2017) the GMs weren’t as impressed.
Along similar lines, Tim Thomas dragged a Bruins team that was certainly not the defensive juggernaut that they would soon become to an 8th seed in the East, a shock considering that they were projected to finish last by the Hockey News. And Luongo’s 35-29-9 record didn’t impress voters after his 47 wins the season prior, but he did excellent work in 73 games behind a transitioning Canucks team that missed the playoffs because of anemic offence. If you really care about games played too much to let 51 GP Huet or 57 GP Thomas win, Lou might be your guy.
Honourable Mentions: Tomas Vokoun (FLA), Ilya Bryzgalov (ANA/PHX)
The Real Winner: Tim Thomas (BOS)
Runners-Up: Steve Mason (CBJ), Niklas Bäckström (MIN)
In 2009, the rest of the league worked out what we already knew last season - Tim Thomas is an elite goalie. He led the NHL in save percentage, led the Bruins to the Jennings trophy, and was a huge reason that they went from listless rebuilder to a 116-point Eastern Conference-winning season in just two years.
Spoiler alert: he’s our Vézina winner too. By a huge margin. So let’s focus on the other two, nominated for the first and only times in their careers.
Steve Mason had one of the weirder careers of the 2010s. His nomination for this award coincided with a Calder Trophy and a 4th-place Hart Trophy finish, one of the most celebrated rookie seasons a goalie has ever had. The main reason is that he led the league in shut-outs with 10 while helping the Jackets sneak into the playoffs for the first time in franchise history (where they were summarily destroyed by the Red Wings). After that, everything went horribly, horrible wrong in Columbus. From 2009 to 2012, Mason cost the Blue Jackets over 104 goals above expected, by far the worst stretch of performance for a netminder by that statistic and 59 goals worse than the second-lowest goalie in that period. Usually when a player struggles that much, they don’t play starter minutes, but Mason’s Calder pedigree kept him in the cage.
Despite this, Mason’s rookie season is still considered lightning in a bottle. But once again, there’s a piece of the puzzle missing: team defence.
“Hold on,” you might say. “The Blue Jackets’ defence was headlined by Klesla, Tyutin, Commodore, and Hejda. They weren’t exactly the 90s Devils.”
No, they weren’t, on paper. But by gum did Ken Hitchcock do everything in his power to make them that way. The Jackets conceded only 2.3 expected goals per hour in all situations and 1.9 at 5v5, both the best numbers in the league. They were boring as sin, as dysfunctional at creating scoring chances as they were good at preventing them. With that in mind, Mason’s 0.916 save percentage is a lot less impressive - he actually saved one goal less than expected and ranked 24th in WAR.
The Wild were almost an identical situation, 5th-best defensively but even more offensively useless than the Jackets in their final season with Jacques Lemaire as head coach. Like last season’s Islanders, the Wild conceded a fair number of shots but limited them to the outside, a strategy which resulted in elevated save percentages for their goaltenders.
Niklas Bäckström had a good season and played a huge number of games, but his 0.923 save percentage isn’t that much higher than his expected save percentage of 0.920.
Our New Winner: Tim Thomas (BOS)
Roberto Luongo (VAN), Tomas Vokoun (FLA)
Thomas’ coming out party was an exceptional season in which he stopped over 42 goals above expected, the third-best campaign on record by that stat and 18 more than the runner-up Luongo.
And then there’s Vokoun, who received only one lonely 3rd place vote. Those who remember the state of the Southeast Division at that point, whose teams basically served the purpose of allowing Ovechkin to score on them, will recall that Florida was not a very good team by any means at this point. In 2008-09, they ranked 3rd-last ahead of only the Islanders and the Kings in expected goals against, but managed to put up 93 points and miss the playoffs by a tiebreaker. How? The answer is Tomas Vokoun, who put up a 0.926 save percentage while playing behind this:
Vokoun scraped out a few Vézina votes in his time but was never nominated due in large part to poor teams in front of him and a lack of gaudy win totals. He should have been nominated this season, particularly over Bäckström whose team also missed the playoffs.
Honourable Mentions: Cam Ward (CAR), Dwayne Roloson (EDM), Jonas Hiller (ANA)
The Real Winner: Ryan Miller (BUF)
Runners-Up: Ilya Bryzgalov (PHX), Martin Brodeur (NJ)
Once again, the voters got it right here. Miller was the best goalie in the NHL by far that season when you factor in the difficulty of his workload. The Sabres were a high-event run and gun team that ranked 22nd in the league at preventing scoring chances, and Miller saved them 44 goals above what could have been expected based on the difficulty of the chances he faced.
The runners-up are a trickier story. Bryzgalov’s rise to prominence was a big story that season, as the Coyotes shocked the NHL by winning 50 games and making the playoffs for the first time since 2002. Bryz was great that season, as he had been in 2007-08, but he certainly was helped by new head coach Dave Tippett’s stingy system, which would fall apart in future seasons (foreshadowing) but which gave the Yotes a top 10 team defence.
Brodeur is a whole other story, as 2009-10 was like 2007-08 on steroids. The Devils were the best defensive team in the NHL, once again limited by their dedication to intensely low-event hockey. Based on the quality of the shots he faced he was expected to put up a 0.918 save percentage and he mustered a 0.916. This was also the last season in which he was an effective starter in the regular season.
Our Winner: Ryan Miller (BUF)
Tomas Vokoun (FLA), Evgeni Nabokov (SJ)
Like I said, this was a legendary season for Miller - arguably one of the best campaigns in NHL history. Not only did he drag the Sabres kicking and screaming back into the playoffs after missing them for two seasons, he did so with a two-week break in between in which he won the most valuable player of the tournament award at the Olympics in Vancouver. All in all he was one of the best goalies of his generation, including in his less-celebrated time in Vancouver and Anaheim later in his career, but 2009-10 was indisputably his masterpiece.
Vokoun just kept thanklessly plugging away down in Florida, keeping his team from utter calamity. He was expected to put up a 0.913 save percentage based on the quality of shots he faced but ended up with a 0.925, a gap that prevented an estimated 25 goals and salvaged 16 points in the standings compared to a replacement-level goalie. That being said, the Panthers were so bad that it ultimately was the difference between last place and third-last place. Team performance in front of the goaltender shouldn’t factor into how we evaluate the goalie himself, but you don’t often see much Vézina love go to guys with 23 wins.
Fortunately, if you prefer your goalies to pile up games played and wins, Nabokov’s final act in San Jose was better than ever. On the stacked 113 point Sharks, the artist formerly known as John was excellent behind a team whose commitment to team defence was getting shakier and shakier each season.
Nabokov ultimately left the Sharks, signing in the KHL before returning midway through the 2010-11 season in one of the most unfortunate re-entry waiver situations ever. He tried to sign with the Detroit Red Wings at league minimum to tandem with a young Jimmy Howard and chase his first Cup, but back in those days a player returning to the NHL from another league had to pass through “re-entry” waivers, making him available to every other team to make a claim. Garth Snow and the Isles didn’t play ball, claiming him and suspending him when he wouldn’t report. Ultimately he turned up at Isles training camp in September 2011 and played three seasons for them before winding down his career with the Lightning.
Honourable Mentions: Tuukka Rask (BOS), Henrik Lundqvist (NYR), Cam Ward (CAR)