In 1982-83, the Venzina trophy was first awarded to the goalie “adjudged to be the best at this position”, after decades of being given to the goalie(s) of the team that allowed the fewest goals during the regular season, introducing the Jennings Trophy to fill that rather trivial niche in the NHL trophy case. This was a great idea… except for the fact that sometimes the adjudgers — the NHL GMs — weren’t always on point with their adjudging.
For example, as I’ve covered before, in 1995-96 Jim Carey won the Vezina despite a 90 GA%−, massively snubbing another fantastic season from Dominik Hašek and breaking up what would have been a run of 5 straight Vezinas. But I don’t think that was the worse snub. It was bad, to be sure, but at least Darren Puppa (the runner-up) had an argument for the award as well.
It pales in comparison to 2003-04. In 2003-04, Martin Brodeur won his second Vezina. And by all accounts he really shouldn’t have. Well, to be honest, he shouldn’t have won his first one either, but this isn’t really about Brodeur so much as it’s about Roberto Luongo. Because he wasn’t just snubbed for the Vezina. I firmly believe he was also snubbed for the Hart as well.
Vezina Snub
Why not Brodeur?
Brodeur’s 2003-04 was very typical of his early 00’s seasons: 71+ games, 4300+ minutes played, GA%− around 95. Solid stuff really. 2003-04 had been his most impressive season since 1997-98.
What’s most impressive about Brodeur’s 2003-04, however, is the fact that he played 75 games, 4555 minutes, had a 94 GA%−, and 10.3 GSAA. Those kind of results over 75 games are incredibly valuable to a team, especially Brodeur’s New Jersey Devils who played a shut-down style. The Devils allowed the 2nd fewest shots against, took the 5th most shots for.
It’s really difficult to maintain an elite SV% of a workload of 70+ games, no matter the era. The only reason it’s more common than you’d expect is that teams with elite goalies are more likely to attempt it.
The top 3 seasons of over 70 games played by GA%− are
Bernie Parent (1973-741) 73 GP, 65.2 GA%−
Tony Esposito (1973-74) 70 GP, 68.6 GA%−
Dominik Hašek (1997-98) 72 GP, 73.0 GA%−
But back to Brodeur, what made his Vezina win this year so egregious?
Well, for one, 10.3 GSAA is a little low for 4555 minutes (i.e. ~93% of the Devils’ total possible minutes!). It works out to roughly 0.14 GSAA/60. Ed Belfour accumulated a very similar GSAA of 10.1 over 3444 minutes, or 0.18 GSAA/60, for Toronto that same season.
Now, where did Brodeur finish in overall by GSAA this season? Well, he finished 10th. This is because, as many will recall, GSAA is (SV%)(Shots Against)−(League average SV%)(Shots Against): and Brodeur didn’t get a lotta shots against. Despite setting the record for most minutes played in a season by a goalie — breaking Marc Denis’s record from the previous season of 4511 minutes played — he was 10th in GSAA. Who was ahead of him? In descending order:
Marc Denis: 66 GP, 13.44 GSAA
Vesa Toskala: 28 GP, 14.68 GSAA
José Théodore: 67 GP, 15.65 GSAA
Evgeni Nabokov: 59 GP, 16.38 GSAA
Miikka Kiprusoff: 38 GP, 21.03 GSAA
David Aebischer: 62 GP, 22.66 GSAA
Andrew Raycroft: 57 GP, 24.24 GSAA
Dwayne Roloson: 48 GP, 28.82 GSAA
Wait, where’s #1? Well, I’ll get to that.
Another good way to measure this would be to compare Brodeur’s shot rate to the league average. In 2003-04, Martin Brodeur faced −3.7 shots/60 more than the average goalie. But again, this is 2003-04 — the tail-end of the dead puck era. How would that translate to a more typical shooting environment (i.e. 30.0 SA/60)? Not too much of an adjustment, actually, but we do get a nice round number in −4.0 SAAA adjusted. Now, we could be charitable and assume that, since NHL GMs — the voters for the Vezina — are supposed to be well-informed about hockey and hockey personnel, that having evaluated these various goalies in as the season unfolded, they could interpret the one element that doesn’t translate into shot quantity: shot quality. It could be the case that, despite the significantly below league-average shot volume on a per game basis, Brodeur actually faced enough high-quality shots to put him on par with the league-average.
But even without a statistical analysis of the likelihood that Brodeur’s fewer shots against were of higher quality than an average-workload goalie with a normal distribution of shot quality across that shot workload, reputation precedes the the late-90s/early-00s Devils. Such an analysis would fly in the face of all previous observations of this team: a stalwart defence, trap-playing team that created offense on the counter-attack and became passive upon taking leads.
Wrapping up on Brodeur, why was he the one the GMs chose as the Vezina winner?
Well, he lead in 2 categories — wins and shutouts — as well as won the William M. Jennings for being the goaltender(s) (minimum 25 games played) that played for the team with the fewest goals allowed. But the two above categories he lead in were due to his extreme usage. I would go into detail about win% or shutout%, but I frankly don’t personally weigh those factors at all in my analysis. Suffice to say, Ed Belfour had a higher win% (57.6%) and shutout% (16.9%) than Brodeur’s (50.6%, 14.7%). So, even within the rational that favours him, it’s not necessarily internally consistent.
Why Luongo?
Let’s start by asking another question: what kind of season should we expect from a Vezina-winner? Well, for pre-xGoal seasons like 2003-04, there should be a balance of 3 factors:
Volume: I don’t have any hard and fast rules for this, but generally for the 1990-2004, I’m typically expecting at least 55 games played (i.e, 2/3rds of an 82-game season). This said, volume is the least important of these three factors.
Shots saved on a rate basis (i.e GA%−): a goalie’s job is to stop pucks. Therefore a stat that tracks the rate at which a goalie stops pucks is ideal to evaluate how well a goalie is stopping pucks (on a surface-level basis). Now, for those among you who think a goalie’s job is prevent goals and thus GAA would be a good stat to evaluate as well, that’s a common misconception. You’ll find that (1) goals are always preceded by shots (2) that the job of the entire defence is to prevent goals — by creating turnovers and suppressing higher-danger shot attempts — and (3) there’s already a trophy for that2.
Shots saved compared to the league average (i.e. GSAA): I’m not going to say that one stat could be looked at to determine a Vezina-winner — that’s not good practice. I am saying that your decision should probably start with GSAA if you don’t have GSAx available. We’ve already been over GSAA. It combines SV%, multiplies it by shots against, then compares that figure the league average SV% times the same number of shots against. It takes a rate and reapplies the volume, then measures against what the average would yield. Is it blind to shot quality? Yes. But so is SV% and GAA. And the win stat is blind to all 3.
So then, given my choice of subject for this article, it should come as no surprise that the NHL leader in GSAA in 2003-04 was no other that Roberto Luongo.
Roberto Luongo: 72 GP, 48.42 GSAA
And in case you forgot what everyone else was at, it’s not particularly close:
Dwayne Roloson: 48 GP, 28.82 GSAA
Andrew Raycroft: 57 GP, 24.24 GSAA
David Aebischer: 62 GP, 22.66 GSAA
Miikka Kipprusoff: 38 GP, 21.03 GSAA
Luongo had saved more than 48 more goals than the average goalie would have facing the same number of shots, and nearly 20 more goals than anyone else, period. And while we’re at it, remember that list we did earlier: top 3 goalies by GA%− (min 70 GP)?Guess who’s number 4
Roberto Luongo (2003-04) 72 GP, 78.0 GA%−
Now, comparing Luongo to the rest of the field in 2003-04, he excels in two somewhat contradictory categories. If we look at the top 5 SV% leaders (min 20 GP)
Dwayne Roloson: 48 GP, .933 SV%
Miikka Kirprusoff: 38 GP, .933 SV%
Roberto Luongo: 72 GP, .932 SV%
Vesa Toskala: 28 GP, .930 SV%
Andrew Raycroft: 57 GP, .926 SV%
And the top 5 goalies in shots against
Roberto Luongo: 72 GP, 2475 SA, 34.9 SA/60
Marc Denis: 66 GP, 1970 SA, 31.1 SA/60
Olaf Kölzig: 63 GP, 1958 SA, 31.4 SA/60
Tomàš Vokoun: 73 GP, 1958 SA, 27.8 SA/60
José Theodore: 67 GP, 1860 SA, 28.2 SA/60
We see Luongo .001 off a tie for the lead in the former and outright leading the latter. Simply put, no one in 2003-04 (1) faced as many shots per 60 minutes as Luongo, (2) stopped as high a percentage of those shots, and (3) did so over as many games as Roberto Luongo.
Roberto Luongo's 03-04 season is simply one for ages. He broke the record for Goalie Point Shares with 20.9, which estimates how valuable players are to their team. According to this stat he accounted for 27% of the Panthers total Point Shares. Now, reading Point Shares in terms of % simply tells you where most of the team’s value is generated. I don’t like the school of thought that thinks “most valuable to his team” should emphasize the “to his team” part. If we are talking about the most valuable players, then that value should be regardless of particular affiliation. Just so we’re clear.
Hart Snub?
Why not Martin St. Louis?
Now, let me start this off by saying that Martin St. Louis is a great player and deserves his flowers. But in hindsight, his Hart trophy and Ted Lindsay trophy wins were actually kind of baffling choices in the first place.
Playing all 82 games in 2003-04, St. Louis scored 38 goals and tallied 56 assists for a league-leading total of 94 points. This sub-100 point total might not really get the people going, but it’s also important to remember that context always matters. Given that this season saw the second lowest goals per game since shots were first tracked in 1955-56, we should probably look at his adjusted goals and assists, which adjust points to an 82-game schedule, rosters to 18 skaters, and a scoring environment of 6 goals/game and 1.67 assists per goal. These figures are courtesy of Hockey reference. Looking at all seasons in the dead puck era — which I personally define as being the seasons of 1997-98 until 2003-043 — his 2003-04 ranks 14th in adjusted points with 109. Keep in mind, however, that it was still the most of the entire season.
Why Luongo?
But how do we compare a great goalie season to a great player season directly? Well, let’s use Point Shares, which is designed to do just that! You might remember Goalie Point Shares from the last paragraph of the last section. A player’s Point Shares are calculated from the sum of their value on offense (Offensive Point Shares) and defense (Defensive Point Shares). Now, from what I’ve seen, Goalie Point Shares can sometimes overemphasize goalie contributions — and in fact, it’s probably designed to do so since for most of NHL history, goalies don’t play full seasons. But even still, between 1997-98 and 2003-04, four goalies had the most overall Point Shares:
Dominik Hašek (1998-99): 64 GP, 16.8 GPS, 0.26 GPS/60
trophies: Vezina
José Théodore (2001-02): 67 GP, 17.4 GPS, 0.27 GPS/60
trophies: Vezina, Hart
Dominik Hašek (1997-98): 72 GP, 18.6 GPS, 0.26 GPS/60
trophies: Vezina, Hart, Ted Lindsay
Roberto Luongo (2003-04): 72 GP, 20.9 GPS, 0.29 GPS/60
trophies: none
Even in an era where leading in Goalie Point Shares correlates to hardware, Luongo is cast aside and ignored.
You might be wondering why this emphasis on Point Shares, and besides the fact that — as I’ve mentioned — it’s the only stat that can compare 1:1 goaltending and skater performances, each Hart Trophy winner lead (or would have lead) in Point Shares for at least one of either skaters or goalies — even St. Louis!
1997-98: Dominik Hašek — lead goalies and league with 18.6 Point Shares
1998-99 Jaromír Jagr — lead skaters with 14.6 Point Shares
1999-00 Chris Pronger — lead skaters and league with 14.8 Point Shares
2000-01 Joe Sakic — lead skaters and league with 15.9 Point Shares
2001-02 José Théodore — lead goalies and league with 17.4 Point Shares
2002-03 Peter Forsberg — lead skaters and league with 14.9 Point Shares4
2003-04 Martin St. Louis — lead skaters with 13.2 Point Shares
What makes this so interesting is that Point Shares is suppose to compare all contributions, so the fact that it mirrors pretty well the voting preferences of the hockey writers in a span of 7 years that saw a defenceman and two goalies win MVP.
Now, my analysis doesn’t punish Jagr for winning the Hart in 1998-995, but it will point out that Martin St. Louis’s 13.2 Point Shares is approximately 62% of Luongo’s Point Shares. 62% of Martin St. Louis’s 13.2 would be 8.2 — or value equivalent to Tampa Bay Lightning teammate Fredrick Modin in 2003-04.
This is why I call this a Hart snub. Luongo had the best season of anyone in 2003-04. He provided the most value to his team and would have made any of the other 29 teams better if you injected his mix of workload and rate quality into their lineup — and perhaps could have thrived more because no other team in 2003-04 gave up as many shots as the Florida Panthers.
Devil’s Advocate
I think there is a large degree of uncertainty that we must face when analysing seasons like this one. One thing we are sorely lacking on Luongo’s 2003-04 is any data indicating shot quality. We can speculate on the concentration of shot quality in New Jersey’s data, but we can guess by their reputation that it’s unlikely to be exceptionally dense with what we would now identify as high- or mid-danger scoring opportunities. At worst, we could assume it’d be comparable to their contemporaries in terms of the concentration of high- or mid-danger scoring opportunities, but their suppression of volume would be what makes their defense elite. It could be that years of received wisdom and contemporaneous analysis was actually incorrect, but I somewhat doubt it, and have no way that I know of to prove it.
With Florida and their league-leading shots against, the picture is much less clear. Looking at their roster, it’s clear that this team had a lot of roster churn. Only 9 skaters played more than 59 games for them — which is weird, because you need 18 of those guys on a roster for all 82 games. For comparison, the Toronto Maple Leafs had 16 skaters play more than 59 games, which makes a lot more sense.
But looking at the quality of their players — at least on D — we can assume that this team was pretty atrocious. By average time on ice, here are Florida’s top-7 Defencemen
Mike Van Ryan — 79 GP, 24:26 ATOI
Jay Bouwmeester — 61 GP, 23:02 ATOI
Adreas Lilja — 79 GP, 19:34 ATOI
Lyle Odelein — 82 GP, 18:53 ATOI
Mathieu Biron — 57 GP, 18:13 ATOI
Branislav Mezei — 45 GP, 17:43 ATOI
Pavel Trnka — 67 GP, 17:07 ATOI
This was not a good defense. Obviously not, it allowed the most shots in the league. But just in case you thought there were household-name defensive defencemen who forced opponents to try for high volume instead of high-quality…no. That wasn’t the case. Behold the anti-Devils: kinda mediocre at offense, terrible at defence. We can therefore conclude, that the Panthers were, at best, in line with the average in terms of the rate of high- and mid-danger opportunities they allowed — which still leads to higher volume, mind you. At worst, they were significantly worse than everyone else and Luongo saved their butts all the same.
Conclusion
The reason why I think people somewhat ignored Luongo is that his value made a bad team into a team stuck between terrible and mediocre. He didn’t push his team into a playoff spot. He didn’t lead an already great team to greater heights.
What Luongo did was show up and face more shots than any other goalie in the league, and stop more of the shots he faced than all but a few goalies, providing more value to his team than any other goalie or player did in the 2003-04 season.
All stats are courtesy of Hockey Reference
With both Parent and Esposito’s 1973-74, it’s important to caveat that the 70s double-dipped in terms of talent dilution with both the WHA and NHL expansion leading to more jobs available for goalies of varying talent-levels. This warrants its own investigation at a later date.
Whose wording of “awarded to the goalie of the team with the fewest goals against” undermines its nature as an individual trophy
From a purely statistical point of view, I believe that this is the correct definition. During this period, Goals per game (per team) dip below and stay under 2.8 and shots hover between 27.5 and 28.3. Stylistically, there were obviously teams already playing the trap from the first lockout and into 1995-96 and 1996-97, but it’s the relative stability of these figures that for me marks the league-wide dead puck era.
Prorated from his 13.6 GPS in 75 GP.
I will, however, passive-agressively point out that Hašek lead the league in Point Shares with 16.8
WTF is with 2004 Vesa Toskala? More GSAA than Martin Brodeur in 28 games? How could that be? Those must've been the most epic 28 games of all time.