Friday, June 13, 2008

Vezina Trophy 2008

Congratulations to Martin Brodeur on winning the 2008 Vezina Trophy. For the first time in his career, I believe that he was deserving of the award. It was a bit of a wide open race with Luongo having an off year. No goalie was able to particularly distinguish himself in save percentage, as 10 netminders ended up finishing between .920 and league leader Dan Ellis' .924. Both Nabokov and Lundqvist had the type of seasons we have traditionally associated with Brodeur - winning lots of games and posting slightly above average save percentages while facing few shots per game on good teams. Those are not the type of seasons that are generally deserving of a Vezina Trophy, as they tend to be a greater reflection of team defensive strength than outstanding goaltending play. Of the 3 goalies who were nominated I would definitely select Brodeur. Brodeur had a high save percentage on a team that was solid but not outstanding defensively, and likely ended up finishing fairly high in shot-quality neutral save percentage as well. I think the best goalie was either Brodeur or Tomas Vokoun, who had a strong year (.919) while facing the most shots in the league on a weak team in Florida.

Overall, Brodeur faced more shots than usual and especially more power play shots as the Devils were not as disciplined under Brent Sutter (only Kiprusoff faced more shots on the penalty kill this year than Brodeur). It has been primarily Brodeur's play on the penalty kill that is responsible for his improvement in recent years, and that was shown again this year with a .903 save percentage when a man down, 4th best among starting goalies. At even-strength, Brodeur was just 12th among starters. His .932 was actually behind rivals like Leclaire, Luongo and Hasek, and it was penalty killing that ended up making the difference.

Can Brodeur keep it up next year having turned 36? Based on standard career curves, we have to expect some decline as he continues to age, and it is doubtful that his team support will be any better next year. Penalty kill save percentages tend to be more variable than even-strength ones, so it is probable that Brodeur will regress to the mean somewhat in that area next season. Another potential warning sign is that according to Hockey Numbers, Brodeur led the league in save percentage against difficult shots (>20% chance of being a goal) with 74%. He was just 21st against average quality shots, which tend to be the most consistent category for goalies and the best one for ranking. Brodeur did exhibit a similar breakdown last year (and subjectively to me has exhibited that tendency throughout his career of making a great save and then letting in a stoppable one), so he probably has some ability to keep making the toughest stops, but will he retain that ability as he gets closer to 40?

Despite this year's setbacks, Roberto Luongo is waiting in the wings to take over the popular mantle of the best goalie in the league (in the eyes of many he already has), and is likely to return to form next season. Especially if Vancouver's new GM can surround him with talent Luongo is more than capable of having an Hasek-type elite season. If anyone can hold off the pursuit of time you would imagine it could be Brodeur, the NHL's goalie iron man, but his age plus likely worsening team factors plus a prime Luongo indicate that there is a fairly good chance this Vezina trophy will end up being Brodeur's last.

33 comments:

Triumph said...

'the devils are a worsening team'?

absurd. they have a ton of cap room upcoming and while they do have a lot of veterans on their roster, none of them are entering serious decline. the defense is especially young.

the devils will undoubtedly be better in front of brodeur - it's just a question of whether brodeur can maintain his play from this past season.

3remarkd said...

Many hockey fans are moaning and groaning about Evgeni Nabokov not winning the Vezina this year. However, his .910 SP (save percentage) ranked 23rd in the League. Decidedly mediocre. He shouldn’t have even been nominated. If he had won the Vezina, then 22 other netminders should have been handed a Vezina at the same time.

Nabokov faced only 24 shots per game, and had a GAA of .910. In the meantime, J.S. Giguere faced 27 shots per game and yet had a lower GAA (2.12) than Nabokov, due to his superior .922 SP. So if anyone was “robbed” of a trophy, it was Giguere.

Ironically, Nabokov’s SP was identical to that of Martin Gerber, often criticized this season for poor play. If the two had switched teams and schedules, with Nabokov facing 30 shots per game and Gerber facing 24 shots per game, Nabokov would be feeling the heat, while Gerber would be an award candidate.

While Nabokov led the NHL in wins, he also tied for the lead in games played, so we would expect him to rack up a lot of wins. Recall that last season Andrew Raycroft played 72 games and tied a Maple Leaf franchise record for most wins in a season, but did it with a miserable .894 SP – hardly an impressive achievement. Most goalies (unless injured) want to play every single game, but that decision is out of their hands. Moreover, wins is a team statistic. If I suggested that Tomas Kopecky was a better player than Vincent Lecavalier because he had more wins this season, I would be laughed out of the rink.

Some Ranger fans protest that Henrik Lundqvist should have won, pointing to his League-leading 10 shutouts. However, his .912 save percentage ranked 18th in the NHL. Lundqvist gave up 4 or more goals 12 times in 72 games, while Giguere (with only 4 shutouts) gave up 4 or more goals only 6 times in 58 games. Moreover, Lundqvist gave up 5 or more goals 5 times, while Giguere went the entire season without giving up 5 or more goals. So who better deserves a Vezina – the goalie who is consistently excellent or the goalie who is perfect one night and lousy the next?

The SP rankings reveal that the three best goalies during the NHL regular season were Dan Ellis (.924 SP), Ty Conklin (.923 SP) and Giguere (.922 SP). If you want to set a minimum games played criterion of 55 games (two-thirds of 82 games), then the three best goalies were Giguere, Tim Thomas (.921 SP) and Marty Brodeur (.920 SP).

At least the Vezina went to a top goaltender this year. In 2003, Brodeur won the trophy despite having a .914 SP that was 15th in the NHL. That same season, Marty Turco led the League with a SP of .932. The following year, Brodeur’s .916 SP was again 15th in the League, and again he was awarded a Vezina. Mika Kiprusoff, Dwayne Roloson and Roberto Luongo had SPs of .933, .933 and .930 respectively, but voters were excited because Brodeur led the League in wins and shutouts. That season, Brodeur played 75 games, facing a little over 24 shots per game, while Luongo played 72 games, facing 35 shots per game, so naturally Brodeur won more games. Does anyone doubt that if Brodeur had played for Florida and Luongo had played for New Jersey, Luongo would have had more wins?

By focusing on wins, you tend to award goalies for middling performances in easy situations, and ignore goalies who put forth amazing performances in difficult situations

Bruce said...

If he had won the Vezina, then 22 other netminders should have been handed a Vezina at the same time.
... The SP rankings reveal that the three best goalies during the NHL regular season were Dan Ellis (.924 SP), Ty Conklin (.923 SP) and Giguere (.922 SP).


3remarkd: Tht's if you drink the KoolAid that it's a simple formula that "Best Sv%= Best Goalies". It's important, but it's just one column in a large table of stats. Voting the Vezina on just that one criterion would be like voting for the Hart based on most goals. Sometimes you'll wind up with the right winner (see: Ovechkin this year), but usually it's a little more complex than that.

voters were excited because Brodeur led the League in wins and shutouts

Those columns matter too, as do GAA and GP, not to mention a bunch of stuff that doesn't even get quantified at all. Brodeur consistently is among the leaders in all those categories, as he was yet again in 2007-08.

The Contrarian Goaltender said...

The SP rankings reveal that the three best goalies during the NHL regular season were Dan Ellis (.924 SP), Ty Conklin (.923 SP) and Giguere (.922 SP).

I have to echo Bruce on being careful with save percentage. It is definitely the best place to start, but team effects and randomness can have a big impact, especially on someone who played only part of the season like Conklin.

You also need to consider the goalie's teammates. Look at the other goalies they play with and the team around them to try to judge the strength of the defence, look at save percentages broken down by game situation (PP, PK, ES) and if possible look at shot-quality neutral stats such as the ones posted by Hockey Numbers or Hockey Analytics.

Tht's if you drink the KoolAid that it's a simple formula that "Best Sv%= Best Goalies". It's important, but it's just one column in a large table of stats. Voting the Vezina on just that one criterion would be like voting for the Hart based on most goals.

I disagree with the analogy. Forwards have 2-way responsibilities while goalie is a defensive position. A forward can trade off goalscoring for defence, but there is not much a goalie can do to replace making saves since keeping the puck out is the point of goaltending.

If we could perfectly measure save percentage with consistency across the league and then adjust it for exact shot location, save location, shooter and game situation, then I think we would have a close-to-perfect stat to evaluate goalies with. At least until we find some evidence that puckhandling or some other skill is important and needs to be included in the analysis.

Those columns matter too, as do GAA and GP, not to mention a bunch of stuff that doesn't even get quantified at all.

What extra insight do all those stat columns really provide? Wins are probably 80-85% the result of the team in front of the goalie. Even in an extreme case where a goalie is creating a lot of wins on an average team, how are they accomplishing it? Primarily through making saves, so the superior performance is sure to be reflected in other stats like save percentage.

Games played is useful in terms of defining the significance of the save percentage sample, but beyond that, does more GP = more better goalie? How do we know it isn't the result of a veteran reputation, coaching preference, or a weak backup?

And what does GAA really provide, since, GAA = (1 - save percentage) x shots per game? It simply restates a teammate-polluted version of save percentage, unless you subscribe to the belief that goalies can significantly influence the number of shots against, which I still haven't seen convincing evidence for.

Shutouts are an arbitrary stat, important only because we have developed a tradition of measuring and posting them. Why do we look at shutouts, but not one goal games? Why is a 16 save shutout (the "Brodeur special") more significant than stopping 50 out of 51? Unless you are going to consider 1-goal games, 2-goal games, 5-goal games, etc. as well, then I don't see any value in shutouts. Goals are already counted in save percentage and GAA, so there doesn't seem to be any reason to award bonus points for 0 GA.

So what is the value of wins, GAA, shutouts, and GP? I've looked at all of them in depth, and I just don't see much value there, which is why I put a lot of weight in save percentage and variations thereof. You may call that "drinking the Kool-Aid", but just assuming that since shutouts and GAA are tracked and recorded then they must be valuable doesn't seem to me to be particularly reasoned either.

Now "the bunch of stuff that doesn't get quantified at all", that is the interesting part. I think we're better off trying to find new ways to look at puckhandling, rebound control, clutch play, etc., and try to evaluate them individually and then add it to our save percentage knowledge, rather than trying to deduce something from the existing set of flawed stats. Of course it is difficult and may never be possible, but it is still the overall goal.

Bruce said...

If we could perfectly measure save percentage with consistency across the league and then adjust it for exact shot location, save location, shooter and game situation, then I think we would have a close-to-perfect stat to evaluate goalies with. At least until we find some evidence that puckhandling or some other skill is important and needs to be included in the analysis.

Aye, there's the rub. I say there's all kind of evidence that stuff is important, as we discussed under Top All-Time Goalies, now in "older Posts". But its inferential rather than statistical. To summarize my case from that discussion, Sv% alone is not a fair way to compare the following:

Goalie A (call him "Marty") excels at puck retrieval and distribution and rebound control and contributes to team strategies that minimize shots on goal.

Goalie B (code name "Roberto" is a great stopper who doesn't leave his crease much and who keeps the puck in front of him because he'll probably stop the rebound anyway.

If Goalie A's contribution to team defence (as in shot prevention) is as little as -2 shots per game, and Goalie B's is +2 shots per game, that's the equivalent of a Sv% difference of ~ .010.

Of course it's just guesswork, by definition shots not taken are not counted. But I've watched a lot of hockey and paid a lot of attention to what the golaies do, +/- 2 or 3 shots seems to me like a reasonable difference between a sweeeper-keeper and a pure stopper. Even if it's as little as 1 shot, that's gonna warp your almighty Sv% into something less than the Holy Grail of goalie stats.

You may call that "drinking the Kool-Aid", but just assuming that since shutouts and GAA are tracked and recorded then they must be valuable doesn't seem to me to be particularly reasoned either.

They're all valuable, some more than others, none perfect. I personally would have a lot higher confidence in Sv% as an indicator than shutouts, say 85% confidence to to 70% confidence. But I'll never have 100% confidence in any stat ... which is what I mean by "drinking the Kool-Aid". Probably closest is Wins since that is the major currency in the league, but of course you are all correct that it's a team stat, it's not just the goalie who wins games. Stats all -- ALL -- have to be considered in context. Just as there are circumstances that will help Goalie A pile up wins, there might be circumstances favouring Goalie B maintaining a high Sv%.

does more GP = more better goalie?

Does more TOI = "more better" defenceman? I say yes, if with the extra workload he keeps outperforming the replacement level even if his rate stats might drop slightly. It's axiomatic that the best players generally log the most ice time. A 70+-GP high-performance goalie is like a 26-minute defenceman or a 22-minute forward -- pure gold. So absolutely, that is an important stat, especially if the guy wins most of his games. The GMs and PHWA sure seem to value it when they're picking the trophy winners and All-Stars, and I tend to agree with that.

The Contrarian Goaltender said...

Of course it's just guesswork, by definition shots not taken are not counted. But I've watched a lot of hockey and paid a lot of attention to what the golaies do, +/- 2 or 3 shots seems to me like a reasonable difference between a sweeeper-keeper and a pure stopper.

Your arguments are reasonable and logical, and they make sense. However, the data simply do not support them. I've looked at backup stats, in considerable depth. The team context far overwhelms individual stats. +2 or 3 is a reasonable difference, between different teams. Like, for instance, Marty Turco's backup goalies have faced 3.0 fewer shots per game when playing on his team than when they were playing on other teams. Brodeur's backups faced 2.3 per game fewer in New Jersey than they did elsewhere. Then when you compare Turco and Brodeur to their backup goalies, the results are a difference of 0.3 shots per game and 0.7 shots per game respectively. So the difference between the two best puckhandlers in the game and their goalie teammates is less than 1 shot per game over their careers. Again, where is the effect?

I looked at 16 recent starting goalies, trying to see who had the fewest shots against compared to their teammates. The leader was Arturs Irbe. Here is how hockeygoalies.org describes Irbe's puckhandling: "Very poor stickhandler", "Can also get into trouble when handling the puck away from the net", "Irbe is one of the worst puck-handling goalies ever." My position remains that puckhandling is a very minor aspect of goaltending.

Even if it's as little as 1 shot, that's gonna warp your almighty Sv% into something less than the Holy Grail of goalie stats.

OK, that's fine. We can, you know, translate save percentage into an absolute number by, for example, multiplying by number of shots and comparing to league average. Then if we figure out that the goalie saves, say, 1 or 2 shots per game hypothetically, then we can easily add in that effect. Either that or just adjust the rate itself like you did in your comments on an earlier post. That is what I am advocating - start with save percentage and make adjustments. I have no problem with making adjustments if they are justified, I am actually advocating it (that's what they do for shot-quality neutral save %). I think that is a far better method than trying to find some evidence in the meaningless traditional stats like GAA (a worse version of save percentage), shutouts (arbitrary), or wins (team stat).

I personally would have a lot higher confidence in Sv% as an indicator than shutouts, say 85% confidence to to 70% confidence.

What is your confidence in "one-goal games" or "5+ goal games"? Is it also 70%? Do you agree with my point that shutouts are arbitrary and meaningless?

But I'll never have 100% confidence in any stat ... which is what I mean by "drinking the Kool-Aid".

I know you were saying that more in reference to one of the anonymous posters, but I won't have 100% confidence in any stat either, at least until we start using my ideal super save percentage that I described above. But if there is a stat that brings nothing to the table, then I'll have 0% confidence in it, whether it is traditionally used or not.

It's axiomatic that the best players generally log the most ice time. A 70+-GP high-performance goalie is like a 26-minute defenceman or a 22-minute forward -- pure gold.

In general, players that play more minutes are better, I agree. But minutes depend both on a player's skill and the skill of their teammates. Vinny Lecavalier doesn't play 30 minutes a game on the Canadian Olympic team like he does on Tampa Bay. Jason Spezza is just as good a player playing 5 minutes a game on the 4th line of a world championship team as he is playing 25 minutes a game in Ottawa. The difference is in their teammates.

Same thing with goalies - the skill of the backup is just as important as the starter. It is much, much easier to have a team with two good goalies on it than it is to have a team with four great lines. So you have Evgeni Nabokov and Vesa Toskala on the same team, and they split the starts down the middle. Then Toskala leaves, and suddenly Nabokov plays every game and is nominated for the Vezina. Either he sure got a whole lot better this season, or maybe, just maybe, his backup got a whole lot worse. Good for Nabokov, he played a lot of games, he was valuable to his team, but I don't see him as any better because his playing time went up.

Anonymous said...

i cant even read your garbage......goes on and on and on.....

you need a life big time.

a witness said...

Contra-Goalie - i love your stuff. Well worded arguments without letting passion get into it - pushing things toward extremes...you stay focussed and clear. well done.

Truth is no numbers fully tell the story and in the case of goalies and defensemen, the numbers we do collect are lazy stats so that SOMETHING is tallied. Neither positions are adequately accounted for. I do think that the goalie is much easier to see in numbers than a defenseman. And i think you are on the right track with the idea of making some modifications to the save percentage stat. I mean, ultimately a goalie is only as good as his ability to stop the puck from going in the net. And sv pct covers that very well.

Rebound control is something that i think makes a goalie better but i think that is almost covered in the sv pct because a goalie that gives up fat rebounds is going to give up more goals while only producing maybe two more shots for that goal to occur.

Stick-handling is the Brodeur-stat. Way overblown as to its actual impact.

anyway, thanks for posting.

Bruce - appreciate your style as well. Even though i lean toward what C-G is saying, you are respectable as are your thoughts on it. The fact is numbers and stats will always leave room for debate...even in baseball, where numbers do almost tell you everything about performance...

Anonymous said...

Interesting site.

It'd be too bad for the Brodeur-worshippers if Saint Marty had ever had a decent-to-good backup (maybe Dunham?), especially at the outset of his career. Instead, NJ's other & pre-M.B. goalies were so terrible (Chevrier? Billington? some of the worst relative svpct's in their respective seasons) that MB was given big-time minutes & was relatively much better than the other NJ goalies.

Play a lot of games & perform generally OK (in the measurable sense; e.g. savepct is by far the best stat, SQN% is even better) for a team that allows few shots & PP opportunities against = the brilliant Marty Brodeur!

(Oh but of course, Marty makes the saves "when he has to", & only allows goals "when they don't matter", haha)

ContraG is right though; MB has been much better since the lockout.

Bruce said...

//Of course it's just guesswork, by definition shots not taken are not counted. But I've watched a lot of hockey and paid a lot of attention to what the goalies do, +/- 2 or 3 shots seems to me like a reasonable difference between a sweeper-keeper and a pure stopper.//

Your arguments are reasonable and logical, and they make sense.


Thank you. I'm glad we are able to maintain a civil tone throughout all this, even as we take opposite sides of the discussion most days. Call me the (ahem) Devil's Advocate.

However, the data simply do not support them.

I agree that is problematic. But I'm still not sure how best to interpret that data. For the sake of further discussion ...

I've looked at backup stats, in considerable depth. The team context far overwhelms individual stats. +2 or 3 is a reasonable difference, between different teams. Like, for instance, Marty Turco's backup goalies have faced 3.0 fewer shots per game when playing on his team than when they were playing on other teams. Brodeur's backups faced 2.3 per game fewer in New Jersey than they did elsewhere. Then when you compare Turco and Brodeur to their backup goalies, the results are a difference of 0.3 shots per game and 0.7 shots per game respectively. So the difference between the two best puckhandlers in the game and their goalie teammates is less than 1 shot per game over their careers. Again, where is the effect?

Some cocktail of:
1) Small number statistics for backups
2) Lower strength of opponent for backups
3) Team tightens up for backup
4) Team plays the same defensive system for both goalies, and backup tries to fill his role
5) Team/organization teaches same system to all goalies as they develop, and in practice

I looked at 16 recent starting goalies, trying to see who had the fewest shots against compared to their teammates.

I'll accept your statistics at face value, knowing as I do that you would never, ever twist a stat to make Brodeur look bad. :-D

The leader was Arturs Irbe. Here is how hockeygoalies.org describes Irbe's puckhandling: "Very poor stickhandler", "Can also get into trouble when handling the puck away from the net", "Irbe is one of the worst puck-handling goalies ever." My position remains that puckhandling is a very minor aspect of goaltending.

Oh yeah, Irbe was bad. Real bad. But he did love to wander, and while he cost his teams goals, he may have saved them shots in the process, which is all you're counting here. He was pretty good at rebound control as I recall, which is an important part of the equation.

BTW, that site you keep referring to seems badly out of date. But I do note this comment about Irbe: "Strong along the ice. Opponents usually aim high against the small goalie." Which makes me wonder if there might be more missed shots against such a goalie, given that high shots can go over the net whereas low shots are unlikely to go under it. Yet another wrinkle which might show up in the modern Missed Shots data for certain goalies if one cared to look. I don't.

That is what I am advocating - start with save percentage and make adjustments. I have no problem with making adjustments if they are justified, I am actually advocating it (that's what they do for shot-quality neutral save %).

I'm with you so far, it's just gonna take a long time if ever before all the appropriate adjustments can be figured out.

I think that is a far better method than trying to find some evidence in the meaningless traditional stats like GAA (a worse version of save percentage), shutouts (arbitrary), or wins (team stat).

Those stats may be less good (lower confidence) but to call them meaningless is a stretch which I don't accept. For the first 65 years of the league's history they're all we've got, so we're better off trying to figure out how reliable those stats are in light of modern metrics, rather than simply write them off.

What is your confidence in "one-goal games" or "5+ goal games"? Is it also 70%?

Those would be interesting stats. Somebody keeps them (Elias Sports Bureau?). I would have similar confidence in one-goal games as shutouts, and I'll bet the leaders are most of the same guys. As for the 5+ goal games, that is a negative stat which would also carry meaningful information, some of it about the tendencies of the coach when his stopper is having an off-game.

Do you agree with my point that shutouts are arbitrary and meaningless?

They are arbitrary, but they are far from meaningless. Raw career numbers don't tell the whole story, e.g. George Hainsworth is third on the all-time list with 94 but only led the league twice; he happened to play in the First Dead Puck Era. Clint Benedict on the other hand led the NHL in each of its first seven consecutive seasons with just 19 total shutouts. Since then, here are the number of times different guys have led the league:

6: Glenn Hall
4: Alec Connell, Tiny Thompson, Jacques Plante, Ken Dryden, Ed Belfour, Dominik Hasek, Martin Brodeur
3: Harry Lumley, Terry Sawchuk, Ed Giacomin, Tony Esposito, Bernie Parent, Patrick Roy
2: Georges Vezina, George Hainsworth, Charlie Gardiner, Dave Kerr, Frank Brimsek, Turk Broda, Bill Durnan, Roger Crozier, Mike Liut


To my eye that's a pretty comprehensive list that includes most of the greatest goalies of all time and very few pretenders. Maybe a small surprise to see, say, Roger Crozier and not Johnny Bower, but for the most part it seems a decent metric. I can't speak to all the old guys like Benedict, Connell, Thompson and Lumley or most of the guys from the last group, but from this list we can divine that Hall, Plante and Sawchuk were the standard bearers from the Late Original Six; Dryden, Parent, Esposito and Giacomin were the best of the 70s; and from the Second Dead Puck Era Brodeur, Hasek, Belfour and Roy had 15 shutout championships among them, and no other modern goalie more than one. That's the exact same four guys we identified earlier as being the best of their generation, which makes the shutout metric a long way from meaningless.

//But I'll never have 100% confidence in any stat ... which is what I mean by "drinking the Kool-Aid".//

I know you were saying that more in reference to one of the anonymous posters,


Yes.

but I won't have 100% confidence in any stat either, at least until we start using my ideal super save percentage that I described above. But if there is a stat that brings nothing to the table, then I'll have 0% confidence in it, whether it is traditionally used or not.

So what stat brings nothing to the table? Ties? They all tell us something, even if we have to provide the context to try to make sense of them.

The Contrarian Goaltender said...

I'll accept your statistics at face value, knowing as I do that you would never, ever twist a stat to make Brodeur look bad. :-D

I don't have to. That's the great thing about being on my side of the debate, the truth doesn't lie :-D

Some cocktail of:
1) Small number statistics for backups
2) Lower strength of opponent for backups
3) Team tightens up for backup
4) Team plays the same defensive system for both goalies, and backup tries to fill his role
5) Team/organization teaches same system to all goalies as they develop, and in practice


I agree that these factors have an effect, but surely they are not particular to New Jersey and Dallas alone? They are very likely to apply to every team across the board, and so shouldn't the marginal impact of a great shot-reducing goalie still show up in the data? Points 4 and 5 suggest that the system has a big impact, which doesn't exactly imply that the elite puckhandling skillset is really that unique or valuable. If you can just teach any goalie the system, then who needs Brodeur, just throw in the backup and show him how to keep shots against low.

He was pretty good at rebound control as I recall, which is an important part of the equation.

Yeah, I think Irbe was pretty good at rebound control, and that is probably the main reason for the difference. Irbe faced fewer shots than his counterparts even when up against much better puckhandling goalies like Tom Barrasso. I agree that rebound control is important and it matters. I also think that it is already substantially reflected in save percentage since rebound chances are some of the highest percentage scoring chances in hockey.

Yet another wrinkle which might show up in the modern Missed Shots data for certain goalies if one cared to look. I don't.

I cared to look. It doesn't.

For the first 65 years of the league's history they're all we've got, so we're better off trying to figure out how reliable those stats are in light of modern metrics, rather than simply write them off.

Yes, that is an important point, for historical analysis. I should have said those statistics were relatively meaningless, since the reason we shouldn't give them much weight is that we have better ones. But all we have to compare, say, Clint Benedict to Alex Connell are wins, shutouts and GAA, so I agree that it would be valuable to make some sense of them and I have indeed made some efforts in that regard.

Your point about shutouts is also well-taken. It is mostly because of the sampling effects (only good goalie play enough games, etc.), but it still does seem to identify most of the greats. But I do wonder about some of those goalies - a lot of the well-known names show up, but how many of them are known as "greats" because of great team stats rather than outstanding play? If we define past greatness in large part because of shutout totals, then it is circular reasoning to claim that shutouts have validity because all of the greats show up. But I think we agree - there is at least some useful information from shutouts in career totals and times leading the league and so on, but comparing goalies head-to-head over the course of a single season does not have a great deal of value.

So what stat brings nothing to the table? Ties? They all tell us something, even if we have to provide the context to try to make sense of them.

Wins. Wins bring nothing to the table, they are a team stat. For every goalie that is fighting his way to a decent win total on a weak team, there is a weaker goalie riding a strong team to a higher total. They are as likely to be misleading as insightful.

The Dark Ranger said...

I think this is one of the greatest sites ever!!!!! Keep up the good work as my little New York Rangers showed Brodeur how well a Vezina Trophy winner he really is this playoff season.

Marty the Whiner has met a true website.

Thank you. Thank you. Please keep it up.

The Dark Ranger

Bruce said...

Points 4 and 5 suggest that the system has a big impact, which doesn't exactly imply that the elite puckhandling skillset is really that unique or valuable. If you can just teach any goalie the system, then who needs Brodeur, just throw in the backup and show him how to keep shots against low.

Agreed, the system has a big impact. The system is largely developed with the starter’s skill set in mind. Or the coach that develops the system develops the goalie himself to play within it; puckhandling is very much a learned skill, but that makes it no less valuable to have a guy who develops that skill to maximum effect.

I would also suggest that defensive systems which result in low shot totals have a more direct influence on GAA than they do on Sv%. If anything, low shots totals likely have a mildly negative effect on Sv%, a point I have not stressed but consider important. It has always been my observation that scoring opportunities against the flow of play tend to be of higher quality, and it's axiomatic that it's more difficult for the goalie to stay sharp when he doesn't face a shot for minutes on end.

I agree that rebound control is important and it matters. I also think that it is already substantially reflected in save percentage since rebound chances are some of the highest percentage scoring chances in hockey.

Seems reasonable. Doesn’t quite explain Luongo. Not sure we have the stats yet to show the number of shots on rebounds and the Sv% on those shots, maybe they reveal something about the quality of rebound. E.g. Some goalies give up fat ones, others give up short rebounds which are likely to just be driven right back into the goalie, still others aggressively kick pucks into corners or right through the scrums to reduce second shot opportunities, while others simply gobble up first shots and allow hardly any rebounds. In fact, every goalie does all of those things, just in very different proportions.

//Yet another wrinkle which might show up in the modern Missed Shots data for certain goalies if one cared to look. I don't. //

I cared to look. It doesn't.


Good work.

//For the first 65 years of the league's history they're all we've got, so we're better off trying to figure out how reliable those stats are in light of modern metrics, rather than simply write them off.//

Yes, that is an important point, for historical analysis. I should have said those statistics were relatively meaningless, since the reason we shouldn't give them much weight is that we have better ones. But all we have to compare, say, Clint Benedict to Alex Connell are wins, shutouts and GAA, so I agree that it would be valuable to make some sense of them and I have indeed made some efforts in that regard.


Good. I hope the input about shutout leaders will prove valuable in this endeavour.

Your point about shutouts is also well-taken. It is mostly because of the sampling effects (only good goalie play enough games, etc.), but it still does seem to identify most of the greats. But I do wonder about some of those goalies - a lot of the well-known names show up, but how many of them are known as "greats" because of great team stats rather than outstanding play? If we define past greatness in large part because of shutout totals, then it is circular reasoning to claim that shutouts have validity because all of the greats show up.

I was frankly surprised by how well the shutout leaders list I posted identified the greats. It did less well in the 80s when shutouts were an endangered species (and a whole bunch of different guys led the league exactly once each), but it did mighty well at picking out the greats of the 60s, 70s, and the Second Dead Puck Era. I watched all of those guys and to my mind it’s not circular reasoning at all – they were great. This gives me greater confidence in the method for identifying the greats of earlier eras.

But I think we agree - there is at least some useful information from shutouts in career totals and times leading the league and so on, but comparing goalies head-to-head over the course of a single season does not have a great deal of value.

Less value in shorter comparison periods for sure. These things accumulate over time and luck plays a role; lots o' nights the goalie can do nothing wrong and still give one up. Or more than one. Even comparing the same goalie against himself shows a pretty significant fluctuation; e.g. in Brodeur's 12 full seasons he has recorded 6, 10, 10, 4, 6, 9, 4, 9, 11, 5, 12 and 4 clean sheets in his inexorable pursuit of another career record. Equally likely to be 4-6 or 9-12 in a given season, and pretty much a yo-yo effect from one season to the next. Looking at larger samples reveals more stable results: note the extraordinarily consistent two-season totals of 16, 14, 15, 13, 16, and 16 shutouts! (Oops, there I go singing Marty's praises again, which certainly is not the point of this website. :-D) Point is, shutouts have a large standard deviation but the cream rises to the top.

//So what stat brings nothing to the table? Ties? They all tell us something, even if we have to provide the context to try to make sense of them.//

Wins. Wins bring nothing to the table, they are a team stat. For every goalie that is fighting his way to a decent win total on a weak team, there is a weaker goalie riding a strong team to a higher total. They are as likely to be misleading as insightful.


It won’t surprise you that I disagree. The correlation is not 100%, but it isn’t random either. Strong goalies tend to make their teams stronger and win more often. Strong teams don’t tend to acquire, or keep, weak goalies.

Using the same method as I did for shutouts, here are the number of times different goalies have led the NHL in Wins:

8: Martin Brodeur
6: Clint Benedict
5: Tiny Thompson, Terry Sawchuk, Jacques Plante
4: George Hainsworth, Bill Durnan, Glenn Hall, Ken Dryden
3: Ed Giacomin
2: Alec Connell, John Ross Roach, Normie Smith, Frank Brimsek, Turk Broda, Harry Lumley, Tony Esposito, Bernie Parent, Grant Fuhr, Patrick Roy


Not exactly a bunch of nobodies, in fact most of the same names from all eras. There’s Sawchuk, Plante and Hall again, the same three Late Original Six greats identified by the shutouts metric. There’s Dryden, Giacomin, Esposito and Parent again too. From the modern era Brodeur dominates this category to the extent that the other guys are under-represented. The split of 8-2-1-1 for Brodeur-Roy-Hasek-Belfour as Wins leaders is less representative in my view than the 4-3-4-4 split of the same foursome as shutouts leaders, where the four of them (correctly, in my view) stand head and shoulders above the rest. But both lists are interesting, and there are not too many bums to be found on either.

Finally, I can't let this list pass without considering team effects. Much has been made around here about how Brodeur benefits from playing on a great team, but I wouldn't put the 1993-2008 Devils within a light year of the Canadiens of Jacques Plante, Ken Dryden, or Bill Durnan; and certainly lesser than Clint Benedict's Senators, Turk Broda's Maple Leafs, Terry Sawchuk's Red Wings, Grant Fuhr's Oilers. Those guys all played a lot of games, too, in many cases virtually all the games. Yet Marty Brodeur ranks ahead of all of those guys in winning games and winning seasons. Hmmm, maybe this guy is pretty good ...

Anonymous said...

Check out Sawchuk's save pct's (the stats are available, although not "official")... an early very good career, followed by years of so-so/up & down seasons. Yet finished with 5 seasons of win-leading.

Dryden led in wins 4 times but only played 7 full seasons? Too bad he didn't hang on & play past 32.

I get a little tired of reading about how many wins Brodeur has, & number of games he's played, & won 2-3 cups, so therefore even if he rarely finished in the top 5 in save pct he somehow is one of the greatest of all time. That's just weak, circular reasoning.

Anonymous said...

Bruce, again you are waaay over-rating goalie "puckhandling"... there is no way that a team's "defensive system" has that much impact on, or impacted by, a goalies supposed puckhandling wizardry.

Goalie puckhandling wizardry might be a big deal in rec league (where such a wizard might even be able to stickhandle past people to centre ice), but at anything above beer league it has very very little impact on "defensive systems", other than the occasional choice to shoot-in/dump-in or not.

Stop grasping at straws, trying to find something, ANYTHING, that validates Brodeur as somehow being one of the greatest of all time.

The Contrarian Goaltender said...

If anything, low shots totals likely have a mildly negative effect on Sv%

It has been verified that the relationship between save percentage and shot totals is either very mildly negative or zero (see here or
here). It is probably correct to say that the number of shots faced has no significant effect on save percentage. The other trends you describe might be very mildly negatively correlated as well in the aggregate, but team shot quality against is far more significant than merely the shots against. You can find teams in all 4 quadrants, if you will, from high shots/high shot quality to low shots/low shot quality. Based on the shot data available, Brodeur has fallen more often than not in the latter category which is obviously a big advantage to have as a goaltender.

Not sure we have the stats yet to show the number of shots on rebounds and the Sv% on those shots, maybe they reveal something about the quality of rebound.

We have some stats, so let's look at what they say. I'm pulling these from The Forechecker.

Based on this definition (a shot within 2 seconds of a previous shot with no stoppages in between), the chance of scoring on a rebound shot was 28% league-wide. The best save percentage of any team was 82% (Chicago), and the worst was 60% (Washington).

So let's say there are 2 goalies, both facing 30 shots a game, both stopping 90% of them. If one goalie starts giving up one extra rebound chance against per game and allows rebound goals at the average rate, his expected save percentage drops to 27.7/31, or .894, a drop of .006.

Rebounds are very dependent on the strength of the team - the 5 worst teams were Toronto, Pittsburgh, Edmonton, Los Angeles, and Florida. Florida gave up almost twice as many rebounds as league-leading San Jose. That, in my view, is why Luongo got a rep for giving up rebounds - he played on the league's worst defensive team. Move him to Vancouver, and suddenly that problem is not so obvious (this year Vancouver had the 5th fewest rebounds against).

Looking at larger samples reveals more stable results: note the extraordinarily consistent two-season totals of 16, 14, 15, 13, 16, and 16 shutouts! (Oops, there I go singing Marty's praises again, which certainly is not the point of this website. :-D)

What about these two-season totals, aren't they also extraordinarily consistent?

26.8, 26.0, 23.6, 24.8, 23.1, 26.6, 27.5

72, 70, 72, 73, 74, 78

Shots against and game played, both strongly correlated with shutouts. The consistency of the defence in front of him certainly contributed to the shutout totals.

Yet Marty Brodeur ranks ahead of all of those guys in winning games and winning seasons. Hmmm, maybe this guy is pretty good ...

Yes, Brodeur must be pretty good.....at playing a lot of games. Sure, a lot of the other guys played every game of every season, at the same time that every other goalie in the league was also playing every single game. Brodeur had the good fortune of being the leading workhorse at a time when the other top goalies did not play every game, and in fact the average has been closer to 55-60 games per season. If virtually any goalie on a half-decent team had played Brodeur-type minutes in the 1980s, they would also have led the league every year in wins.

By the way, don't underrate the Devils. Sawchuk's Wings and Dryden's Habs I'll give you, but the other teams weren't that much better than the New Jersey, as these winning percentages show:

Edmonton (w/Fuhr): .639
Montreal (w/Plante): .636
Ottawa (w/Benedict): .622
New Jersey (w/Brodeur): .620
Montreal (w/Durnan): .619
Toronto (w/Broda): .562

New Jersey has been an elite team for almost Brodeur's entire career, and in terms of defensive play they have been one of the very best in league history. Combine that with a good goalie who plays a lot of games, and that's why you have a guy who leads a lot in wins and shutouts.

Bruce said...

Anonymice:

Check out Sawchuk's save pct's (the stats are available, although not "official")... an early very good career, followed by years of so-so/up & down seasons. Yet finished with 5 seasons of win-leading.

Yup, that was Sawchuk. Great for five years, led the league in Wins every year and never did again. Had a GAA between 1.90 and 1.99 every year and never again got below 2.38. He had problems with severe stress, mononucleosis, depression, alcoholism and injuries. He took early retirement in Boston in 1956-57, and while he returned to play until his untimely death, he was never again the phenom that took the league by storm.

Dryden led in wins 4 times but only played 7 full seasons? Too bad he didn't hang on & play past 32.

Yeah, he could have led the league 6 or 7 times instead of just 4. Durnan was similar, a short but utterly brilliant career. Both retired way too young.

I get a little tired of reading about how many wins Brodeur has, & number of games he's played, & won 2-3 cups, so therefore even if he rarely finished in the top 5 in save pct he somehow is one of the greatest of all time. That's just weak, circular reasoning.

Well some people rely on Sv% to tell them everything they think they need to know, and some of us look at the whole bundle of statistics, which include unimportant things like winning games and championships. If you get tired of reading about that, then I suggest you ignore reading my comments, because I refuse to drink the KoolAid that Sv% tells the whole story.

Bruce, again you are waaay over-rating goalie "puckhandling"... there is no way that a team's "defensive system" has that much impact on, or impacted by, a goalies supposed puckhandling wizardry.

The above comment includes exactly one reference to puckhandling. If that's waaay overrating it, then I guess I'm guilty as charged.

Goalie puckhandling wizardry might be a big deal in rec league (where such a wizard might even be able to stickhandle past people to centre ice), but at anything above beer league it has very very little impact on "defensive systems", other than the occasional choice to shoot-in/dump-in or not.

It affects how the defensive team lines up at the blueline, which frequently forces the dump-in choice. Jersey has lined four guys up on its own blueline for years and basically dared teams to dump-and-chase. Having a reliable sweeper back there has greatly facilitated this style of play as I see it.

Stop grasping at straws, trying to find something, ANYTHING, that validates Brodeur as somehow being one of the greatest of all time.

I guess you didn't read the parts about the league leaders in shutouts and wins? Neither of which has (much) to do with puckhandling, they just help us identify who have been the great goalies over the years. Brodeur ranks high on one of those lists and atop the other one. But it isn't Sv%, so I guess it's just a mirage.

BTW, I have never once claimed that Brodeur is the greatest goalie in history, but he's certainly one of them. He ranks second all-time in Wins and fifteenth in Losses. He's second in Shutouts, third in Games Played, fifth in Saves, sixth in Points Percentage, seventh in Save Percentage, eighth in Goals Against Average. To have a discussion about the best goalies of all time and not include a guy like this because you don't happen to like him is patently ridiculous.

Anonymous said...

Bruce: Brodeur is "certainly one of them" (greatest goalies in history)?

2nd in Wins & SO's - yep
3rd in GP - yep

... & as proven over & over, these are HEAVILY dependant on playing in front of a great defensive team that allows few shots against & few powerplays against.

It's also misleading to point out his career saves (only officially kept track of since '83); his UNadjusted career save pct; & his UNadjusted GAA. Where do the goalies who played the majority of their career in the 80's rank? Oh right, the league avg save pct & GAA were relatively poor (compared to today), so they will be passed by one of the all-time greats, Martin Brodeur. Another accomplishment by Brodeur!

Did I ever say I didn't happen to "like" Brodeur? Since the lockout he's definitely been one of the top 5-10 starters in the league. Prior to that... not so great, & usually a bit above avg & occasionally below avg.

It's obvious you look to his superficially impressive career stats & think that somehow, someway, he MUST be one of the all-time greats. I suspect that for years you "drank the Koolaid" & admired the mighty Marty & marvelled over his stats, & so when someone rationally discusses actual #'s & makes it apparent that MB is not quite so "great", you just can't accept it.

I imagine this is what it's like trying to discuss the theory of evolution & the fossil record with a die-hard creationist, haha.

Anonymous said...

p.s. Bruce: your bias towards crude stats like GAA & wins for Sawchuck... it may be that a 2.38 avg (whether by Sawchuk or someone else) could be comparatively better than a 1.90 season... depends on the # (& type) of shots faced.... that's why save pct is far better than GAA & wins & shutouts to rate goalies.

Anonymous said...

re: Bruce, again you are waaay over-rating goalie "puckhandling"... there is no way that a team's "defensive system" has that much impact on, or impacted by, a goalies supposed puckhandling wizardry.

The above comment includes exactly one reference to puckhandling. If that's waaay overrating it, then I guess I'm guilty as charged.

*Actually, in your other posts you went on & on about the positives of having a 'great' puckhandling goalie (however that is defined or measured... we don't know), trying to ascribe some special talent to Brodeur that would make him appear better than his stats indicate. Once the site author shot that down, you sorta stopped mentioning it...

Bruce said...

It's also misleading to point out his career saves (only officially kept track of since '83); his UNadjusted career save pct; & his UNadjusted GAA. Where do the goalies who played the majority of their career in the 80's rank? Oh right, the league avg save pct & GAA were relatively poor (compared to today), so they will be passed by one of the all-time greats, Martin Brodeur. Another accomplishment by Brodeur!

The point is that everything that has been kept track of, since 1983 or since 1917, Brodeur ranks amog the leaders, whether it be counting numbers (wins, shutouts) or percentges (Pts%, Sv%, GAA). Sure the guys from the 80s are outta luck, guys from other eras not so much. e.g. the all-time list of GAA leaders:

1. Alec Connell 1.912
2. George Hainsworth 1.933
3. Charlie Gardiner 2.024
4. Lorne Chabot 2.039
5. Tiny Thompson 2.077
6. Marty Turco 2.148
7. Dave Kerr 2.149
8. Martin Brodeur 2.201
9. Dominik Hasek 2.202
10. Ken Dryden 2.235


Looks like an era advantage to me ... just not this era. Being in the top ten is no mean feat.

I imagine this is what it's like trying to discuss the theory of evolution & the fossil record with a die-hard creationist, haha.

I would imagine it is. I'm kinda feeling the same way myself.

Thanks for the compliment tho', I really appreciate it.

Anonymous said...

Bruce, there is a very simple way to era-adjust for save pct. & GAA (compare to league avg, etc.).

But you continually fall back on unadjusted stats to say, 'see, Brodeur is high on this list, therefore he must be an alltime great' without making any attempt to explain that save pct. & GAA has varied widely (as I'm sure you're aware) or make the adjustments.

It's lame to just say that 80's goalies are "outta luck" because the league avg. save pct & GAA were relatively poor, & then put Brodeur above them all because he has had the fortune of playing in a low-scoring, high-save pct. era.

What's next? Your thesis on why Joe Malone was far greater than Gretzky because he scored 2 goals per game whereas Gretzky could barely score above 1 per game?

Yeesh.

Bruce said...

Bruce, there is a very simple way to era-adjust for save pct. & GAA (compare to league avg, etc.) But you continually fall back on unadjusted stats to say, 'see, Brodeur is high on this list, therefore he must be an alltime great'

My major point above is that he is high on all the lists -- GP, Wins, Win%, Sv%, GAA, SO. I don't see how that is compatible with him not not being an all-time great. (Key word: "an", not "the")

without making any attempt to explain that save pct. & GAA has varied widely (as I'm sure you're aware) or make the adjustments.

Of course I am aware, and I have stated so more than once. As for making adjustments, I'm no computer whiz and am still just tapping in to some of the sources and resources available. Other information is not available ... e.g. the effect of puckhandling and the goalie's role within team systems to limit shots against totals. You ridicule me for trying to evaluate these unquantified attributes which are not yet in the statistical lexicon, and then you ridicule me for being old-fashioned and sticking to traditional statistics. I like to consider the whole shmear, from what we've always had to new information which is emerging to what we still don't have but maybe should try to develop. I particularly value traditional stats simply because they are all we have for historic comparisons. And I refuse to accept there is a single Superstat which captures everything, no matter how sophisticated the method. Hockey isn't a one-dimensional game, even for goaltenders.

The main reason I refuse to kneel at the altar of the Church of Save Percentology is that there are only 25 years of data, and two widely distinct eras that make comparisons even within that quarter century span very difficult. By all means, compare against league average and it will reveal that Patrick Roy had some years that were greater than they appear. So let's add Roy to our list of all-time greats. Hey wait a minute, he's been on my list all along.

As for Brodeur, he has a career Sv% of .9135 during a span when the league average was .905. At league average rates his GAA would be 2.42 rather than 2.20, and we probably wouldn't be having this discussion. Even in this, supposedly his weakest area, he is far above league norms.

It's lame to just say that 80's goalies are "outta luck" because the league avg. save pct & GAA were relatively poor, & then put Brodeur above them all because he has had the fortune of playing in a low-scoring, high-save pct. era.

I don't think I put Brodeur above them all ... I merely cited on-the-record career statistical totals that did so. My "lame" comment was meant to acknowledge that some goalies were hurt by their era, while others were helped. Brodeur certainly belongs in the latter group, but even he was helped less than the likes of Connell, Hainsworth, Gardiner, Chabot and Thompson, every last one of whom started their careers between 1926 and 1928. Of course there are era effects!

What's next? Your thesis on why Joe Malone was far greater than Gretzky because he scored 2 goals per game whereas Gretzky could barely score above 1 per game?

Nope. I'm an Oiler fan, and Gretz was the best ever! But it's no stretch to say the Great One was helped (although not as much as Malone) by era effects: he was fortunate to play at a time when league-wide scoring was at an all-time high, he was fortunate to play on a great team with five other Hall-of-Famers, he was fortunate to be healthy. All of these factors helped to maximize his production. You could -- and CG has -- make similar arguments about Marty Brodeur, that he has played on the "right" team in the "right" era, which have helped to maximize his statistical impact in his particular position. And guess what? I agree with that. His statistics make him look better than he really is. In fact, when he retires he will likely be the top statistical goalie of all time. Whereas I'm not claiming he's the best goalie of all time, just that he deserves to be considered when the discussion takes place.

The Contrarian Goaltender said...

The point is that everything that has been kept track of, since 1983 or since 1917, Brodeur ranks amog the leaders, whether it be counting numbers (wins, shutouts) or percentges (Pts%, Sv%, GAA).

This is true. However, in virtually every modified stat that I have looked at that includes a team element or a league-average normalizer, Brodeur does not rank among the leaders. This includes league-adjusted save percentage, performance compared to backup goalies, shot-quality neutral statistics, playoff winning percentage compared to regular season winning percentage, even-strength save percentages, head-to-head save percentage results vs. opposing goalies in the playoffs, and a bunch more that I don't even remember right now but that you can find somewhere on this blog or in work done by other stats guys. So that is the justification for my ranking. It isn't that I unilaterally dismiss Brodeur as an all-time great, I just think that taking everything into account he doesn't measure up to the very best of the best.

Now, granted, one of the problems is that I do know a lot more about Brodeur's team situation that I know about, say, Dryden's team, or Plante's or Durnan's, so maybe I'm unfairly penalizing Brodeur for advantages that a lot of other guys enjoyed as well, but I'm just going on the available information.

Anonymous said...

But Bruce, it is misleading when you list Brodeur as being in the top 10 "alltime" in things like GAA & save pct -- & indicate that this is another reason for him to be 'one of the alltime greats' -- when you don't adjust for seasonal differences.

You use these misleading, unadjusted stats to "prove" your point, then when called on it you shrug your shoulders & indicate that you are just using the data you're given, & that goaltending is some sort of nebulous, "black art" that is far too difficult to measure properly --- as if just simply stopping the puck wasn't the OVERWHELMING duty of a goaltender (puckhandling, passing, filling up the waterbottles... whatever... excluded).

It's not that difficult to compare each goalie's SVPCT to the league avg for each year (at least, since 1983), then weight each goalie's year by minutes played, then calculate a career comparison to league (as a % of league avg, NOT as + or - the league avg).

The last time I checked (prior to the lockout), I don't think Brodeur was even in the top 30-40 of goalies who played a certain # of minutes in their career. And this is just since 1983... ignoring all the goalies that were likely better than him, pre-83.

Since the lockout, Brodeur's performance has been quite good, but I doubt that has (yet) weighted his career stats such that he is anywhere near the top 10 (since 1983).

Of course, SVPCT itself is a bit of a raw measure, & the better quality shotqualityneutral SVPCT has only been available for the past 3-4 seasons... & that indicates that Broduer has faced EASIER shots than the avg goalie (rendering his raw SVPCT not as impressive as it appears).

Anonymous said...

Bruce - "The main reason I refuse to kneel at the altar of the Church of Save Percentology is that there are only 25 years of data, and two widely distinct eras that make comparisons even within that quarter century span very difficult. By all means, compare against league average.."

Correct me if I'm wrong, but I don't think you have compared to the league avg SVPCT for all (regular) goalies since '83... haven't you just looked at Brodeur's own (1994-08) era? If so, that's only using about 1/2 the available (1983-08) era.

Bruce said...

It's not that difficult to compare each goalie's SVPCT to the league avg for each year (at least, since 1983), then weight each goalie's year by minutes played, then calculate a career comparison to league (as a % of league avg, NOT as + or - the league avg).

First of all, if this was my day job, maybe I would do that ... since it's so easy, maybe you want to do it. Oh wait a minute, scratch that idea, you don't even have time to sign up with even a pseudonym. (I hate never knowing which "Anonymous" I may be talking to around here)

I have already admitted I am no computer whiz which is what you need to be for that kind of across-the-board number-crunching. Otherwise it's mighty goddam time-consuming. Thanks for volunteering me, though.

Second of all, why do you want to weight this stat on minutes played? It's Sv%, which is Sv/SOG without regard to GP or MP. If it has to be weighted, do so on shots faced.

Not sure weighting it by anything accomplishes a lot, besides a bunch of extra work. The first-string goalies tend to play a consistent # of minutes from year to year and the league-wide Sv% doesn't fluctuate that much year-over-year. If the NHL average Sv% is .905 for Brodeur's career and his is .9135 and he plays 70+ games every year, do you really think a weighted average is going to be more than +/- .001 of the much-more-easily-established rate of +.0085 above the league norm? For an obvious exception like Hasek, who was a backup for some seasons, injured for others, and retired for still others, and who therefore played a disproportionate number of his games during the Deadest Puck Era, I guess weighting makes sense. Ultimately rate stats like Sv%-over-league need to be converted into a true number of goals saved to matter much. Obviously the more games played and shots faced by the above-average goalie, the better.

Any goalie-vs.-goalie or goalie-vs.-league weighting that is done by minutes played, gross shots against, total saves, or any other "workload" metric will work in Brodeur's favour, since he is among the league leaders in all those categories every year. e.g. since the lockout:

Season -|- GP | SOG | Sv
----------------------------------
2005-06 | 3rd | 2nd | 2nd
2006-07 | 1st | 2nd | 1st
2007-08 | 1st | 4th | 2nd
-----------------------------------

The ability to finish higher on the Saves list than the Shots list is no mean feat, btw.

Of course, SVPCT itself is a bit of a raw measure, & the better quality shotqualityneutral SVPCT has only been available for the past 3-4 seasons... & that indicates that Broduer has faced EASIER shots than the avg goalie (rendering his raw SVPCT not as impressive as it appears).

SQNSv% is more than a little raw itself, judging by this damning essay by Alan Ryder, "Product Recall Notice for 'Shot Quality' ",

http://tinyurl.com/5xkcad

... which shows that the stat is subject to systematic and significant bias by the home shotcounter. In it Ryder reveals home/road differentials in shot quality for and against from +0.408 (!) to -0.172; apparently shots in Ranger games are 41% higher quality in games in MSG than they are in away games, whereas for the Sabres the comparable figure is 17% lower. While the Devils' differential isn't as extreme as some teams, the guy in the Swamp tends to count a significantly lower quality of shots for and against the Devils by a much greater margin than occurs in their road games.

Yet somehow I am supposed to have great faith in this newly developed and highly problematic stat rooted in the notorious RTSS with no past history to compare it to, while consigning traditional measures like GAA, shutouts, and yes, Wins to the trash bin of obsolescence??

as if just simply stopping the puck wasn't the OVERWHELMING duty of a goaltender (puckhandling, passing, filling up the waterbottles... whatever... excluded).

Stopping the puck is still the most important duty of the goaltender, but the position has gotten a little more complex since the 1940s when the guys would never leave their crease, and goalie coaching was restricted to "I don't care how you do it, just stop the puck." I don't happen to agree with the assertion that puckstopping is 100% of the modern goalie's job, or 99.9% or even 99.5%. Depending on the goalie, 90 or 95% maybe. But we've been over that already, you think goaltending is a one-dimensional job and I don't. I see it like a baseball pitcher, whose primary duty will always be to throw the ball but also has to field his position, hold the runners, control the tempo of the game, and do numerous other small things which don't look hard when done well, but will impact runs allowed (both earned and unearned) and ultimately Wins and Losses. Some pitchers are just pitchers, and others are ballplayers. Other things being close to equal, give me the ballplayer every time.

Baseball does a better job measuring that stuff (see my Martinez vs. Maddux comparison at the bottom of "The Masses Weigh In On Goalies" below), but to me the hockey goalie has a roughly equivalent set of supplementary duties besides simply stopping the puck. Like pitchers, some goalies are just goalies and others are hockey players.

The Contrarian Goaltender said...

Sorry to butt in on your debate, but just had to take exception with this comment by Bruce:

Any goalie-vs.-goalie or goalie-vs.-league weighting that is done by minutes played, gross shots against, total saves, or any other "workload" metric will work in Brodeur's favour, since he is among the league leaders in all those categories every year.

He is now, but that certainly wasn't always the case. Take, for example, 2001-02:

1st in GP
14th in SOG (!)
17th in saves (!!)

When you are comparing, say, Brodeur vs. Luongo in 2003-04, and the difference in shots against is something like 10 per game, it makes a big difference whether you weight by minutes or shots against. For most of his career, weighting by shots against does not favour Martin Brodeur. Most people see shots against as a team effect, so that is why it is normal to weight by minutes played.

Anonymous: I appreciate your input, and I think we share a lot of the same views on goaltending, but I generally find the best way to convince someone is to calculate and post the numbers myself, rather than to tell them how to do the number crunching. If you have any numbers available, or would be willing to do the math, I'd be interested in seeing them posted. I've done things like this myself, of course, but it always is good when it can be verified by someone else's work.

Bruce said...

Sorry to butt in on your debate,

Any time, CG, it's your site. :)

For most of his career, weighting by shots against does not favour Martin Brodeur. Most people see shots against as a team effect, so that is why it is normal to weight by minutes played.


I agree SA favours him less than by MP, but he's still among the leaders most years. My point was if you're going to weight Sv% at all, you might as well weight it by its own denominator rather than some unrelated metric, whether that is more in Brodeur's favour or not (it isn't).

But the post-lockout stats I posted -- which include the three greatest shots totals of Brodeur's career by a wide margin -- were in specific reply to these comments:

The last time I checked (prior to the lockout), I don't think Brodeur was even in the top 30-40 of goalies who played a certain # of minutes in their career. ... Since the lockout, Brodeur's performance has been quite good, but I doubt that has (yet) weighted his career stats such that he is anywhere near the top 10 (since 1983).

Citing anecdotal stats that are four years out of date and using qualifiers like "I don't think" and "I doubt" are not going to win me over.


I generally find the best way to convince someone is to calculate and post the numbers myself, rather than to tell them how to do the number crunching. If you have any numbers available, or would be willing to do the math, I'd be interested in seeing them posted.

Agreed ... not that anybody around here ever seems very convinced by any of the numbers I post. :)

Anonymous said...

It's me... Anonymous... I'm currently away from the city (& my stats pile) so I'll post the minute-weighted savepct compared to league stats one of these days. Could be several weeks as I'm in the middle of a move & will have to dig my 'stuff' out of boxes. You can either trust me on my previous statements, or copy/paste the stats into Excel & do it yourself... arrgh, back when I started it there wasn't a suitable website to copy/paste from, so I entered the data by hand.

Anonymous said...

As for whether or not goaltending is 'only' 90%+ puckstopping, it would be interesting to see how many times a typical goalie (& Brodeur) touch the puck per game/60 min's.


"I see it like a baseball pitcher, whose primary duty will always be to throw the ball but also has to field his position, hold the runners, control the tempo of the game, and do numerous other small things which don't look hard when done well, but will impact runs allowed (both earned and unearned) and ultimately Wins and Losses. Some pitchers are just pitchers, and others are ballplayers. Other things being close to equal, give me the ballplayer every time."

But how does a goalie control the tempo of the game? While pitchers can get outs by catching, throwing to base, etc., I don't see goalies making that many plays with the puck... other than stopping it from going in the net, which is reflected in SVPCT. A pitcher (in the NL) can even contribute to offense by hitting... goalies do very little directly to contribute to offense.

Bruce said...

It's me... Anonymous... I'm currently away from the city (& my stats pile) so I'll post the minute-weighted savepct compared to league stats one of these days.

'K. No worries, mon, you got more important things on your plate.

As for whether or not goaltending is 'only' 90%+ puckstopping, it would be interesting to see how many times a typical goalie (& Brodeur) touch the puck per game/60 min's.


I agree, it would be interesting. I have asked myself that question more than once, but have never taken the time to enumerate and answer it. At a guess, the goalie touches the puck more often on "other" plays than on direct shots on net. Many of these are routine plays, granted, but then again, many saves are pretty routine as well.

But how does a goalie control the tempo of the game?

By freezing the puck for a faceoff, or not. By stopping the shootaround, or not. By teeing it up for the defenceman and getting the hell out of his way, or not. By reversing the play to the open guy who can walk it out of the zone, or not. By intercepting the icing by the other team's PK in front of the red line and exploiting the line change by hitting the long bomb to the far blueline. Or not.

While pitchers can get outs by catching, throwing to base, etc., I don't see goalies making that many plays with the puck...

No offence, but I think you need to watch more games.

A pitcher (in the NL) can even contribute to offense by hitting... goalies do very little directly to contribute to offense.


Using Brodeur as an example, he has scored 1-31-32 in his regular season career. That's a little over 1.5 points for every 100 goals he's allowed. In the playoffs, he is 1-8-9, or >2.6% of the 344 goals he has allowed. A small contribution, but on the other hand have you watched most pitchers hit?

Goalies are a little more likely to get third or fourth assists than first or second, but to be fair a lot of the recorded assists are pretty incidental, they're not exactly tape-to-tape breakaway passes most of the time. (But they can be.) But if you help your team break out of the zone enough times you're bound to get rewarded occasionally.

Anonymous said...

Bruce: Freezing the puck, stopping a shootaround, teeing up for a skater, intercepting the icing... these are all fairly routine goalie actions. Aside from perhaps stopping a shootaround, I can't remember the last time I (or any other person/commentator) remarked, "wow, what a great icing stoppage! That's why he's an all-star!" haha

I mean, really... I've watched thousands of games at all levels... at the pro/NHL level, the types of goalie actions you describe as somehow dictating the tempo of a game are very routine & don't have much impact on tempo.

When was the last time a coach said something like, "well, we tried to keep the tempo of the game high but that darn opposing goalie didn't allow us to" ? Or, a coach saying, "hopefully our goalie will be on his game & therefore increase (or decrease) the tempo of the game so we can win"?

Doesn't happen. Tempo of the game is largely determined by the skaters on the ice... the ones who actually play the entire ice surface, not the player who rarely skates more than 20 feet.