Tuesday, May 6, 2008

Top All-Time Goalies

I have been asked in the comments to list my top all-time goalies just as a reference point for my analysis on this blog.

My overall premise is that in the majority of cases, the goalie is as good as his team. When I evaluate somebody, I am looking for evidence that they were better or worse than the team around them. I don't care very much about things like wins or shutouts or Cups, as they are very strongly influenced by the team. Instead, I prefer to focus instead on save percentage, individual awards, playoff performance, performance relative to era, and performance relative to teammates.

All-time rankings are always a tough task because of different eras, incomplete historical information, uncertain team contexts, and other challenges, so it always comes down to a bit of guesswork. I am pretty settled on my top 4 guys, but after that I'm still trying to sort out the rest. I tend to think that ranked lists are very subjective and often not very meaningful (is there really a big difference between, say, the guy ranked 12th and 13th on pretty well any list?), so I present my top 4 ranked goalies along with a list of honourable mentions to others who are also in the argument.

1. Dominik Hasek

Most MVP awards by a goalie, most times finishing first among all goalies in MVP voting, highest career save percentage, best official single season save percentage, best ranking compared to his backup goalies, 6 Vezina Trophies.

2. Jacques Plante

Did very well compared to his backup goalies throughout his career. Played on some great Montreal teams, but his late career performance with St. Louis and Toronto was also outstanding. Has a very good playoff record. Won a Hart Trophy. Tied with Hasek for most times leading the league in "Goals Saved".

3. Glenn Hall

Named the best goalie in the league seven times in end of the year All-Star balloting, more times than any other goalie. Wasn't as outstanding in the playoffs, but his teams were likely a major factor.

4. Patrick Roy

The performances of his backups indicate that he was definitely helped by his teams, but Roy still has an impressive record: Excellent playoff resume, very good era-adjusted save percentage results, and 3 Vezinas and 3 Conn Smythes that suggest a strong individual contribution to his teams' success.

Other goalies I think were very good include: Bernie Parent (very high peak value), Ken Dryden (not completely sold on him because of the huge team effect, but his performance numbers are outstanding), Turk Broda (exceptional playoff performer), Tony Esposito (great regular season play makes up for lack of playoff success), and Johnny Bower (did well in save percentage stats put together from the Original Six era).

29 comments:

Bruce said...

Interesting stuff, CG. I've been following goalies pretty closely since I "broke in" back in 1963, and the exceptional six-pack of Bower, Plante, Hall, Sawchuk, Worsley and Johnston were all in their prime to late prime.

I think I have to agree that Dominik Hasek is the best goalie I have ever seen, with Bernie Parent's spectacular run to back-to-back Smythe Trophies challenging the Dominator in quality if not quantity. Transcendent.

Not a bad goalie on your (longer) list, though if I had a quibble it would be naming Tony Esposito and not Martin Brodeur, who has great regular season play AND playoff success on his ample resume. While these comparisons are always imperfect, even Tony O's strengths -- durability, longevity, consistency -- are outstripped by The Fraud. Those Hawks of the 70s were every bit as good a team as the modern Devils but couldn't win the big one even once.

Bill Durnan had some pretty nice numbers too, as did Johnny Bower. Then there's the likes of George Hainsworth, Alec Connell, Clint Benedict, and the immortal Georges Vezina himself who deserve consideration, even as comparisons to more modern goalies are near impossible ...

Anonymous said...

there is no merit to any of this site. you basically pick and choose when stats matter and when they dont. you take into account strength of team sometimes but not all the time, and you never give brodeur more credit than the occaisional, he was decent. to not have him in your top 3 is ridiculous, but to not even have him in your second list is mind boggling.
again, your reasons for you list are good stats here or there, or individual awards, without consideration for anything by which you evaluate brodeur on. brodeur has better regular and postseason stats than pretty much anyone on that list, so to say stats are the reason for any of those goalies is bogus. for other goalies you use era adjusted numbers which again would put brodeur at the top. talk about being subjective.

Anonymous said...

Anonymous: Brodeur "has better regular and postseason stats than pretty much anyone on that list" -?? huh?

You must be just looking at basic stats, not adjusted stats such as 'shot quality neutral save pct.'

The most notable thing about Brodeur (other than his inflated reputation) is the consistently high number of minutes he plays. But I wonder how high he ranks in terms of shots faced per yr? Traditionally, NJ has not allowed very many shots.

Bruce said...

Anonymous2: When you're looking at 90 years of goaltenders, "just" basic stats are all we got. Sometimes, not even those.

Certainly there are more sophisticated stats maintained these days (somewhere). I'm interested in the "shot quality neutral save percentage", but am wary as to the methodology. Is "shot quality" a function of distance? That is a judgment call by individual scorers within each arena, the guys who keep the infamously unreliable RTSS. Besides, shot distance is hardly the best indicator of the quality of the scoring opportunity.

-- How many 50-foot wristers into traffic do you see that are real dangerous? Lots. If the shot makes it through it's a tough stop for the goalie; if it doesn't, he gets no credit.

-- How many three-foot shots do you see that have no chance? Lots. There's plenty of short-rebound goalies who are happy to absorb a whack-whack-whack into the pads, in the process "padding" their Sv% and their SQN Sv%, but how much better are they than the guy who deflects the first shot into the corner or who allows no rebound at all?

And, how much better are they than the guy who faces no shot at all because he made a good play with the puck on the intial dump-in? There's no way to quantify that. But to consider the goalie as a guy who does absolutely nothing but stop the puck, well, that's simplistic and just plain wrong. Certainly that's his most important, and most measurable, contribution; but a goalie who doubles as a sweeper is invaluable, as is one who minimizes second chance opportunities, and to completely disregard that in evaluating him is a mistake.

As for Brodeur and shots faced per year, since the lockout he has finished 2nd, 1st, and 2nd in saves per season. That seems like a lot of work to me, no matter how many minutes it's spread over.

It's instructive to note that (by far) Brodeur's three highest career shots against totals (2105, 2182, 2089) have occurred in the last three years since the lockout. That coincides with the introduction of the no-go zone, and I wonder how much of a "coincidence" it really is. Brodeur used to handle a lot of pucks in those areas, not just dump-ins but cycle-busting. In fact his name was the one I heard cited most often when the league instituted the new rule. (There was a Brodeur Rule long before there was an Avery Rule. :) )

Of course the post-lockout seasons also coincide with the departure of Niedermayer, Stevens and Daneyko and the subsequent mediocritization :) of the Devils' defence. As to how much of the shots increase is caused by which factor, who knows?

All I do know is that Brodeur has responded extremely favourably to the extra workload, and has posted two particularly outstanding seasons in a row. His .920 Sv% was the best of any goalie in the top 10 most saves this year; his .922 Sv% was the best of any goalie in the top 25 most saves in 2006-07. He has responded well to the increased workload of mounting shots totals.

Anonymous said...

Yeah, Brodeur has been better since the lockout. Not sure why. If he had played at this level the previous 10+ yrs he might deserve more of the accolades he receives.

A goalie's main (99.5%?) function is stop shots. If you bring up any other skill (e.g. clearing the puck/puckhandling) you better have some stats to illustrate why this is so important. Even if Brodeur were somehow 10 times better than any other goalie at puckhandling (very doubtful), I doubt this would make much of a difference in terms of goals saved. And most NHL goalies are half-decent to very good at puck-clearing anyway; it's hardly an astonishing talent. It was probably pretty easy to clear the puck when Stevens, Daneyko et al were riding the forecheckers into the sideboards (aka interference).

As this blog has illustrated, Brodeur's first approx 10 yrs: some good, some avg, a couple of below avg years.... since the lockout: above avg years. Plays a lot of minutes for a team that has mainly been otherwise quite good. Therefore lots of career wins. But he's still probably the most overrated player in the NHL today. He's rarely if ever been the best goalie in the league in any particular year... not much of a legend, in that respect.

Bruce said...

Anonymous2, is that you? It’s hard to tell you Anonymice apart. You raise some interesting objections which I will address one by one.

Yeah, Brodeur has been better since the lockout. Not sure why.

I would say the imporvement you have noted (presumably Sv%) is because his job description has changed, the rules have changed, his team is considerably weaker, but he's a great enough goalie to adapt and continue to be successful.

If he had played at this level the previous 10+ yrs he might deserve more of the accolades he receives.

Yeah, all he did in those 10 years was win. You make the focus on various shot-focussed statistics, when if you ask the goalies themselves, their first focus is on winning. Sure, the best thing they can do to influence that is to stop the puck, but if the goalie plays on a good team that is expected to win and they do win, that is a goalie who is doing his job.

A goalie's main (99.5%?) function is stop shots.

99.5 % ???!!! Sorry, not buying. The goalie has major responsibilities back there above and beyond stopping shots.

If you bring up any other skill (e.g. clearing the puck/puckhandling) you better have some stats to illustrate why this is so important.

I better have stats or you’ll … ? Of course I don’t have stats, they don’t keep such stats, never have. Now if I could, I would travel through time and restructure the method the league keeps RTSS, but barring that I will simply use my eyes to watch the game and my brain to interpret what’s going on.

Much of the game isn’t quantified. Do you really think a defensive defenceman’s game is really summarized in hits and blocked shots, or maybe he does other stuff that’s important like move the puck out of the zone. So tell me … how many breakout passes did Nik Lidstrom make this year? How many per game? You “better” have stats or you’ll never convince me he’s good at it.

Even if Brodeur were somehow 10 times better than any other goalie at puckhandling (very doubtful)

Nope, he’s not 10 times better, he’s just Nik Lidstrom. He collects the puck, he reads the forecheck, he moves the puck, he rarely makes mistakes. There are others good at this skill, notably Turco who is more aggressively offensive minded than Brodeur, but when it comes to making the solid, safe play, Brodeur has been elite ever since he broke in. And that plays a huge role in the system Jersey has been able to deploy.

, I doubt this would make much of a difference in terms of goals saved.

On this point we will have to disagree. I think the difference is probably pretty significant, but immeasurable.

Let’s put it this way: if I were to list the top goaltenders over the course of Brodeur’s career, the short list would include Hasek, Roy, Belfour, Joseph, Luongo, Giguere, Turco. Of that list three jump off the page as exceptional puckhandlers, namely Belfour, Turco and Brodeur himself. Well guess what, those three have faced an average of 24-25 shots per game over the course of their careers, while the other five are all in the range of 28-32 shots/60. Do you think that’s just a coincidence of what team they play on, or does their good puckhandling help their team play the way it does? I’m not saying the difference is 10-20% as those averages suggest, but it sure wouldn’t surprise me if it was 5%.

And most NHL goalies are half-decent to very good at puck-clearing anyway; it's hardly an astonishing talent.

The good ones make it look easy. The range of skill is fairly wide, however; some guys handle the puck dozens of times per game, and get it to the defence with time and space; others are more inclined to just leave it for the D to handle under pressure.

It was probably pretty easy to clear the puck when Stevens, Daneyko et al were riding the forecheckers into the sideboards (aka interference).]

You are right that interference used to play a bigger role, and that rule was changed at the same time as the trapezoid was adopted. Both would affect the goalies’ range. However, I will turn your point right around and suggest it was probably easy for them to run interference when they didn’t have to worry about the puck because they trusted their goalie to handle it.

As this blog has illustrated, Brodeur's first approx 10 yrs: some good, some avg, a couple of below avg years.... since the lockout: above avg years.

Depends what numbers you look at. If you drink the Koolaid that it’s all about Sv%, Brodeur’s career Sv% is .913; the league-wide Sv% during those 14 years has been .902. When Jersey had a particularly strong offensive team was exactly the time Brodeur’s Sv% normalized, still average or a little above. But great GAA and league-leading wins totals year after year. What more can a goalie do on a strong team but keep winning? (See: Hasek and Osgood this year, no fantastic Sv% but they won the Jennings Trophy and the President’s Trophy.)

Plays a lot of minutes for a team that has mainly been otherwise quite good. Therefore lots of career wins. But he's still probably the most overrated player in the NHL today. He's rarely if ever been the best goalie in the league in any particular year... not much of a legend, in that respect.

3 Vezinas in the last four years would suggest that the GMs think otherwise about who’s the best goalie. But what do they know?

The Contrarian Goaltender said...

What do the GMs know? Sorry if I'm not convinced that GMs are goalie experts. The Vezina Trophy is basically turning into an award for the most games played, and meanwhile every season multiple goalies like Kiprusoff or Bryzgalov or Huet are either coming out of nowhere or are getting traded for nothing and becoming stars.

I just wanted to weigh in on the puckhandling debate. The debate is all subjective, of course, since as you point out it is difficult to quantify. I just don't see much margin in puckhandling, based on my viewing and of course my practical experience. I think that goalie puckhandling plays are generally small positive events, whereas turnovers are very negative events that outbalance a large number of good plays. So goalies that rarely play the puck don't add much but don't give much away, and goalies that play the puck add a lot of small positive gains but give a lot back on the corresponding increased rate of turnovers. This view is partially supported by what I found when I looked at giveaway rates for Brodeur and other goalies in this post.

The thing that struck me when looking at your list of the top recent goalies was that two of them jump off the list as exceptionally awful puckhandlers, Hasek and Roy, yet they are still easily the two best on the list. If there is anything approaching the margin you suggest in goalie puckhandling, then how did Hasek and Roy manage to be so successful?

those three have faced an average of 24-25 shots per game over the course of their careers, while the other five are all in the range of 28-32 shots/60. Do you think that’s just a coincidence of what team they play on?

In a word, yes. Brodeur and Turco are one-team guys, but their backups didn't face many shots either. Take Belfour, though, and look at his shots against by team: Chicago 24.6, San Jose 29.4, Dallas 24.3, Toronto 28.4, Florida 28.3. Better defence=fewer shots, seems pretty clear cut.

Finally, I've seen you try to make save percentage arguments for Brodeur before, Bruce, but I'm afraid they are never very convincing. I think you are understating the era effect, first of all - the numbers I've seen give the league average save percentage over Brodeur's career at .905. Secondly, league average is not a particularly impressive benchmark for any good starting goalie, much less a supposed all-time great. And it certainly isn't impressive for someone who was playing on the best defensive team in the league, facing probably the lowest quality scoring chances against as well as the fewest opposing power play chances. In that scenario, league average should be just about the basic requirement to keep the job.

But to really understand the Dead Puck Era effects, you have to look at the other goalies. Here are two lists, the first the all-time save percentage leaders (source: Hockey Reference), and the second is the save percentage for the first 8 seasons of a number of goalies who began their career around the same time or after Brodeur (see this post).

Career:
1. Hasek .922
2. Luongo .919
3. Giguere .915
4. Kiprusoff .915
5. Legace .914
6. Vokoun .914
7. Brodeur .914
8. Turco .913
9. Fernandez .912
10. Biron .911
10. Nabokov .911

First 8 NHL Seasons:
Luongo .919
Fernandez .913
Vokoun .913
Brodeur .912
Dunham .912
Nabokov .911
Hebert .911
Biron .911
Kolzig .911
Theodore .910
Roloson .910

Again, Brodeur is a good goalie. But so is Turco, so is Giguere, so is Kiprusoff, so is Vokoun, and so on. And that is the point, that Brodeur is not as outstanding as he is usually made out to be.

Anonymous said...

Bruce, I didn't know goalie puckhandling was so important to a team's chance of winning! :)

[Although the NHL doesn't (yet) keep track of stats like that, I wonder if someone informally does out there on the 'Net.] But seriously, in any typical game how many times does a goalie make a really dumb play with the puck, especially one that has serious consequences? It rarely happens... it's so routine for a goalie to just stop the puck behind his net, or make a short pass. (I suspect the impact of Lidstrom's passes - as compared to an avg d'man - far outweighs Brodeur's supposed puckhandling wizardry - as compared to avg goalie.)

I think you are way, way overrating the impact of goalie puckhandling, & in particular Brodeur's supposed mastery over this fine art.

As for the 'all that matters is winning' argument, I guess that would make a guy like Dryden (who probably has an amazing win pct.) the best all time goalie. Although then what are we to make of Bunny Larocque, who probably had a great win pct with Mtl, & then a terrible one through the rest of his career (e.g. Toronto).

Yep, Brodeur has played a ton of minutes, has racked up big win #'s, & in his primary function (to simply stop the puck) he hasn't even been the best in the league during his prime years. Again, a good to very good goalie -- but an alltime great? Not really.

Somewhat of a comparison is Mike Gartner.... although if Gartner had a couple below-avg yrs (e.g. 50 pts as a top line winger in his era) & a few avg yrs (e.g. 65 pts as a top line winger in his era) that would be more accurate.

Maybe he's been playing better the last 3 yrs because his "home life" has been resolved somewhat...

Bruce said...

CG: Some interesting points. I think the discussion about the modern goalie’s responsibilities is more important than where any particular guy stands on an imaginary list. For now I will answer just one section of your latest rebuttal.

In a word, yes. Brodeur and Turco are one-team guys, but their backups didn't face many shots either.

I’ll propose several reasons for this. The team’s system is designed with the #1 guy in mind, when he gets a day off the team continues to play the same way, and his backup is encouraged to play the system as well. He may well have been chosen with these attributes in mind, if not he sure as hell will be trained through repetition in practice. I hardly think it’s a coincidence that Dallas has churned out a number of excellent puckhandling goalies – Turco, Ellis, Smith – in recent years; since Belfour replaced Andy Moog in 1997-98 the Dallas system has relied on the goalie handling the puck.

Take Belfour, though, and look at his shots against by team: Chicago 24.6, San Jose 29.4, Dallas 24.3, Toronto 28.4, Florida 28.3. Better defence=fewer shots, seems pretty clear cut.

Not so fast. The San Jose experience was an absolute blip of 13 games; Belfour’s career average was below 25 shots per 60 until he landed in Toronto, whose gawdawful defence would overwhelm any goalie. In the context of this discussion, Belfour’s experience in Florida is particularly illuminating, a single season sandwiched between the incumbencies of Luongo and Vokoun, two excellent modern goaltenders not noted for their great puck and/or rebound control. It might be interesting to compare their individual performances over Florida’s three post-lockout seasons. Let’s stir the statistical ashes and see if we can find a spark of insight.

2005-06: Luongo: 34.7 SA/60; .914 Sv%;
2.97 GAA; 35-30-9 W-L-L, .534

2006-07: Belfour: 28.3 SA/60; .902 Sv%;
2.77 GAA; 27-17-10 W-L-L, .593

2007-08: Vokoun: 30.3 SA/60; .919 Sv%;
2.68 GAA; 30-29-8 W-L-L, .507


The Panthers were coached by Jacques Martin throughout and posted 85, 86, and 85 points with the three different starters. Belfour posted the worst Sv% of the three by a significant margin, but the team allowed far fewer shots so his GAA was nicely comparable and his actual winning record was significantly superior to his predecessor or his successor. But even after a pretty decent season he was sent packing, no doubt in part because he didn’t have a great save percentage, the stat of choice among the cognoscenti of the cord cottage.

Not to mention much superior to his backups, who posted a net 8-14-6 W-L-L mark, with an average of 30 SA/60. That’s a significant number of games both with and without Belfour in the net, so for arguments sake let’s postulate that Belfour’s puckhandling, rebound control and so forth was actually responsible for that reduction of 2 shots per game. With a .900 Sv% that equates to 1 GA every 5 games and a reduction of 0.20 off his GAA, without having any impact whatsoever on his Sv%.

That’s just one example, and I’ll search for more. The “problem” of Brodeur and Turco being one-team goalies (cuz they’re indispensable?) makes it hard to find comparables.

Anonymous said...

I haven't looked at it for awhile, but there is/was a common trend of the backup goalie(s) facing fewer shots/60... likely because the team will choose to use their lesser goalie(s) against the worst teams, when they have to use the lesser goalie(s) at all.

When comparing the Florida goalies over the years, & considering shots against/60, what about times shorthanded? Couldn't this vary significantly from year to year, & have an impact on shots against?

The Contrarian Goaltender said...

Anonymous, Michel "Bunny" Larocque is one of the classic cases for the "goalie is as good as the team" argument. In Montreal he was 144-48-31, 2.83. Outside of Montreal, 16-41-14, 4.87.

Bruce, interesting points. I don't think single seasonal examples are particularly meaningful here. I looked at Belfour's career and there isn't much evidence that he is better at shot prevention than his backups. Year-by-year average of shots against for Belfour: 26.9. Year-by-year average of shots against for his backups: 27.3.

I imagine that is probably typical. I looked at two other guys, Hasek and Luongo. Hasek did better than his backups, 27.8 to 28.8. Luongo did considerably worse, 32.2 to 30.2.

So, do some goalies face more shots than others? I think in most cases the effect is marginal, but there are guys who seem to be special cases, like for example Luongo. Now we have to figure out what is driving that. Is it puckhandling? Is it rebound control? What about team style of play, do they tighten up with the backup in net? If you were a bad team that struggled to score goals, wouldn't you take more chances with Luongo in net than with, say, Trevor Kidd? Are the backups only playing the weak teams? These are all factors blurring the analysis.

One other thing is that we don't really know who is a good puckhandling goalie. Is it better to only make 100% plays and never give it away? We think Brodeur is good, but do we really know that? What is the value of a good play compared to a turnover? Some people might say Rick DiPietro is good, but when I looked at the giveaway numbers DiPietro's were so far off the charts that I think he could be the worst puckhandling goalie in the league, no matter how far he can shoot a puck.

When I did my study comparing goalies to their backups, I used GAA, which I chose partially because it would capture anything that a goalie was doing to decrease shots against. Hasek still ranked #1, and I didn't particularly notice puckhandling goalies ranking any better than non-puckhandlers.

Maybe we should set up a control group. Name 4-5 good puckhandling goalies, and then 4-5 who aren't so good. I can look at shot differentials and see if there is any noticeable difference. If I had to bet, I'd guess there isn't much of a difference, and that the majority of any shot discrepancies can be explained by rebound control. But it would be interesting to find out.

Bruce said...

We think Brodeur is good, but do we really know that?

CG: Judging from your resistance to his glittering statistical record, you're going to be hard to convince on something as ethereal as puckhandling and flow of play.

What is the value of a good play compared to a turnover?

Variable in both cases. A simple teeing up of the puck behind the net is less valuable than a tape to tape breakout pass. Similarly, a turnover into an open corner is less damaging than the one that leads to an open netter.

I happened to see live New Jersey's game here in Edmonton this past Dec. 21, Brodeur was charged with two turnovers and I remember both of them. He came out to handle the puck, his passing options were covered, so he just softly dumped the puck into an open corner and got back in the crease. Both times it was the best play available, even though he got charged with a negative for making the correct decision. Both times the "turnover" -- where Brodeur collected the puck first, then "gave" it back in a non-threatening position -- amounted to nothing.

I suspect if goalie giveaways were tallied as a ratio of total touches, it would normalize in a hurry. DiPietro is crazy aggressive, but he does make a ton of real good plays, not just "small positives". And he's getting better at it.

When I did my study comparing goalies to their backups, I used GAA, which I chose partially because it would capture anything that a goalie was doing to decrease shots against.

Not sure I understand your reasoning here. I thought Sv% and SQN Sv% were the kings of the statistical hill? Please explain how using GAA "captures" an effect on shots against. (My own preference would be to compare the whole package of statistics.)

I don't think the backups are that good a test case. It's misleading to compare a 70-game starter to a guy who plays a handful of games that could range from "protected" home games against weak opponents to cleaning up the mess during blowouts. Small number statistics.

If I had to bet, I'd guess there isn't much of a difference, and that the majority of any shot discrepancies can be explained by rebound control.

Well, a big part of my position in the current discussion is that the goalie has many ways of contributing to shots against, of which puckhandling is just one area. My larger point is that no one stat, even a good one like SQN Sv%, is going to account for all that the goalie does back there.

The list includes:
-- communication with the defence
-- puck retrieval
-- teeing up the puck where the defenceman can move it safely
-- clean pass to the open man
-- direct clearance of the zone
-- rebound prevention
-- rebound control
-- housekeeping of loose pucks around the crease and behind the cage
-- interception of goal mouth passes
-- knowing when to freeze the puck
-- knowing when to not freeze the puck but maintain possession by keeping the play alive
-- etc.

All of these details pale in comparison to the primary job of stopping the puck, but they are nonetheless important.

there are guys who seem to be special cases, like for example Luongo. Now we have to figure out what is driving that. Is it puckhandling? Is it rebound control?

I'd say some of both. Luongo is a pure stopper, and he's got the impressive save percentages to prove it. Ditto Hasek, Roy, Giguere, Kiprusoff. Whereas Brodeur, Turco, Belfour, diPietro are more proactive, more "complete hockey players" if you will. Of that second group, Brodeur stands out to my eyes as an extremely efficient netminder, who is above- to way-above-average in all of those little categories. Add 'em all up and maybe it reduces shots against by 2 or 3 a game? Just an educated guess, but unlike Anonymous2, I'm convinced it's significant.

Whereas Luongo is probably the best pure stopper, but sometimes he has to stop two or three in succession. It's not necessarily a bad thing when you makes the stop(s) as frequently as he does; like any other position there's different ways of getting the job done, but that's the very reason why there's no one stat which adequately compares the different styles of goalies.

The Contrarian Goaltender said...

CG: Judging from your resistance to his glittering statistical record, you're going to be hard to convince on something as ethereal as puckhandling and flow of play.

My comment was questioning the certainty of our ability to evaluate rather than Brodeur's skill. I think Brodeur is a very good puckhandler, for the same reasons you state. But how much does he contribute to his team compared to a goalie who is just average with his stick? I'm still not convinced it is very much.

I suspect if goalie giveaways were tallied as a ratio of total touches, it would normalize in a hurry.

I'm sure it would normalize to some degree. More touches = more faults. I think the only way to really break puckhandling effects down would be film study.

Not sure I understand your reasoning here. I thought Sv% and SQN Sv% were the kings of the statistical hill? Please explain how using GAA "captures" an effect on shots against. (My own preference would be to compare the whole package of statistics.)

If a goalie can prevent shots, then it will impact his GAA, since fewer shots against means fewer goals allowed. It would not affect his save percentage, unless the shots that he prevented were either much more dangerous or much less dangerous than normal. Note that I don't necessarily believe this to be a significant effect, but if there was one there I think it makes more sense to look for the effects in GAA. Just like you did when comparing Belfour to Luongo in your earlier comment.

I don't think the backups are that good a test case. It's misleading to compare a 70-game starter to a guy who plays a handful of games that could range from "protected" home games against weak opponents to cleaning up the mess during blowouts. Small number statistics.

Over one season, of course it is small number statistics. But what about over the course of a 10 or 15 year career? If we have, say, a 150 game sample of how somebody's teammates performed, doesn't that become significant? No one statistic will tell you everything, but I think it can be useful as evidence in the overall analysis.

Brodeur stands out to my eyes as an extremely efficient netminder, who is above- to way-above-average in all of those little categories. Add 'em all up and maybe it reduces shots against by 2 or 3 a game?

There are many reasons why I focus on Brodeur, and one of them is what you just described here, that he is rated at the top in basically every category of intangibles there are. My question all along has been, so where's the effect? I'm not completely ruling out the possibility of goalies being able to prevent shots or contribute to the team in other ways than mere puckstopping, but unless something prevents goals then it isn't really valuable. And if it prevents goals, then it should show up or be measurable in one way or another.

Bruce said...

My question all along has been, so where's the effect? I'm not completely ruling out the possibility of goalies being able to prevent shots or contribute to the team in other ways than mere puckstopping, but unless something prevents goals then it isn't really valuable. And if it prevents goals, then it should show up or be measurable in one way or another.

I would suggest the effect IS fewer shots against. And as you say:
If a goalie can prevent shots, then it will impact his GAA, since fewer shots against means fewer goals allowed.

In the case of Brodeur, we typically have:
1) well below the league average in shots against, and
2) good to very good save percentage, ergo
3) excellent goals-againts average, which combined with
4) outstanding durabilty, leads to
5) league-leading wins totals

Now we can debate 'til the cows come home how much of effect 1) is Brodeur's doing, how much is the effect of the Devils' skaters, and how much of it is the system in which he is an important cog in the team. It's impossible to measure accurately, essentially we're looking for residuals, an island of signal in a sea of noise.

But without quantifying exact numbers, I would suggest fewer shots against, fewer goals against, and more wins are certainly "effects". Brodeur's career-long record demonstrates all of them on an annual basis.

2) and 4), save percentage and durability, are more directly attributable to the man himself. In the former he is good to excellent, among the very best in the league as you have outlined above; in the latter he is without peer for sustained longevity at a high level of play.

Bruce said...

It would not affect his save percentage, unless the shots that he prevented were either much more dangerous or much less dangerous than normal.

... or the shots that he created were more dangerous than normal. Again I have no numbers, so let's use logic instead. e.g. Dom Hasek, a guy you cited as a terrible puckhandler. Hasek takes lots of chances, especially when he's bored, going after pucks in the corner or aggressively charging at loose pucks squirting into the zone with an opponent in pursuit. Let's be generous and say that 9 times out of 10 he prevents a shot from happening at all by getting to the puck first or wiping out the guy who might have had a breakaway. But that 10th time there might be an extremely dangerous opportunity, a shot at an open net or with a defenceman or Hasek himself scrambling to make an emergency save. If the shot happens to be on goal (also not a given in a chaotic play) it has a much-better-than-usual chance of going in. Meanwhile, the goalie has been credited with zero saves on all the plays he has made safely. So the effects of dangerous wandering and bad puckhandling "should" have a negative effect on Sv%, moreso than on GAA.

unless something prevents goals then it isn't really valuable

How about if it contributes to goals for? I think of Saturday's game in Detroit where Marty Turco started the play for both Dallas goals and also set up one of their best scoring chances with a breakaway pass. The aggressively offence-minded goalie is a rare breed (Turco, DiPietro, NOT Brodeur), but the long bomb to the far blueline during a shift change on the PP is a weapon for a number of goalies. I'd like to see a little more of it at even strength, the goalie negating the icing and starting the counter attack with a long pass. However the no-change icing rule is advantageous enough that most (all?) coaches would prefer the whistle, the faceoff and the choice of match-up to the long bomb pass against a surprised and scrambling opponent. Since both rule changes -- obsolescence of the red line and no-change icings -- came in at the same time there's no way to know how the game might have evolved differently.

Anonymous said...

If Brodeur supposedly reduces the # of shots against due to his supposed stickhandling mastery, then it should be simple enough to weight his reduced-shots-against & apply it to his save pct & then compare this product to the product of an 'average' stickhandling goalie & his non-reduced-shots-against who has a higher save pct.

And yes, M.B. has had good save-pct. yrs since the lockout, but was less impressive the previous 10 or so years.

Bruce said...

I agree, Anonymous, that the calculation should be simple. The tricky part is to quantify the # of shots reduced by Brodeur's "supposed stickhandling mastery", not to mention all the other things I cited above that affect shots like rebound control etc. Such effects can only be measured imperfectly at best, as we have discussed above.

CG lists the top career Sv% leaders as follows:

Career:
1. Hasek .922
2. Luongo .919
3. Giguere .915
4. Kiprusoff .915
5. Legace .914
6. Vokoun .914
7. Brodeur .914
8. Turco .913
9. Fernandez .912
10. Biron .911
10. Nabokov .911

Most of us can agree that Brodeur is pretty darn good at most of those intangibles and immeasurables that affect flow of play. For argument's sake, let's estimate that all those factors have reduced his shots against by 2 per 60 minutes for his entire career. Adjust his save percentage for these two unrecorded "saves", and it's .920, right up there with Hasek and Luongo.

But wait! as CG notes above, Luongo's record suggests that he faces more shots than he "should". He gives out more rebounds, amnd is not a noted puckhandler. Let's makes an adjustment for him of -2 shots a game. Now his effective Sv% drops from .919 to .914, and the position of the two on the list is reversed.

NOTE: I am not saying that the effect is this great, or this small for that matter. I don't know what it is. I am merely doing the simple part, the math, based on a highly speculative guesstimate. But in a category where the difference between goalies is rated in thousandths (tenths of a percent), a small impact on the flow of play goes a long way.

Anonymous said...

Let’s imagine that Brodeur’s stickhandling wizardry reduces the # of shots faced per game from 30 to 28 (a reduction of 7%)… debatable that he would have that much impact on a game, but… if the save pct. leader was at 92.0% & Brodeur was 91.5%, over 70 games Brodeur would save 1.4 more goals - negligible. If the leader had 92.0% & Brodeur had 91.0%, over 70 games Brodeur would allow 8.4 more goals – a few wins (& losses) here & there. If Brodeur saved 90.5%, over 70 games he would allow 18.2 more goals – very significant.

Aside from the last 3 yrs, when Brodeur has seemed to have stepped up his play for some reason (above-avg save pct.), he often had mediocre save pct’s as compared to the elite. So even if he somehow is reducing the number of shots against (as compared to other goalies; verifiable?) he still would be costing his team wins due to his save pct. Perhaps NJ would have won more Cups if they had a superior save pct. goalie who unfortunately didn’t prevent as many shots against because he was not the dazzling stickhandler & passer that Brodeur apparently is.

As for rebound control, I have yet to see any stats indicating which goalie leads in this category. Just as with the shots-against theory, just more wild guesses.

The Contrarian Goaltender said...

Anonymous: Other hockey bloggers have done some fine work on the subject of rebounds, try searching around. Rebounds are a lot easier to quantify than puckhandling - you can figure them out from the game logs. If I was good at stripping data from the gamesheets that would be one of the first things I would focus on. However, other bloggers have done some fine work on the subject. Check out this post at the Forechecker, or this one at Mc79hockey. In the last one, Tyler measured Brodeur in 2006 as giving up .038 rebounds per shot while Luongo gave up .1, which would equate to about 2 extra shots per game against Luongo. This year, however, the Forechecker has Vancouver 5th in the league in rebounds against, while New Jersey is just 18th. Florida is last in the league, which probably explains a bit why Luongo's numbers were so bad as a Panther and have improved as a Canuck.

This illustrates the problem that only rebounds leading to a shot are counted. A good defence will clear most of the rebounds away and make the goalie look better. Dirk at the Forechecker tends to assign blame for rebounds to the defence, while Tyler at Mc79hockey tends to assign it to the goalie. The truth is probably somewhere in the middle.

The other issue is that goalies facing low quality shots against have a much easier job controlling rebounds than someone facing more difficult shots. This is probably a major reason why New Jersey ranks substantially lower in rebounds allowed in 2007-08 compared to prior years.

The Contrarian Goaltender said...

In the case of Brodeur, we typically have:
1) well below the league average in shots against, and
2) good to very good save percentage, ergo
3) excellent goals-againts average, which combined with
4) outstanding durabilty, leads to
5) league-leading wins totals


But without quantifying exact numbers, I would suggest fewer shots against, fewer goals against, and more wins are certainly "effects". Brodeur's career-long record demonstrates all of them on an annual basis.

Sure, fewer shots against, fewer goals against, and more wins are "effects" of something. The difficulty is whether they are effects of goaltending, or team defensive play, or offensive puck possession, or something else. Especially when most half-decent goalies on very good teams will show points 1, 2 and 3 on your list.

Take the example of someone like Chris Osgood, who we already established was a pretty average goalie. Let's look at Osgood and Brodeur, who are roughly the same age, looking at their formative goalie years side by side, at a time when they were both playing on successful teams (ages 22-27 seasons). Which one is which?

Goalie A: .632 win %, 2.17 GAA, .913 save %, 24.9 shots/gm

Goalie B: .654 win %, 2.30 GAA, .911 save %, 25.7 shots/gm

Not much between them, even though Goalie A is the future first-ballot Hall of Famer and master of every soft goaltending skill imaginable, while Goalie B is just a solid starter on a great team, not particularly noted for any of those things. The clear difference between them is of course durability and longevity, where Brodeur excels. Opinions of course vary on how much that should contribute to the legacy and greatness of a goaltender.

... or the shots that he created were more dangerous than normal.

Good point, and I agree with this one. I have repeatedly referred to the cost of turnovers because I think that negative events (like turnovers and rebounds) are much more impactful than positive ones (breakout passes, pokechecks, etc.). Goalies can create good scoring chances for the opposition through their own mistakes, resulting in more goals against, and that will show up in their numbers. This fits into my philosophy that save percentage captures nearly everything that is important about a goalie. If Brodeur is far better than everyone else at puckhandling and rebound control, why has he usually been in the middle of the pack of decent goalies in terms of save percentage? If there is a significant effect from these things that must mean Brodeur is a weaker puckstopper than the other guys are, and that is not even taking into account the fewer power plays and easier shots that he faced.

How about if it contributes to goals for?

Turco was indeed impressive against the Red Wings, and I think everyone would agree that assists are valuable plays. But they are still quite rare events. For example, over his career Brodeur has faced about 1,000 shots for every assist, and Turco's ratio is about 700 to 1. Like any skill in hockey, goalie puckhandling is good and desirable and can at times contribute to winning games, the debate is over how valuable it is.

Bruce said...
This comment has been removed by the author.
Bruce said...

if the save pct. leader was at 92.0% & Brodeur was 91.5%, over 70 games Brodeur would save 1.4 more goals - negligible. If the leader had 92.0% & Brodeur had 91.0%, over 70 games Brodeur would allow 8.4 more goals – a few wins (& losses) here & there. If Brodeur saved 90.5%, over 70 games he would allow 18.2 more goals – very significant.

Anonymous: An interesting point. After some initial confusion about your math, I agree with it. A reduction of 2 shots per game would be the equivalent of an effective Sv% ~.005 better than the official one. But rather than guesstimate differences on imaginary seasons, since we're considering career accomplisments I prefer to look at career rates. For one thing, Brodeur's Sv% has never been as low as .905 in any season that he has played as many as 50 games (which is to say, every one since 1995-96). He has consistently stayed within +/- .008 of his career rate of .914, a fairly narrow range that is among the most consistent of stoppers (Luongo, Giguere, Legace and Biron are right there around +/- .008; the rest have a wider range of Sv% over the years). And when you consider that Brodeur is just .001 out of third place on the all-time Sv% list, that "small" difference of .005 is pretty significant if you're all keen to rank these guys.

But since you raised the flag of the Sv% league leader, let’s go there.

Aside from the last 3 yrs, when Brodeur has seemed to have stepped up his play for some reason (above-avg save pct.), he often had mediocre save pct’s as compared to the elite.

You mean compared to the elite like Cristobal Huet? Niklas Backstrom? Dan Ellis? Because if Sv% is the only stat that matters, then those season leaders have been the best goaltenders in the NHL over the three post-lockout seasons. Or maybe you should call them half-seasons, since those guys played 36, 41, and 44 GP respectively. But they sure did lead the league in Save Percentage. When it came to actual saves, not to mention other contributions such as wins, they were a long way down the list. Sure, those guys probably had pretty good years, but I’d rather put my stock in the guys who are among the leaders every year rather than getting caught up in who might bubble up to the top. And when I say “among the leaders”, I include Sv% but I also consider GP, GAA, shutouts and especially Wins and points percentage. All of them are imperfect, and all of them need to be put in context of team. The good goalie is a team player first and foremost.

As an indicator of who is really the best goalie, in recent years I’d put Sv% somewhere behind shutouts – which frankly astonishes me, but compare these two lists of league leaders since 2000:

Season --- Shutouts | Sv%
-------------------------------------
2000-01 – Hasek | Turco
2001-02 – Roy | Theodore
2002-03 – Brodeur | Turco
2003-04 – Brodeur | Roloson
2005-06 – Kiprusoff | Huet
2006-07 – Brodeur | Backstrom
2007-08 – Lundqvist | Ellis


The shutout list 2000-07 includes 5 of the past 6 Vezina winners, and 6 of the last 6 first All-Stars (current season pending, of course, but I like Lundqvist’s chances a lot better than I do Ellis’s). Of the Sv% leaders, only Theodore won the Vezina, and he wasn’t even a first All-Star that rather astonishing season; the shutout leader, Roy, was. The other five years the two groups of voters -- GMs and PHWA -- agreed as to who the best goalie was, and it wasn't the Sv% leader.

No doubt you’ll reply that shutouts merely impress voters, but I’ll counter by saying forget about their votes and cast your own: just look at those two lists and tell me which group is better.

I’ll add that my tongue is partly in my cheek, it’s probably a cyclical thing, in his six consecutive Vezina winning seasons Dominik Hasek led the league in Sv% every year, but led or co-led in shutouts in “just” 3 of those years. The voters (and I) were able to agree who was the Dominant goalie in those days too.

My greater point is there are lots of aspects to goaltending, and no one stat captures the black art perfectly. That’s why I prefer to look at the whole bundle of stats rather than fixating on one particular column. It's not that Sv% isn't important, it's just not all-important.

Anonymous said...

There's no way that Wins, Pt Pct, GP & even shutouts & GAA are as important as Save Pct. when evaluating a goalie. All other things being equal (especially number of shots faced), I want the goalie who will stop the highest % of shots. Simple.

Give a good (but not necessarily "great") goalie 70+ GP playing for a great defensive team, & of course he'll have a lot of wins & good GAA & shutouts. Even if his save pct. is not outstanding. Ergo, Brodeur.

Have the save pct. leaders you listed (Huet, Backstrom, Ellis) shown themselves to be consistently at the top over many years? No, not yet. Maybe they will, maybe they won't.

[BTW, I prefer to compare a goalie's save pct to the league avg, especially since it has varied so much over the decades. Brodeur's career 90.5% sounds really impressive when you look at the 80's goalies, for instance, when 90% would make you the best in the league.]

Bruce said...

There's no way that Wins, Pt Pct, GP & even shutouts & GAA are as important as Save Pct. when evaluating a goalie.

Ask a goalie if wins are important. Ask a coach if GP are important. If you'd rather have a guy with a .924 Sv% in 2229 minutes (Ellis) over a guy with a .920 Sv% in 4617 minutes (Brodeur) because Sv% is the only way you like to evaluate goalies, be my guest.

All other things being equal (especially number of shots faced), I want the goalie who will stop the highest % of shots. Simple.

It might be simple if all other things were equal, but they never are. Every team presents different challenges, plays a different style under a different coach, with different degrees of offensive and defensive support.

Give a good (but not necessarily "great") goalie 70+ GP playing for a great defensive team, & of course he'll have a lot of wins & good GAA & shutouts. Even if his save pct. is not outstanding. Ergo, Brodeur.

Of course he will. That's why there's so many of those guys out there.

I happen to think a great goalie is more than a passive beneficiary and actually contributes to a great defensive team, but obviously when you got Johnny Oduya and Mike Mottau on your backend and you got all of one guy up front with more than 55 points, any Gumper the Goalie should at least win his division. Hell, the bum shoulda won the Cup.

I prefer to compare a goalie's save pct to the league avg, especially since it has varied so much over the decades.

I did that, showed Brodeur's Sv% to be some .011 above league average, and CG said league average didn't mean much, compare him to the "elite" so I did that.

Brodeur's career 90.5%

... that would be 91.4%

sounds really impressive when you look at the 80's goalies

Do you see anywhere where I compared his Sv% to an 80's goalie?
Anywhere?

We'll just have to agree to disagree on this one, Anonymous. You clearly have no use for the guy and will call him down no matter what. I on the other hand am not a Jersey fan at all, don't even think they're all that good, but I respect Brodeur as one of the "top all-time goalies". So I don't think we will find much common ground.

Bruce said...

For example, over his career Brodeur has faced about 1,000 shots for every assist,

CG: I'm guessing you're getting that from the excellent new resource Hockey-reference.com, but their stats for regular season goalie points only date back to 1997 and should be highlighted as incomplete. For the record Brodeur is 1-31-32 in the regular season, or about 1 point every 750 shots, or 1for every 65 goals against (a more equitable comparison than shots since assists are scoring plays). In the playoffs his 1-8-9 equates to 1 point every 36 GA. I very much agree it's small number statistics, obviously it's not the goalie's primary function but scoring at the other end (in two passes or less) is the tip of the iceberg when it comes to moving the puck.

The Contrarian Goaltender said...

Bruce: Yes, I just quickly pulled that from Hockey Reference, good catch.

No doubt you’ll reply that shutouts merely impress voters, but I’ll counter by saying forget about their votes and cast your own: just look at those two lists and tell me which group is better.

The answer to the shutouts vs. save percentage question is that the shutouts group has a built-in selection bias. Only the better goalies play enough games to have enough chances to record enough shutouts to lead the league. The average number of games played by the shutout leaders in your list was 72, the average number of games played by the save percentage leaders was 48. There is obviously more randomness in a smaller sample.

So here's a simple solution: set the minimum games to 50. Here is the revised list of save percentage leaders:

2007-08: Giguere
2006-07: Brodeur
2005-06: Kiprusoff
2003-04: Luongo
2002-03: Turco
2001-02: Theodore
2000-01: Burke

That list looks pretty good to me.

Sample bias is the reason you can't compare rate stats to counting stats. If we were looking at the league leader in shutouts per game, instead of total shutouts, then we would also see more outliers, such as Pascal Leclaire this year, Cristobal Huet again, Jocelyn Thibault, Roman Cechmanek, etc.

Ask a goalie if wins are important.

Of course wins are important. Winning the game is the point of playing hockey. But that doesn't mean that wins are the best way to evaluate goaltenders. The primary way that a goalie helps win games is by stopping the puck, and the rest is simply decided by his teammates and opponents. So it makes way more sense to focus just on stopping the puck, i.e. the critical process, than wins, i.e. the result that is predominantly determined by external factors.

If you'd rather have a guy with a .924 Sv% in 2229 minutes (Ellis) over a guy with a .920 Sv% in 4617 minutes (Brodeur) because Sv% is the only way you like to evaluate goalies, be my guest.

I agree that taking Ellis over Brodeur based on this season's results is statistical folly. But there are many ways to make mistakes with numbers - there are some who make the opposite mistake by claiming Brodeur is better than Hasek because Marty has more career shutouts. I think if we are looking at save percentage over a large enough sample size, and we are aware of the team factors that are at work, then that will provide a pretty good picture of a goalie's play.

Bruce said...

First some unfinished business:

Take the example of someone like Chris Osgood, who we already established was a pretty average goalie.

I didn't say Osgood was an average goalie, in fact I compared him favourably to a Hall of Famer (Cheevers). I called him "another great player on a great team". He ranks fourth among active goalies and 15th all-time with 363 regular season victories, and fourth and 14th respectively with 57 playoffs wins. Tonight he broke Terry Sawchuk's Red Wing record with his 12th career playoff shutout, and became just the fourth goalie in NHL history to start the SCF with two shutouts (the last was, ahem, Marty Brodeur in 2003).

Let's look at Osgood and Brodeur, who are roughly the same age, looking at their formative goalie years side by side, at a time when they were both playing on successful teams (ages 22-27 seasons).

Nice dig, and a reasonable comparison. Judging from that Pts%, Osgood played on the more successful team. Despite that his Sv% and GAA were not quite as good. And of course he played about 100 fewer games.

The clear difference between them is of course durability and longevity, where Brodeur excels. Opinions of course vary on how much that should contribute to the legacy and greatness of a goaltender.

... or of any player. Some might rank Mario Lemieux above Gordie Howe based on peak value (which is debatable), but I'll go with Mister Hockey based on durability and longevity, not to mention consistency. At a certain point, such as Gordie's 20 consecutive years in the Top 5 scorers or Glenn Hall's 503 consecutive starts, durability must be recognized as an exceptional talent in its own right. Brodeur's 10 consecutive seasons of 70+ GP is such an accomplishment in my view ... exceptional enough that, when combined with across-the-board excellent rate stats (Sv%, GAA, Pts%), clearly sets him apart from the field and warrants consideration among the all-time elite of his position.

Bruce said...

CG: Nice recent post, with some good points to which I’ll respond.

The answer to the shutouts vs. save percentage question is that the shutouts group has a built-in selection bias. Only the better goalies play enough games to have enough chances to record enough shutouts to lead the league.

“Only the better goalies” indeed. Seems to me some of the best goalies play a hell of a lot. Their save percentage might dip a little with the big workload, but they still rack up some pretty important counting numbers: GP, Shutouts, Wins …

The average number of games played by the shutout leaders in your list was 72, the average number of games played by the save percentage leaders was 48. There is obviously more randomness in a smaller sample.

Obviously. Ergo Roloson, Huet, Backstrom, Ellis as our last four league-leaders … Which is “obviously” one of the reasons I don’t just accept Sv% as the be-all and end-all when evaluating netminders.

So here's a simple solution: set the minimum games to 50.

I agree with this, and thanks for doing the research to generate the below list. Alternately, the minimum could be set as a certain number of shots faced, a la plate appearances in baseball.

Here is the revised list of save percentage leaders:

2007-08: Giguere
2006-07: Brodeur
2005-06: Kiprusoff
2003-04: Luongo
2002-03: Turco
2001-02: Theodore
2000-01: Burke

That list looks pretty good to me.


… and to me. Give or take a Roy or a Hasek, almost as good as the shutout list. Hey lookit! there’s good ol’ Marty Brodeur!

Sample bias is the reason you can't compare rate stats to counting stats.

… and the reason you can’t just accept rate stats as the only relevant tool.

If we were looking at the league leader in shutouts per game, instead of total shutouts, then we would also see more outliers, such as Pascal Leclaire this year, Cristobal Huet again, Jocelyn Thibault, Roman Cechmanek, etc.

Say no more. I’m not making the case for shutouts particularly -- I keep saying no one stat covers the waterfront –- but you’ve hit on the reason why that category does a better job of who the real difference-maker goalies are. There's obviously less randomness in a larger sample.

Of course wins are important. Winning the game is the point of playing hockey. But that doesn't mean that wins are the best way to evaluate goaltenders.

It’s one way to evaluate them. If a goalie isn’t winning there may be reasons, but if he is, it’s a sure sign he’s doing the job.

The primary way that a goalie helps win games is by stopping the puck, and the rest is simply decided by his teammates and opponents. So it makes way more sense to focus just on stopping the puck, i.e. the critical process, than wins, i.e. the result that is predominantly determined by external factors.

Not sure what you mean by “predominantly”. The goalie is just 1 player out of 12 out there; I would say he has a disproportionately high effect on the game, but it’s not any 75% or 100% or whatever stupid number the “experts” spew. He’s the key man in his own zone, but certainly he has very little to do with what happens in 2 zones out of 3. OTOH, a superstar forward who plays 20 minutes a game is on the bench 2/3 of the time.

I agree that taking Ellis over Brodeur based on this season's results is statistical folly.

Thank you. Unlike our anonymous friend, you have shown you will listen to reason at least occasionally. :D In the current instance, one goalie is .015 above the league average, and the other one .011 better for ~twice as many shots. Not too hard to figure which one prevented more “expected goals”.

But there are many ways to make mistakes with numbers

You can say that again.

- there are some who make the opposite mistake by claiming Brodeur is better than Hasek because Marty has more career shutouts.

That would be a mistake. It’s a point in Marty’s favour, but that’s all. There’s lots o’ columns to consider and shutouts are just one.

I think if we are looking at save percentage over a large enough sample size, and we are aware of the team factors that are at work, then that will provide a pretty good picture of a goalie's play.

I agree with “pretty good”. Those team factors throw a huge wrinkle into the discussion, as they do when considering virtually any player but probably goalies the most. They include numbers of shots against (counted), quality of shots against (estimated at best), and offensive support (variable in Brodeur’s case, currently weak).

In the above discussion I have tried to make the case that as an important part of the team, the goalie himself plays a significant role in those team factors above and beyond stopping the puck. If you accept the premise that Brodeur’s play outside of facing shots is responsible for reducing opposition shots by a couple per game compared to the other guys above him, he effectively ranks behind only Hasek on the career Sv% list. Even if the difference is as little as one shot per 60, that puts him in third place, all-time, in the category he’s supposedly weakest.

The Contrarian Goaltender said...

Judging from that Pts%, Osgood played on the more successful team. Despite that his Sv% and GAA were not quite as good. And of course he played about 100 fewer games.

Of course they weren't quite as good. It is Chris Osgood vs. Martin Brodeur. If I compared Dominik Hasek to Tom Barrasso, or Patrick Roy to John Vanbiesbrouck, who do you think would come out ahead? I don't think either of those comparisons would be as close as the Ozzie v. Marty one. The point is that fewer shots against and fewer goals against are often observed on great teams. Martin Brodeur played on mostly great teams. That could be a coincidence, or it could be a result. We have to try to dig deeper to find out which one it is, and until then you can post as many win totals and low shots against totals as you want, but it won't really prove anything.

Some might rank Mario Lemieux above Gordie Howe based on peak value (which is debatable), but I'll go with Mister Hockey based on durability and longevity, not to mention consistency.

I'd take Gordie Howe over Mario Lemieux too, because of both peak value and longevity. Howe's peak is very underrated - he had some dominant scoring years where he left everyone else in his dust. Another reason is that most of Lemieux's scoring dominance took place on the power play. Special teams scoring is valuable, of course, but I think the greater test of a player is how they perform at 5 on 5.

Hey lookit! there’s good ol’ Marty Brodeur!

Yes, there is. In his outstanding 15 year career, in which he repeatedly led the league in games played, wins, shutouts, and best goalie in the league votes by NHL general managers, Martin Brodeur managed to lead the league once in save percentage (50 game minimum), tied with the likes of Sean Burke, Bob Froese and Jose Theodore. Truly a historically significant achievement. And because I'm in a generous mood, I won't even get into how Luongo was actually the better puck stopper in 2006-07 as well, although you can read about it here if you are so inclined.

Seriously though, I have written about how Brodeur's play has improved post-lockout. I consider it highly unlikely that he was just playing at the same level all the time and something else changed - his save percentage splits are substantially different. He is very good now, but that doesn't mean he was very good all along. I am not going to change the title of this blog any time soon, but the reality is that I do rank him a lot higher now than I did three years ago.

If you accept the premise that Brodeur’s play outside of facing shots is responsible for reducing opposition shots by a couple per game compared to the other guys above him, he effectively ranks behind only Hasek on the career Sv% list. Even if the difference is as little as one shot per 60, that puts him in third place, all-time, in the category he’s supposedly weakest.

I don't accept the premise yet, as I'm still looking for evidence of those types of effects. But even if I did, you are putting way too much stock in the all-time save percentage list. Save percentage is an extremely era-dependent statistic. As such, the current list is largely a ranking of goalies based on age, i.e. youngest at the top, oldest at the bottom. This is what I have tried to show already by posting the example of the results for mid-'90s goalies in their first 8 seasons. There are exceptions, of course, but the guys at the top of the all-time list are the ones who played since 1995, and 1995-2004 was obviously a very different era than the mid 1980s. The other obvious point is that official save percentage only goes back to 1984, so we don't even have 25 years of data.

Take for example Patrick Roy, who played over 28,000 minutes in a high-scoring era. His career save percentage is .910. Nowadays that is just a bit above league average. That's why we see average goalies who played in the late 1990s/early 2000s ahead of him, like Martin Biron. If you normalize for era, Roy comes out way ahead of Brodeur and probably everyone else except Hasek.

You have pointed out some of the flaws in save percentage already, and this is another one - you can't compare across eras. League average save percentage has moved around a lot over the years, and will likely keep cycling up and down in the future.

But even if you could somehow prove that Brodeur ranked high in save percentage, so what? That's where he should be in the first place, because he played most of his career behind a dominant defence, not to mention he is supposedly such an all-time great. The team adjustment effect would probably be at least as large as your shot prevention adjustment, and it is going to be in the opposite direction, putting him where I think he should rank - a cut below the elite goalies (e.g. Hasek, Roy, Luongo), and right there with the other decent goalies of his generation, the Belfour/Cujo/Giguere types. And in the end, Brodeur will probably end up edging all those guys out because of his greater longevity and durability. That's still a great career for anybody to have, but one of the very best ever? Not in my view.