Thursday, June 28, 2007

Hasek vs. Brodeur: Why It's Not Even Close

In 1993-94 season Martin Brodeur and Dominik Hasek both became full-time starting goalies in the NHL, Brodeur because he was a rookie, and Hasek because he had managed to get out of the Czech Republic, and was finally given the chance at a starting job. For the next decade, until Hasek retired for the first time in 2002, they were both considered among the premier goalies in the National Hockey League. Which one was better? Let's see what the numbers say.

From 1993-94 until 2001-02, Dominik Hasek faced 1,060 more shots than Martin Brodeur, and gave up 135 fewer goals.

I had to check those numbers again because I thought I had made a mistake at first. It is sometimes easy to shrug off save percentages, since there doesn't look like that much of a difference between Hasek's .926 and Brodeur's .911, but the difference shows itself in the totals. To try to quantify the gap between Hasek and Brodeur, I looked for a goalie that faced about 1,000 fewer shots than Brodeur and gave up 130 more goals in the same time period. There wasn't one, because no goalie that bad would get enough playing time to qualify. The two closest were Arturs Irbe (1870 fewer shots, 62 fewer goals against) and Jocelyn Thibault (1948 fewer shots, 77 fewer goals). Brodeur was much closer to guys like Irbe or Thibault than he was to Hasek in the 1990s. The Dominator was just on a completely different level.

Hasek was even more dominant in a team context. The 2002 Red Wings were great, but his Sabres teams were pretty average. Altogether, his teams won 343 out of 706 games from 1994-2002. The Devils were a consistently dominant team in front of Brodeur, with 380 wins in the same time period. They were not just better defensively but offensively as well, outscoring Hasek's teams by 116 goals.

Hasek's backups were 79-95-21, with a 2.96 GAA and a .900 save percentage. Brodeur's backups were 58-62-12 with a 2.63 GAA and a .900 save percentage, facing 3.4 fewer shots per game. Hasek's backups were more talented goalies, as most of them had been or became starters in the NHL: Grant Fuhr, Martin Biron, Dwayne Roloson, Steve Shields, Manny Legace. Brodeur had two decent backups, Mike Dunham and John Vanbiesbrouck, and they combined for a .911 save percentage. The mediocre Chris Terreri, backup for 5 of the seasons, was at .898, and the rest were minor-leaguers (.882).

What about the playoffs? Well, Brodeur certainly had more playoff opportunities because of the strength of his teams, playing 114 playoff games to Hasek's 90. His 67-47 record was also slightly better than the Dominator's 52-37. However, Brodeur's winning percentage of .588 was below New Jersey's regular season average of .615, while Hasek's playoff win mark of .584 was much better than his team's seasonal rate of .559, indicating that the Dominator carried his team in the postseason. In total, Brodeur faced 260 more shots and gave up 33 more goals than Hasek, which meant Hasek had a better save percentage, .927 to .922. Hasek only had one Cup win to Brodeur's two, but from 1994 to 2002 Hasek was the better playoff goalie.

So Brodeur is almost totally outclassed by that comparison. However, it is not really fair to him since the analyzed period includes his early seasons and none of his later Vezina-winning years. To compare apples to apples, let's put the two of them side-by-side at a similar age and look at their records from the age of 29 to the age of 34 (which because of Hasek's late start are the only seasons we can use to compare the two as starting NHL goalies):

Age 29-34 seasons: Dominik Hasek faced 1,494 more shots than Martin Brodeur, and gave up 41 fewer goals.

So Hasek faced almost an extra season's worth of shots, and still gave up fewer goals. Brodeur did miss his age 32 season because of the lockout, so I guess hypothetically he might have played an entire season without giving up a goal, while using his puckhandling skills to create an additional 40 goals of offence. But anything less than that, and it has to be conceded that Hasek was better than Brodeur at a similar age.

The team context is even more skewed in this sample. Between Hasek's age 29 and 34 seasons, Buffalo went 211-180-69 for 491 points in 460 games. In the same career point for Brodeur, New Jersey went 225-124-61 for 511 points in just 410 games. Buffalo scored 2.9 goals per game and gave up 2.5, while New Jersey scored 2.7 and gave up 2.3. One striking difference was in penalty minutes. Buffalo averaged 22 PIM per game, while New Jersey averaged just 11, another advantage to Brodeur who likely faced about half as many power plays.

Hasek's backups went 40-55-13 with a 3.31 GAA and an .897 save percentage, while Brodeur's were 17-15-6 with a 2.21 GAA and a .910 save percentage. Again, this despite Hasek's backups being better goalies (Fuhr, Biron, Roloson et al), as Brodeur's teammates in this period were Scott Clemmensen, Corey Schwab, J.F. Damphousse, and, for one season, the 38-year old John Vanbiesbrouck. The shot totals reinforce New Jersey's defensive strength: Buffalo's backups faced 32 shots per game, while New Jersey's faced just 24. New Jersey was a much better team than Buffalo, meaning that even if Hasek and Brodeur had similar statistics, Hasek would have been the better goalie. The fact that Hasek outplayed Brodeur by a wide margin despite the team factors shows that he was a decisively better goalie.

Hasek was also better in the playoffs in this sample. Brodeur went 44-36 with a 2.00 GAA and a .919 save percentage, stats that were about average given the low-scoring era he played in. Hasek was 28-20 with a 2.01 GAA and a .933 save percentage. Again, Hasek's playoff winning percentage (.583) was higher than his team's during the regular season (.534), while Brodeur's Devils went from a .623 regular season rate to just .550 in the postseason.

Some of Hasek's performance statistics are almost mind-boggling. It is probably only the fact that he came to the NHL late that prevents him from being a near-unanimous choice as the best ever, or at least one of the top 2 or 3 in history. He was much better than Brodeur in the late '90s, and much better at a similar age. Frankly, I think it is astonishing that anyone who followed NHL hockey in the 1990s would rank Brodeur ahead of Hasek. Subjectively and objectively, the Dominator was in a league of his own.

Monday, June 25, 2007

Is It Harder To Face Fewer Shots?

There is a debate over whether save percentage is influenced by the number of shots faced. Some argue that the more shots a goalie faces, the more difficult it is to save them. This position was taken by Klein and Reif in their influential Hockey Compendium, and led to their Goaltender Perseverance stat that rewards goalies who face more shots per game. Others maintain the opposite, that goalies who face fewer shots have a more difficult task. This is because it is harder to stay focused, and their teammates tend to prevent the long, easy shots that other goalies face. This has often been used in support of Martin Brodeur, including in comments posted at this blog.

Neither of these viewpoints are correct. In an article on Puckerings.com, Ian Fyffe dismantled Klein and Reif's viewpoint and argued that Perseverance is meaningless. There is very little correlation between shots faced and save percentage, and the evidence is that shots faced are overwhelmingly a function of the team, rather than the goaltender. To further investigate this finding, I have looked at the top save percentage seasons of the save percentage era, which began in 1983 when the NHL began officially publishing shot totals. My findings supported Fyffe's, that save percentage appears to be independent of the number of shots faced, and that goalies can have outstanding save percentages both when facing many shots per game and when facing very few.

The average number of shots faced by goalies who had save percentages of .925 or better was 28.4 shots per game. The top 4 seasons were all by goalies facing fewer than 30 shots per game. In fact, the only goalies to face 30 or more shots and stop at least 92.5% of them in at least 50 games were Theodore, Luongo, and Hasek (3 times). Hasek is bringing up the average significantly; take him out, and the average shots against per game for the top goalies falls under 28 per game.

Several goalies made the list despite facing 26 shots or less, including Brodeur's .927 in 1996-97. Others were Kiprusoff (.933), Turco (.932), Cechmanek (.925), Tugnutt (.925), and Roy (.925).

That is just absolute save percentage, however. Save percentages have gone up and down with scoring levels in the NHL, which is why there were no 1980s goaltenders on that list. Perhaps a better way to look at it is through relative save percentage, which is save percentage divided by the league average save percentage for that season. Looking at the top 30 in relative save percentage (none of whom, of course, are Martin Brodeur), the goalies faced almost exactly a league average number of shots, and 10 of them faced 2 or more shots less than league average, including the top two (Bob Froese in 1986 and Dominik Hasek in 1994).

This list is not dominated with goalies that face a lot of shots. Of the top 30 in relative save percentage, only a few seasons fit the "Luongo" mold (lots of shots faced on a weak team): John Vanbiesbrouck on Florida in 1994, Bob Essensa on Winnipeg in 1992, Jose Theodore for Montreal in 2002, and Dominik Hasek several times on the Buffalo Sabres. Most of the goalies played on good teams.

The evidence suggests that save percentage is largely independent of the number of shots faced, and that shot quality is a more significant variable than the number of shots against. Facing a high or a low number of shots is therefore no excuse for a goaltender.

Monday, June 18, 2007

Did the Vezina Voters Disrespect Brodeur?

Martin Brodeur recently added a third Vezina to his trophy haul, adding another strong argument to those who defend his greatness. Others would argue that none of those Vezinas were deserving, so they should not be given heavy weighting in a career perspective.

I sympathize with the latter view, but the problem is that picking and choosing Vezina trophy winners can easily dissolve into blatant bias. For example, it is difficult to brush aside Brodeur's three Vezinas and then on the other hand, hold up Patrick Roy's three awards as proof of his greatness. In addition, goaltenders often get delayed recognition - trophies usually go to well-accomplished goaltenders, and younger goalies often lose out. This means that just because a goalie won Vezinas late in his career (like, for example, Brodeur), does not necessarily mean that he is undeserving of those honours, because he may have actually deserved some earlier awards. Many have claimed that Brodeur was ignored and overlooked early in his career, mostly because of the Devils' defensive system. When he won his first Vezina in 2003 it was widely seen as long overdue. To properly evaluate Brodeur, therefore, we need to see whether he was perhaps underappreciated to some degree in his younger days, which may balance out some of the possible overexposure he has received in recent years.

I have investigated the Vezina voting from 1994-2002, and conclude that claims that Brodeur was overlooked and discriminated against are revisionist nonsense. Brodeur got everything he deserved and much more than that as a Vezina candidate. He nearly always was ranked ahead of other goalies with almost identical statistics, even some of the best goalies in the league. He was just never deserving enough to win.

After winning the 1994 Calder Trophy in his rookie year, Brodeur was already recognizable enough to collect Vezina votes in 1994-95 with two seconds and a third place vote. Look at how eerily similar his stats were to Boston goalie Blaine Lacher, who received no votes:

Brodeur: 2.42, .902, 19-11-6, 3 SO, 24.7 shots/game
Lacher: 2.41, .902, 19-11-2, 4 SO, 24.5 shots/game

In 1995-96, Brodeur and the Devils failed to qualify for the playoffs. To be fair, the main reason for this was their lack of scoring, but not making the playoffs is often held against goalies even if they play on weak teams, and Brodeur played on the defending Stanley Cup champions. It was definitely held against Hasek, who did not win the Vezina that year despite leading the league with a .920 save percentage. This was because the Dominator was only credited for 22 wins, facing 35 shots per game on a weak Sabres team. As a result, Hasek finished 8th in Vezina voting. Brodeur was decent (.911 save percentage), but the only thing he led the league in was games played. Nevertheless, Brodeur received 4 firsts, 3 seconds, and 2 third place votes to finish fourth in the Vezina race. In the season-end All-Star Team voting, Brodeur had the second most first-place votes after Jim Carey, and narrowly missed out on a Second Team berth. Carey was a poor choice as a winner that year, and it is justifiable to claim that Brodeur should have beat him, but the best goalie in 1995-96 was Dominik Hasek, so it was not really an injustice to rank Brodeur third.

In 1996-97, Brodeur had a great season, possibly his best ever (.927, 37 wins). Unfortunately, even that wasn't enough to match Dominik Hasek. The Dominator had the same number of wins as Brodeur in the same number of games, and had a better save percentage (.930) while facing almost 7 extra shots per game. The mediocre Sabres won their division, and Hasek's brilliance was shown by him receiving 50 out of 54 first-place Hart Trophy votes. Brodeur finished a deserving second in the voting.

In 1997-98, Martin Brodeur and Ed Belfour had virtually identical seasons. The only difference was that Brodeur played in a few more games.

Brodeur: 1.89, .917, 43-17-8, 10 SO, 22.8 shots/gm
Belfour: 1.88, .916, 37-12-10, 9 SO, 22.4 shots/gm

The Vezina voters apparently saw a huge difference. Brodeur finished second to Hasek again in Vezina voting with 57 points including 2 first place votes. Belfour finished a distant fourth with only 10 points, with just one second place vote and seven thirds. This was repeated in the All-Star voting, where 47 out of 50 voters put Brodeur second on their ballots, while nearly a third of them left Belfour off their ballots entirely.

Martin Brodeur was obviously developing a great reputation, and that was reinforced the next year based on his voting results against the legendary Patrick Roy. Brodeur and Roy had identical GAAs, and although Brodeur had more wins, Roy had more shutouts and a much better save percentage. Roy received only one third place vote for the Vezina, while Brodeur was named on 11 ballots, including one first place vote.

Brodeur: 2.29, .906, 39-21-10, 4 SO, 24.5 shots/gm
Pat Roy: 2.29, .917, 32-19-8, 5 SO, 27.5 shots/gm

In 1999-00, Brodeur again beat Roy in the voting, despite posting similar numbers (2.24 GAA and .910 save % compared to Roy's 2.28 and .914) on a team that was better both offensively and defensively.

The best New Jersey Devils team of the era was probably the 2000-01 version. Playing for that juggernaut, Brodeur again piled up the wins, and the Vezina voters, as usual, took notice. Three other goalies, including Belfour again, had very similar stats.

Brodeur: 2.32, .906, 42-17-11, 9 SO, 26.9 shots/gm
Belfour: 2.34, .905, 35-20-7, 8 SO, 26.9 shots/gm
Dafoe: 2.39, .906, 22-14-7, 2 SO, 27.8 shots/gm
Irbe: 2.45, .908, 37-29-9, 6 SO, 29.0 shots/gm

Together, Belfour, Dafoe and Irbe combined for a mere two Vezina votes. Martin Brodeur got 7 first place votes and finished third in the vote totals, and in the All-Star voting he picked up 18 first-place votes and narrowly missed out on a Second Team nomination.

In the Olympic year of 2001-02, Brodeur did not have a very good year, and Jose Theodore took home the Vezina. Once again there were several close comparables to Brodeur, including Curtis Joseph and Dafoe, and Brodeur received more votes than any of them.

Perhaps the favouritism displayed towards Brodeur is best expressed by a comparison with Ed Belfour. Like New Jersey, Dallas played a strong defensive system, so the two goalies played in very similar environments. Between 1997-98 and 2000-01, both goalies faced 24 shots per game, and both got around 3 goals per game in support from the teams in front of them. As might be expected given these similarities, their performance statistics were very similar as well:

Goalie stats (1997 to 2001):
Brodeur: 284 GP, 167-75-37, 2.19, .910, .665 win %, 29 SO
Belfour: 247 GP, 139-68-33, 2.08, .914, .648 win %, 26 SO

If either of them has a slight performance edge it's probably Belfour, but the Vezina voters repeatedly sided with Brodeur, who received 124 voting points and 10 first place votes over those four seasons. Belfour had just 40 points and was ranked first just twice. In media voting for the seasonal All-Star Team, Brodeur again received way more recognition than Belfour, totalling 371 voting points compared to 123 for Eddie the Eagle. Over the four seasons, only one sportswriter ever ranked Belfour as the best in the league, while Brodeur was ranked first on 26 different ballots.

There may have been some observers who dismissed Martin Brodeur as a team creation in the mid- to late-1990s. If so, they were clearly in the minority. Brodeur got more votes than he deserved almost every year. A number of times there were other goalies with almost identical stats, and virtually every time they ended up ranking well behind Brodeur, even the best of his peers such as Roy, Belfour and Joseph.

Despite all that, Brodeur didn't win a Vezina and only twice finished as the runner-up in his first nine years in the league. The end-of-season awards evidence therefore shows that Martin Brodeur was far from dominant in the prime of his career, which makes it that much more difficult to make a compelling argument that he was the best goaltender of his generation, much less one of the greatest of all-time.

Friday, June 15, 2007

Martin Brodeur Wins His Most Deserving Vezina Ever

I would argue strongly that he did not deserve it this year either, but he certainly had a stronger case this year than in either of 2002-03 or 2003-04. The voting was pretty close as well (16 first-place votes for Brodeur, 14 for Roberto Luongo).

I guess Luongo, who should have 2 Vezina Trophies on his shelf by now, may just have to wait a little longer until the point where he starts getting awarded Vezinas just because of his past successes.

I'd still bet on him for 2007-08, though. I think a key point in this debate is that while Brodeur may have had a career year this year, Luongo certainly did not, despite all the wins and accolades and so on. All he did was play on a better team. Looking at this season in the context of their recent careers, Brodeur's season looks like an outlier and Luongo's season is pretty much the same thing that he has been doing since 2000:

2000-01: Brodeur .906, Luongo .920
2001-02: Brodeur .906, Luongo .915
2002-03: Brodeur .914, Luongo .918
2003-04: Brodeur .917, Luongo .931
2005-06: Brodeur .911, Luongo .914
2006-07: Brodeur .922, Luongo .921

When Goalie A has a career year, and Goalie B plays the same way he always has, and their performances are similar, then the logical conclusion to make is that Goalie B is the better goalie, Vezina or no Vezina.

Thursday, June 14, 2007

Vezina Trophy: Backup Goalies

Another way to compare goalies is to look at the performance of their backups. The backup goalies play behind the same defence, so if there are any strong team effects they often will come through in the statistics of the backup goalies. Vancouver's backup goalie is Dany Sabourin; New Jersey's is Scott Clemmensen. Neither of them played very many games, so their statistics are almost the very definition of a small sample size.

Here are their statistics:
Dany Sabourin: 2-4-1, 2.63 GAA, .906 save %
Scott Clemmensen: 1-1-2, 3.15 GAA, .889 save %

One interesting stat is the shots faced per game by the backups, which often gives us some indication of the strength of the defence. Both teams played more defensively with their backups in, giving up fewer shots against. Dany Sabourin faced 27.3 shots per 60 minutes, compared to Luongo's 29.0. Clemmensen faced 26.6 shots per game, which is about the same difference from Brodeur's 27.9. So that indicates that New Jersey is better at shot prevention than Vancouver; the difference is not just attributable to Brodeur's puckhandling or rebound control.

Backup goalies usually face an easier schedule, so that has to be taken into account. Here are the teams both faced:

Sabourin: EDM, ANA, NAS, CGY, BOS, BUF, ANA, PHO, PHO
Clemmensen: NAS, OTT, BOS, WAS, CAR, NYI

The schedules are roughly even. Sabourin faced 5 playoff teams in 9 games, including 4 elite teams. Clemmensen faced 4 playoff teams in 6 games, 2 that would be considered among the league's elite.

Taking their stats in common games only (vs Nashville and Boston), Clemmensen comes out well ahead, but that is a very small sample size.

Clemmensen: 124 minutes, 55 saves, 62 shots, .887
Sabourin: 112 minutes, 36 saves, 43 shots, .837

A single season for a backup goalie is a very small sample size. It is therefore perhaps beneficial to look at the goalies' careers for some more context.

Scott Clemmensen is 29 years old. In 2005-06, he went 3.35 and .881 in 13 games. He posted solid numbers in the AHL for the three years before that, peaking in 2004-05 with 2.81 and .916 in 46 games on a weak team. Clemmensen's AHL stats were very similar to those of Ari Ahonen, another Devils goaltending prospect who has never succeeded in breaking into the NHL.

Dany Sabourin is 27 years old. He had only 5 career games played in the NHL coming into this season. His 2005-06 AHL season looks excellent (2.26, .922), but the Wilkes-Barre Penguins were an outstanding 113 point team, and he actually had the worst save percentage of the 4 goalies on the team. However, in 2004-05 he put up similar numbers (2.22, .921) on a weaker team, outperforming current Pittsburgh starter Marc-Andre Fleury, who had a .901 mark in 54 games.

In summary, it is difficult to make any definite conclusions based on the backup goalies because of the small sample size. Dany Sabourin had a better season than Clemmensen, but past performance indicates that he is also a better goalie. Clemmensen faced fewer shots, suggesting that New Jersey is better than Vancouver at preventing shots by the opposition.

Wednesday, June 13, 2007

Vezina Trophy: Strength of Teammates

Strength of teammates is an important consideration for goaltenders. The stronger the team around them, the more likely they are to win games. A strong defensive team is also likely to benefit the goalie's performance statistics, such as save percentage and goals against average.

Both New Jersey and Vancouver took slightly more shots than they gave up. New Jersey took fewer shots and allowed fewer. The difference was even greater once you factor in the differences between the Eastern and Western Conferences. The Canucks took 0.3 more shots and gave up 0.6 shots more than their conference average, pretty much an average team in the defensive West. The Devils took 1.5 fewer shots per game than their conference average, and gave up 2.2 fewer, indicating that their style of play contrasted greatly with the rest of their conference rivals.

Both taking and allowing fewer shots than average suggests that the Devils intentionally play a lower-scoring type of game, and that one of the reasons they don't score many goals is because of that style of play. New Jersey scored a few more goals than they allowed in all three periods, which indicates that they only scored as many goals as they needed. A defensive style of play is particularly obvious in their first period results. Whereas second and third period scoring is often driven by the score at the time, the first period always begins 0-0. That makes the chosen style of play more decisive than the particular game situation. New Jersey's chosen style of play was clearly defensive: they scored the second fewest goals in the league in the first period, and gave up the fewest.

On ice statistics also support the idea that New Jersey intentionally plays a low-scoring game. Centers Gomez, Madden, and Elias all averaged about 19 minutes a game. Madden actually played more than either Gomez or Elias at even-strength, averaging 15:24 per game. Heavy use of their checking line reduces New Jersey's goal-scoring while simultaneously reducing the goals against. For Vancouver, their top two centers Henrik Sedin and Brendan Morrison averaged 18 minutes per game, while Ryan Kesler averaged 16, and Josh Green 11.

New Jersey was also better at coming from behind when trailing after one period than Vancouver was, which is a sign of offensive ability since even a defensive team has to attack when they are behind. The Devils were 9-11-4 when trailing after one, the Canucks 6-15-4. New Jersey was also more effective at closing out leads than Vancouver was, winning 31 of 34 games they led after 2. Vancouver won 30 out of 36.

New Jersey's forwards scored 185 goals and 430 points in 2005-06. Vancouver's scored 178 goals and 459 points last season. Based on career scoring rates, New Jersey's forwards would be expected to score 2.45 goals and 5.89 points per game. Vancouver's forward line that began the season would be expected to come in at 2.20 goals and 5.48 points per game. At the deadline, the Canucks acquired Bryan Smolinski, which added a bit of scoring depth and raised the expected totals to 2.36 and 5.88, very close to New Jersey.

Perhaps the most controversial aspect of the strength of teammate debate is the argument over whether Vancouver or New Jersey has a better defensive corps. Many would say Vancouver, but a key part of that is their offensive play - Vancouver's defencemen scored 149 points this season to New Jersey's 122. From a goalie performance perspective, it doesn't matter how good the defencemen are at scoring goals, just how good they are at preventing them. To try to evaluate this, I looked at the personal goals against averages of each team's defencemen over the last 4 seasons, based on the research done at the Hockey Analysis Group.

The average goals against while they were on the ice per 60 minutes over the last 4 seasons for New Jersey's current defencemen was 3.02. For Vancouver, it was 3.60.

This season was par for the course for New Jersey's defencemen. None of them, other than rookies Andy Greene and Johnny Oduya, had personal best goals against averages this season. Paul Martin and Colin White were over their 4-year PGAA average, and Brad Lukowich was just slightly under his, despite Martin Brodeur playing much better this year than he did over the 3 previous seasons.

For Vancouver, Brent Sopel, Sami Salo, Mattias Ohlund, Lukas Krajicek, and Rory Fitzpatrick all had a lower goals against rate this season than in any of the previous 3 seasons. Other than Willie Mitchell, every Canucks defenceman was below his average personal goals-against over the last 4 seasons.

The difference appears to be mostly from goaltending. The Canucks have in years past played a more offensive style, which certainly resulted in more goals against, but last year the Canucks gave up 29.9 shots per game, and according to Ken Krzywicki had average shot quality against. This year, they gave up 28.9 shots per game with easier-than-average shot quality. That is certainly an improvement, but not a huge one. However, their team save percentage went from .900 to .920, which is a bigger reason for the increase.

Several of the Canucks defencemen weren't even with the team last year. Of all the minutes played by Canucks defencemen, 54% were by players that were with the team in 2005-06, a substantial turnover. For New Jersey, 80% of the minutes played were by players who had been with the team last year. As might therefore be expected, New Jersey's defensive performance in 2006-07 was very similar to that of 2005-06. The Devils gave up 27.9 shots per game with easier-than-average shot quality, compared to last year's 28.8 shots per game (also easier than average).

The statistics indicate that New Jersey's defence is better at goal prevention, and that the main reason for many Vancouver defencemen having career defensive years was the goaltending. Based on past scoring history, the two teams' forwards are very similar in talent. New Jersey plays a more conservative, defensive style, which is why they scored fewer goals. The two teams are similar in makeup, style and talent. Overall, there is little talent difference between these two teams to tip the scales towards one goalie or the other. Perhaps one of the biggest differences concerning teammates was that Roberto Luongo had to adjust to a new team in a new conference with a revamped defence corps, while Brodeur played on the same team and system that he always has. Perhaps that explained some of Luongo's early season troubles; if he could play an entire season in his post-Christmas form, he would certainly be the best goaltender in the NHL.

Vezina Trophy: Strength of Opposition

Strength of opposition is a justification that has been used to credit both Luongo and Brodeur. Luongo backers claim that the Western Conference is superior, and that the Northwest Division is the best in the league. Brodeur fans argue that the Eastern Conference is more high scoring, with 14 of the top 20 goalscorers in the league, so it is therefore a tougher place for a goalie to play. Which side is right?

On this issue, the Luongo side is right. I have already dealt in detail with the differences between the Eastern and Western Conferences this year, and the higher scoring per game is a result of a more open game, rather than more talent. At the top end, the teams are just as good at scoring - 6 of the top 10 teams in the league in scoring are from the Western Conference.

That the Western Conference is stronger overall is very clear. The Western Conference teams had a winning record against Eastern Conference teams, and the Stanley Cup Finals were a mismatch in favour of the Western side. The New Jersey Devils were 2-6-2 against Western opponents, while the Canucks went 8-1-1 against the East.

Let's look at overall strength of opponent. I found these statistics on the Canucks.com forum, but even though the source may be biased I have verified that the numbers are correct.

Avg point total, Northwest Div: 94.2
Avg point total, Atlantic Div: 90.8

Avg point total, West Conf playoff teams: 106.5
Avg point total, East Conf playoff teams: 100.8

Sabermetrician Jeff Sagarin from USA Today keeps a detailed ranking of all the divisions, based on game results. According to his calculations, the Northwest Division is the toughest in the NHL. The Atlantic ranks 5th out of the 6 divisions.

Another consideration is that the last time the New Jersey Devils traveled outside of the Eastern time zone was November 27 (Source: ESPN). The Devils never had to battle jet lag, and would have been as rested as their opponents. Playing in Vancouver is a very different story, where a tougher travel schedule makes it more difficult to play well. Better teams and tougher travel, therefore, made Luongo's opposition this year tougher than Brodeur's.

Monday, June 11, 2007

Vezina Trophy: Goal Games

Since shutouts are worthless to evaluate goalies (link), let's look at the frequency of all goal-games for Luongo and Brodeur.

Here are the breakdowns, taken from an HFBoards thread:

0 goals: Luongo 5, Brodeur 12
1 goal: Luongo 20, Brodeur 15
2 goals: Luongo 19, Brodeur 19
3 goals: Luongo 20, Brodeur 18
4 goals: Luongo 8, Brodeur 8
5 goals: Luongo 3, Brodeur 4
6+ goals: Luongo 1, Brodeur 2

These types of charts aren't seen too often, so we need to introduce some context and make sure we are making fair comparisons. I used a binomial calculator (http://faculty.vassar.edu/lowry/binomialX.html) to calculate the expected number of shutouts, one-goal games, etc., for both goalies, as well as for an average goalie, based on save percentage and average number of shots faced.

Brodeur:

0 goals: 8.0 expected, 3.7 average, 12 actual, +4.0
1 goal: 19.0 expected, 11.8 average, 15 actual, -4.0
2 goals: 21.7 expected, 18.3 average, 19 actual, -2.7
3 goals: 15.9 expected, 18.4 average, 18 actual, +2.1
4 goals: 8.4 expected, 13.3 average, 8 actual, -0.4
5 goals: 3.4 expected, 7.4 average, 4 actual, +0.6
6+ goals: 1.5 expected, 5.1 average, 2 actual, +0.5

Luongo:

0 goals: 7.0 expected, 3.6 average, 5 actual, -2.0
1 goal: 17.4 expected, 11.5 average, 20 actual, +2.6
2 goals: 20.0 expected, 17.9 average, 19 actual, -1.9
3 goals: 16.1 expected, 17.9 average, 20 actual, +3.9
4 goals: 9.0 expected, 13.0 average, 8 actual, -1.0
5 goals: 3.9 expected, 7.3 average, 3 actual, -0.9
6+ goals: 1.3 expected, 4.9 average, 1 actual, -0.3

Martin Brodeur had 4 more shutouts than expected, but 4 less one-goal games. Roberto Luongo had 2 fewer shutouts than expected, and three more one goal games. The similarity of those numbers indicates that in all likelihood it just came down to luck - Brodeur got some breaks to preserve his shutouts, while Luongo was unlucky in losing his.

Luongo was slightly more consistent than Brodeur. In 59 of his 76 starts, he gave up 1-3 goals. In terms of quality starts (2 goals against or less), the two goalies were very close: Brodeur 46, Luongo 44. Brodeur had more bad games (gave up 4 or more 0.7 more times than expected). Luongo had fewer bad games (gave up 4 or more 2.2 times less than expected). Luongo was actually also slightly more likely to let in 2 or less (1.3 times fewer than expected, compared to Brodeur's 2.7 fewer than expected).

As you would expect from two goalies with very similar goals against averages, there isn't much between them. It just reinforces that the difference in goals against average is almost entirely from Luongo facing more shots per game. If I had to pick one distribution ahead of the other, I'd go with Luongo's, although the difference is very slight.

Vezina Trophy: The Worthlessness of Shutouts

Shutouts are an arbitrary stat. For forwards and defencemen no single-game performance, no matter how good, is treated specially in any way. Instead, it is the end of the season goal and assist totals that are used to judge how good they were. For some reason, goalies get bonus points when nobody scores on them through 60 minutes, but give up just one goal and they get no credit at all.

The closest player equivalent to a shutout is probably a hat trick, or a multi-goal game. But nobody would use the number of hat tricks a player has scored to determine who is a better scorer. They would look at the entire season's performance as a whole.

Teemu Selanne had 12 multiple goal games this season, including one hat trick. Dany Heatley had 11 multiple goal games, including 3 hat tricks. Both of them also had many games where they didn't score at all. They are the equivalent of a goalie who racks up a lot of shutouts. But neither of them are going to receive the Art Ross.

Vinny Lecavalier is, and he had only 6 multiple goal games, one of them a hat trick. His goal scoring was consistent, and didn't come in bunches. Nobody cares how he scored them, though. What matters is that he scored 52 times, the most in the league, and therefore he was the league's top scorer.

There are many who claim that Brodeur should be ranked ahead of Luongo, because he has 12 shutouts to Luongo's 5, and that should be the tiebreaker given the closeness of the other goalie stats. This is really absurd. The shutouts are already taken into account in the GAA and save percentages of the two goalies. If goalies have the same save percentage, and one has more shutouts than the other, that just means he also had more bad games than the other. Luongo is like Lecavalier, Brodeur is like Heatley. Who would you rather have, the guy more likely to score a hat trick, or the guy more likely to score?

Shutouts are also a team stat. Teams that allow fewer and easier shots make it easier for their goalies to get shutouts. A league average goalie playing behind a defence that only gives up 23 shots per game is more than twice as likely to get a shutout as an average goalie that faces 30 shots per game.

Shutouts have the advantage of guaranteeing a win, but in most games teams score 2 goals or more, meaning that there is very little difference between a shutout and giving up one goal, or even two. I have already discussed in a previous post how shutouts add very little to the expected team winning percentage, as compared to one goal games.

Shutouts are arbitrary, improperly weighted, strongly team-dependent, and poor measures of goaltender ability. They are an anachronistic goalie statistic that should be given almost no weight when evaluating goalies.

Vezina Trophy: Why Games Played Shouldn't Matter

One of Brodeur's biggest strengths is his durability. He plays nearly all of his team's games, and is able to maintain consistent performance in doing so. This is one of the main reasons that he has been able to pile up so many career wins. But how much should that be taken into consideration in Vezina Trophy voting?

In my view, not at all. This is for two main reasons: Games played is largely outside of the goalie's control, and the award is for the best goalie, not the most valuable one.

Goalies do not decide when they play and when they sit, the coach does. Some coaches platoon goalies, while others ride their starter. Obviously, goalies that are better tend to play more games, but it also depends on how good the backup is. For example, Tomas Vokoun's performance this season (.920) was very similar to last season (.919), when many thought he should have been a Vezina nominee. This year, however, he played in only 44 games because he missed some time to injury and because his backup is Chris Mason, who finished 2nd in the league in save percentage. The result is that very few would rate him even close to as highly as last year, even though he has been the same goaltender. Another example is San Jose, where the combination of Evgeni Nabokov and Vesa Toskala, combined with a coach in Ron Wilson who is not afraid to platoon them, means that neither goalie has a chance to play in 70+ games. This is even the case for Martin Brodeur: 1996-97 was possibly his best season (1.88 GAA, .927 save percentage). That year he played in only 67 games, down from his usual total of about 72-75, because he had a decent backup in Mike Dunham. There are several other factors why a goalie can play in more games than another, such as for example scheduling (e.g. number of back-to-back games) or whether a team is fighting for a playoff spot or not late in the season.

The official description of the Vezina Trophy includes the following: "The Vezina Trophy is an annual award given to the goalkeeper adjudged to be the best at his position". It does not say the most valuable, or the goalie with the greatest contribution, or the goalie with the most games played. Playing more games at a lower level does not make a goalie better.

For example, imagine a goalie holds out for most of the season, and only plays in the last 10 games, where he posts 10 shutouts in a row for an average team. Is he the best goalie in the league? Of course. His performance in the games he did play would have been way, way better than everyone else. He should unanimously win the Vezina Trophy. That is obviously hypothetical, but a real-life example of this may have been Miikka Kiprusoff in 2004. He played in only 38 games, but led the league with an excellent .933 save percentage. He was nominated for the Vezina Trophy, but lost to Martin Brodeur, mainly because he did not play enough games. Niklas Backstrom is a similar case this year, with a league-leading .929 save percentage in 41 games. He was not nominated for the award.

I am not saying that a goalie who plays 1 game and gets a shutout should win, nor would I prefer a goalie with 30 games played over another one with 75 just because the first one had a slightly higher save percentage, since the gap is not likely statistically significant. But if there is a clear difference in performance (after adjusting for team factors of course), the better goalie should be rated more highly even if they played many fewer games. Among starting goalies, say those who have played between 50 and 82 games, the games played mark is determined mostly by scheduling, the coaching philosophy, the talent of the backup, and the durability of the goalie, none of which have anything to do with how talented the goalie is or how well he stops the puck when he is in the game.

Brodeur has started three more games than Luongo. That is the main reason why he has more wins, since Luongo has a better winning percentage. It is also the only reason why he has faced more shots, since Luongo faced a higher number of shots per game. Therefore, those arguments should be thrown out the window. In this year's Vezina race, only rate stats should matter.

Sunday, June 10, 2007

Vezina Trophy Analysis - Luongo vs. Brodeur

I will be doing a series of posts breaking down the Vezina Trophy decision in detail. It is clear that either Luongo or Brodeur will win it, so I will focus only on those two goalies. This will allow for some detailed number-crunching that will hopefully provide a better perspective on which goalie is the most deserving of being named the best goaltender in the game.

Sunday, June 3, 2007

How Martin Brodeur Lost the Stanley Cup

With the Stanley Cup Finals underway, and with either J.S. Giguere or Ray Emery set to become immortalized as a Stanley Cup winner within the next two weeks, I thought to take a look back at one of Martin Brodeur's seasons in an attempt to demonstrate how misleading the "Stanley Cup winner" label can sometimes be.

What is one of the most-used arguments for Martin Brodeur's greatness? "He's won three Stanley Cups". This is often stated as if he singlehandedly led his team to victory. But how come you never hear about the Stanley Cup that Brodeur lost? The playoff year where one of the best teams in recent memory couldn't bring home the Cup mainly because of the poor play of their goaltender? Shouldn't that count for something?

In 2000-01, the New Jersey Devils, fresh off a Stanley Cup win in 2000, finished first in the Eastern Conference with 111 points. They led the league in goals scored (295), and tied for 5th in goals against (195). They had the best goal differential in the NHL, and their goals for/against suggested that they were actually major underachievers that year. Only two teams in the last 25 years had a better goal ratio than the 2001 Devils in a full season: the 1996 Red Wings, and the 1989 Flames, and that includes the mid-80's Edmonton Oilers dynasty.

The team was stacked, with two 40-goal snipers (Patrik Elias and Alex Mogilny), Petr Sykora (81 points), Scott Gomez (63 points), and Jason Arnott (55 points in 54 games). The checking line was outstanding at both ends, as Selke Trophy winner John Madden and Randy McKay both scored 23 goals and along with Bobby Holik shut down the opposition's best. The defence corps was led by the Scotts, Stevens and Niedermayer, and supported by Brian Rafalski, who had a breakout year with 52 points, and veteran Ken Daneyko. In goal was Martin Brodeur.

In the first playoff round, New Jersey eliminated 8th seeded Carolina in 6 games. The Devils dominated the 'Canes throughout, but Arturs Irbe had a couple of excellent games to delay the inevitable, despite facing 33.5 shots per game. Brodeur barely broke a sweat, facing just 19 shots per game and allowing just 8 goals.

The Devils moved on to play the #7 seeded Toronto Maple Leafs, who had done Devils a big favour by upsetting the second-seeded Ottawa Senators. One of the biggest reasons was goalie Curtis Joseph, who continued his excellent play against the Devils. New Jersey strongly outplayed Toronto in the first three games, but barely managed to take a 2-1 series lead on the strength of two overtime victories. Toronto then won games 4 and 5 to push New Jersey to the brink of elimination as Joseph continued to outplay Brodeur. In game 6, Brodeur finally played a good game, stopping 24 of 26, and New Jersey won 4-2. In the decisive game 7, the Devils put together a dominant performance, holding Toronto to just 16 shots in a 5-1 rout.

During the series, Brodeur faced just 20.5 shots per game, and had a dismal .878 save percentage against the 13th ranked offence in the league. Curtis Joseph outplayed Brodeur in 5 of the games, facing 28.6 shots per game and stopping pucks at an .898 clip. Despite the huge edge in shots and play, New Jersey outscored Toronto just 21-18.

The Conference Final opponent was the 6th seeded Penguins, a 96 point team during the regular season. With offensive stars including Mario Lemieux and scoring leader Jaromir Jagr, the Pens had finished second to New Jersey in goals scored. The problem was that the defence was weak, and the goaltending was so bad that Pittsburgh turned to untested rookie Johan Hedberg throughout their surprising playoff run.

That run would end against New Jersey. Pittsburgh's high-scoring offence ran into the brick wall of New Jersey's defence. In game one, the Penguins managed just 15 shots on Martin Brodeur in a 3-1 loss. In game 2, they got 23, but beat Brodeur 4 times in a 4-2 win. In games three and four, Brodeur posted back-to-back shutouts, making 20 and 21 saves respectively. Pittsburgh finally scored a couple of goals in game 5, but New Jersey closed out the series with a 4-2 win. Over the series, New Jersey outscored Pittsburgh 17-7 and outshot them 139-99. Brodeur had his best series of the playoffs with a .929 save percentage and 2 shutouts, although he was again rarely tested with just 20 shots against per game.

That led to a much-hyped goalie showdown in the Stanley Cup Finals: Martin Brodeur vs. Patrick Roy. Colorado was the President's Trophy Winner with 118 points, but they did it in a weaker division, had a worse goal differential than New Jersey, and were missing their best playmaker in Peter Forsberg who was injured for the Finals. Against the weakened Avalanche, the Devils were the better team.

In game one, however, the Devils came out flat, and Colorado jumped all over them, lighting up Brodeur for 5 goals in a 5-0 win. Game two was a tighter defensive affair, with both teams putting just 20 shots on net. With a 2-1 win, New Jersey grabbed a split of the games in Colorado.

The Avalanche took home ice advantage right back again with a 3-1 win in game 3, Brodeur giving up 3 goals on 21 shots. Facing a must-win situation, the Devils dominated game 4, outshooting Colorado 35-12. Brodeur let in 2 goals, despite being rarely tested, while Roy nearly stole the game, holding off the Devils until he made a costly puckhandling error to give up the tying goal, and the Devils found a late winner to tie the series. In game 5 in Colorado, New Jersey again outplayed the Avalanche, putting 4 goals past Roy. Brodeur stopped 22 of 23 shots in his best game of the series. The Devils were a win away from the Stanley Cup.

In game 6, Patrick Roy came ready to play. The Devils had early pressure, with three of the first four power plays and a 12-5 edge in shots through one period of play. But Roy stopped everything. Colorado scored 4 goals on Brodeur on only 18 shots, and Roy and the Avs blanked the Devils to send the series back to Colorado for game 7.

The season therefore came down to one game. The media focus was on Ray Bourque, attempting to win his first ever Stanley Cup, but the key players would be Martin Brodeur and Patrick Roy. In this high-stakes showdown, Brodeur blinked first, giving up a first period goal to Alex Tanguay, and in the second period was beaten again by Tanguay and then by Joe Sakic. New Jersey kept attacking, outshot Colorado yet again and managed to get one goal back, but that was all Patrick Roy would allow and the Avalanche took the Cup with a 3-1 win (Youtube).

The Stanley Cup Finals came down to goaltending. Over the 7 games, New Jersey outshot Colorado 178-146 and carried most of the play. Patrick Roy gave up only 11 goals in 7 games for a glittering .938 save percentage. Brodeur's numbers were mediocre: .870 save percentage and 2.71 GAA. The difference was even more glaring in the crucial games 6 and 7, where Brodeur let in 7 goals on 40 shots while Roy stopped 49 of 50. This is not just the stats either - from watching the games it was clear that New Jersey was the better team, but Colorado won because they got much better play from the goaltender position.

Brodeur was not merely outplayed in one playoff series by one of the greatest goalies of all time, he was actually one of the worst goaltenders in the 2001 playoffs. He ranked 15th out of 18 goalies in save percentage, and dead last among all goalies who made it past the first round. Brodeur's poor stats were despite New Jersey having a creampuff run to the Finals against the bottom three seeds in the East. He finished 3rd in GAA and tied for 1st in shutouts, but this was simply because he finished first in fewest shots per game with far and away the lowest total at 20.2 shots per 60 minutes. The next lowest was 23.7. On a game-by-game save percentage basis, Brodeur was outplayed by the opposing goaltender in 13 of his 25 games, including 10 out of 14 times against Roy and Joseph combined.

For comparison's sake, in 2004 Patrick Lalime was chased out of Ottawa after posting a .906 save percentage and a 1.95 GAA in 7 games against Toronto. Famed choker Marty Turco's career playoff record is a .909 save percentage and 2.21 GAA. Roman Cechmanek, last playoff season in the NHL: .909 save percentage, 2.15 GAA. In Curtis Joseph's single worst playoff year with the Oilers, Leafs or Wings he posted a .907 save percentage. In 2001, Brodeur went to game 7 of the Finals with an .897 save percentage, .017 worse than the average of all of the other playoff goalies combined, facing the weakest conference opponents he could possibly face and then a Finals opponent missing one of the best players in the league.

Despite the erratic performance by Martin Brodeur, the New Jersey Devils almost won the Stanley Cup. Dominant at both ends of the ice, the Devils were probably one of the best teams of the last two decades. But they did not win it all, and the main reason was goaltending. Brodeur may have been a bigger factor in his team's loss in 2001 than he ever was a positive factor in his team winning in 1995, 2000, or 2003. By my accounting, therefore, Brodeur should get credit for no more than 2 Stanley Cup wins. He still owes New Jersey one from 2001.