Friday, August 31, 2007

The Insightful Serge Savard

SI's Michael Farber recently gave us his all-time NHL team.

In his article, he included this interesting paragraph on the importance of goaltending:

"The goaltending position on the historical team -- or the Ghost of Christmas Goodies Past, as I like to think of it -- vexed me the most. Recently the esteemed Serge Savard, who narrowly missed making this team as a reserve defenseman, advanced the counterintuitive argument to me that goaltending is not the most important position because, in almost every case, teams make goalies more than goalies make teams. Remembering the 1999 Buffalo Sabres and Dominik Hasek, Savard still has some convincing to do."

Savard is not completely right - goaltending is the most important position, since the potential effect of a goaltender is much greater than that of any other single player. However, I generally agree that the importance of goalies is generally overstated. The position is important, but performance is actually quite similar among most goalies at the top levels, especially after removing team effects. The difference between many good goalies is just small fractions of goals per game, which means that you often need a substantial edge in the crease to make a losing team into a winner, or vice versa. This is because the rest of the team combined is three or four times more valuable than the goalie in terms of their contribution to the final result. So it is actually fairly rare that a goalie wins or loses games because of their singular efforts. Pointing out one exception to the rule (Hasek) doesn't invalidate that entire viewpoint. In fact, it is good evidence for that particular goaltender as one of the all-time greats.

I wonder what Savard thought about his former teammate Ken Dryden. If you take his statement to be true in all cases, then Savard probably thinks Dryden was lucky to be playing on the teams he did.

Farber should have paid more attention to Savard, because he picked Terry Sawchuk as the backup on his all-time team, a goalie who is very overrated because of the strong Detroit teams he played on early in his career. Farber's top goalie ever was Roy, which isn't a bad choice, but probably not the correct one, and the reason again is the hidden factor of team strength.

It is difficult to argue with choices for all-time teams, because different observers rate things differently. For example, some focus more on career numbers and longevity, while others heavily weight a goalie's prime. Some look at playoff performance above all else, while others just see it as part of the mix.

I don't have numbers yet that I believe are conclusive in terms of ranking the best of all-time. However, I think Hasek should probably be in the top spot, and would probably lean towards Jacques Plante in a very tight decision over Glenn Hall, Tony Esposito, and Roy for the #2 position.

Friday, August 24, 2007

Testing the Teammate Theory

To test how well we can evaluate goaltenders by comparing them to their teammates, I decided to look in some more detail at the last two seasons. This allows for a direct comparison with shot quality data to see if there are parallels, and to see whether the two methods deliver similar results.

I calculated GAA compared to backups and save percentage compared to backups (with adjustments) for all the goalies who had been starters over the last two seasons. I also used shot quality information available from Alan Ryder and Ken Kryzwicki at Hockey Analytics (taking care to attempt to correct for the variance in shot distance reporting around the league recently discovered by Ryder), and used that to find out whether each goaltender's save percentage was above or below predicted levels, and by how much.

It turns out that the methods have quite similar results. The correlation coefficient between save percentage vs. predicted save percentage from shot quality and save percentage vs. teammates' save percentage was 0.73. Six of the top 10 goalies in performance vs. backups were also in the top 10 in performance against the predictions of the shot quality model (Huet, Hasek, Lundqvist, Lehtonen, Kiprusoff, and Legace). In addition, both methods agreed on the top man: Montreal's Cristobal Huet. Here are both lists:

Performance vs. Predicted (Shot Quality), 2005-07:
1. Cristobal Huet
2. Dominik Hasek
3. Tomas Vokoun
4. Henrik Lundqvist
5. Kari Lehtonen
6. Miikka Kiprusoff
7. Roberto Luongo
8. J.S. Giguere
9. Manny Legace
10. Ray Emery

Performance vs. Backups, 2005-07:
1. Cristobal Huet
2. Miikka Kiprusoff
3. Henrik Lundqvist
4. Martin Brodeur
5. Tim Thomas
6. Kari Lehtonen
7. Dominik Hasek
8. Ryan Miller
9. Manny Legace
10. Marc-Andre Fleury

Vokoun illustrates one of the problems with using backup stats: His backup is Chris Mason, a good goaltender, which makes it difficult for Vokoun to rate highly against his own teammates. Giguere is in a similar situation. Emery is knocked down by having backed up Dominik Hasek in 2005-06, while it is likely that Luongo's backups simply overachieved in the few games Roberto was given off, probably also against soft opposition.

On the other list, Brodeur and Miller were just outside the top 10 in shot-quality adjusted save percentage, so they aren't way out of place, although they both made it on because their backups did not do very well compared to their predicted save percentages. Thomas and Fleury benefitted from some very poor play by their backups (mostly Toivonen and Thibault, respectively).

Therefore, this seems to show that while using teammate data cannot replicate shot quality data, it can in many cases do a pretty good job of estimating it, as the results are correlated and the rankings similar. There will always be a few goaltenders rated too high or too low because of the strength or weakness of their teammates, even after adjustments, so this method requires some care in interpreting the results. Nevertheless, a discerning application of the method to the goalies of seasons past should yield some useful and interesting results.

Sunday, August 19, 2007

The Best Regular Season Goalie of All-Time

I have completed a comprehensive study of every goalie who has played 10,000 or more minutes in the NHL since World War II. I measured their performance in terms of GAA in each season against the other goalies on their team. This was done because goalie results are strongly impacted by the team around them. Goalies who cannot outplay their teammates are not very valuable to their teams, since their performance can be easily replaced. Therefore, the best goalies should be ones who outperform their teammates the most, after adjusting for controllable factors like era, minutes played, strength of backup goalies, etc.

All GAA results were adjusted to league average to allow comparisons across eras. I adjusted for the strength of their teammates with an adjustment factor based on their career length, performance and reputation. I ignored seasons where the goalie played less than 120 minutes, or his backups played less than 120 minutes, because those samples are too small to be meaningful yet could easily skew the results. I took the teammate goalies' combined performance and calculated how well they would have been expected to do, given the starters' minutes. I also further adjusted the starter and teammates' performances based on the number of minutes played, given that the more minutes played, the greater the statistical significance of the performance.

That left me with an adjusted starter GAA, and an adjusted backup or teammate GAA. Each goalie was then ranked by how well they performed relative to their teammates' adjusted performances. I also calculated these results for each goalie's five best consecutive seasons, i.e. their career prime. The results are described below.

There are a few limitations to this study I would like to outline here:

The adjustment factors for each goalie's backups were subjective. Changing them does not change the results substantially, but it would likely impact the rankings. I found a lot of goalies that I considered to be good based on results and reputation were indicated to be not so good by the results, and vice versa. Perhaps I will go back and revise the factors based on some of my new findings and see how the results change.

Also, it was more common in the 1940s and 1950s for a single goalie to play all of his teams games. Goalies like Glenn Hall, therefore, are underrated in this analysis because in their best seasons they did not have any backup results to compare with, so those years unfortunately could not be considered.

The way a team conducts its goalie rotation can have an impact on the results. For example, some teams only play their backup against weaker opponents, while others essentially platoon. It would be incredibly time-consuming to adjust for opponent on a game-by-game basis for every goalie, so this has been left out. I think it is common for backups to play weaker opponents, so any adjustment would likely be in a similar range for most of the goalies, but it could be that there are some goalies who were advantaged or disadvantaged because of difficulty of opposition.

Another adjustment I would have liked to have made was to introduce an age curve. Several goalies, such as for example goalies who are still active, rank much higher than they should because they do not have a decline stage, a number of games late in their career when they are no longer playing at their peak. Such an adjustment would not penalize goalies with longer careers.

On to the findings:

The best regular season goalie ever: Dominik Hasek. Hasek ranked first overall and had the best prime. Hasek was 29.9% better than his backups in terms of adjusted GAA, and 41.5% better during his prime. Given that Hasek came over to the NHL late and missed out on a number of potential seasons in his mid-20s, and is still active beyond the age of 40, for him to take first place on this list is very strong proof that he is the best goaltender to ever play NHL hockey, at least in the regular season.

Second place was a surprise: Al Rollins. Rollins took second both overall (27.4%) and prime (37.0%). Rollins toiled on weak Chicago teams in the 1950s, but was clearly a strong goaltender, as shown by his Hart Trophy award in 1954. Rollins did not have a long career (9 seasons), and is not in the Hall of Fame. However, these results indicate that his only limitation was his team, and that his performance was outstanding.

In third place was Miikka Kiprusoff (26.0%). Kiprusoff ranks high because the only results we have are from the prime of his career. He has no decline stage, and he did not play much early on. This shows how well he has played over the last few seasons, but he should definitely be ignored in terms of talk of the greatest of all-time. There were a number of other modern goalies who similarly popped up higher than they should have for the same reasons. However, Kiprusoff did rank 5th overall in terms of career prime (26.2%, basically the same as his overall career), so he is clearly an outstanding goaltender.

In fourth place was another expected name: Jacques Plante. I was not sure what to expect with Plante, whether he was a creation of some powerful Montreal teams or if he really lived up to the billing. The results indicate that he clearly was one of the greatest goalies of all-time. He had a very long career, and still easily outperformed his backups (20.9%). In terms of prime, Plante ranks third overall (31.8%). Therefore, looking at his career as a whole, Plante has a strong case to be ranked as the second best regular season goalie of all time after Dominik Hasek.

In fifth place was another well-known name, but possibly a controversial one. Ken Dryden played for some amazing teams, which has caused speculation that he was just a product of those teams. However, Dryden was far better than his backups, letting in 20.1% fewer goals. Again, Dryden had the advantage of a short career without much of a decline stage, so he ranks a little higher than he maybe should. In terms of prime, Dryden is 11th (19.8%). However, the numbers show that Dryden should probably be considered in the top 10 all-time. He played for great teams, but his performance was also apparently great.

In sixth place comes a bit of a shocker: Roman Cechmanek. Cechmanek played just four whole seasons in the NHL, most of them behind a good defence in Philadelphia, but he easily outperformed his backups (19.5%). However, there are reasons to be cautious about this result. First, Cechmanek had a very short career. Secondly, some of Cechmanek's teammates (Esche, Boucher) ranked very low, indicating that his results were probably somewhat inflated because his backups were very poor. Cechmanek's prime (which is essentially the same as his career) ranks 12th. So Cechmanek's awkward style was probably more effective than it looked during his time in the NHL, at least during the regular season.

In seventh place is Bernie Parent (18.2%). Parent also had a dominant prime, ranking 4th (30.6%). Parent is another goalie that played on some great teams, but also put up some great performances.

Eighth was another virtual unknown: Roy Edwards (17.5%). Edwards had 7 NHL seasons in the late 1960s and early 1970s. He is a classic case of a goalie that never got a chance because of the 6 team league, and also one who was underrated because he played on weak teams. Upon expansion, he won a job and proceeded to put up some relatively outstanding numbers on a weak Detroit team that missed the playoffs in 5 out of 6 seasons Edwards played for them. This was despite being 31 years old by the time he broke into the NHL. Edwards' prime ranks 6th overall (24.7%).

Ninth is another active goaltender, and a very interesting result: Manny Legace (17.0%). Legace is the most underrated goalie in the NHL today. He has had an outstanding career as a backup on Detroit, since he nearly always outperformed his more highly-rated teammates (Curtis Joseph, Chris Osgood, Dominik Hasek). He had a very good season last year on a weak St. Louis team as well. Legace ranks 21st in terms of career prime (16.9%). Legace will drop some in the years ahead as he ages, but he is definitely a top goaltender in the league.

Rounding out the top 10 is Mario Lessard. Lessard had a short career in Los Angeles in the early 1980s, but put up excellent numbers. However, his prime ranks 23rd overall, so one of the reasons he comes in so high here is because he did not have a decline stage. Nevertheless, he is another underrated goalie.

The rest of the top 10:
11. Dan Bouchard (prime: 8th)
12. Marty Turco (prime: 33rd)
13. Tony Esposito (prime: 10th)
14. Arturs Irbe (prime: 9th)
15. Bob Froese (prime: 13th)
16. Chico Resch (prime: 16th)
17. Charlie Hodge (prime: 14th)
18. Rick DiPietro (prime: 37th)
19. Fred Brathwaite (prime: 15th)
20. Daren Puppa (prime: 62nd)

OK, enough of that. Now let's go to the fun stuff: pointing out the frauds.

First, let's see how our blog's namesake did:

Martin Brodeur:

vs. backups: 4.5% better, 52nd place
career prime: 18.4% better, 18th place

So Brodeur has a pretty solid prime (1996-2000), but his career as a whole is not that impressive. Just more evidence to throw on the already huge pile that most of his success is owed to the team in front of him.

How about Grant Fuhr:

vs. backups: 2.2% worse, 115th place
career prime: 1.7% better, 93rd place

Terrible results for Grant Fuhr. He was just as good as his backups in terms of numbers. As far as a regular season goalie, it is virtually impossible to make a statistical case for Grant Fuhr as even one of the best of his era, much less all-time. However, I don't think many of his backers would even look at regular season results when arguing for Fuhr's greatness.

Most people probably noticed the absence of Patrick Roy in the top 20. So where does he come in?

vs. backups: 7.3% better, 31st place
career prime: 13.9% better, 29th place

Patrick Roy is overrated. There, I said it. Roy was a great goalie, no question, but his teams were almost always very good, and that was a big advantage throughout his career. He definitely should be considered in the greatest of all-time argument because of his unbelievable playoff record, but he is not one of the best regular season goalies ever.

How about Billy Smith:

vs. backups: 3.0% worse, 121st place
career prime: 1.1% worse, 115th place

Most people don't really see Smith as a regular season goalie, of course, because he is remembered almost entirely for playoff results. Nevertheless, I believe Smith to be one of the most overrated goalies of all time. His playoff results were only great for four seasons when the Islanders were an unstoppable dynasty, and his regular season results were very mediocre.

Terry Sawchuk:
vs. backups: 4.6% worse, 135th place
career prime: 17.5% worse, 153rd place

Terry Sawchuk, considered by many to be the greatest goalie of all-time, had the single worst career prime rating of any goalie in my study. How is this possible? Well, let's look a little closer at Sawchuk's career. Sawchuk played five seasons as the starting goalie in Detroit from 1950 to 1955. In all of those seasons he had GAAs under 2, he led the league in wins all five seasons, and he won 3 Vezinas and 4 Stanley Cups. I think pretty much everyone would agree that was the prime of his career. So was it the goalie or the team? Well, let's look at his backups. In those 5 seasons combined, Sawchuk's backups were 8-1-3 with a 1.47 GAA. This indicates that Sawchuk's Detroit teams were amazing.

Sawchuk had a long career, so that pulls his results down a lot. But over half of his career shutouts came in a five-year stretch with a Detroit team that was clearly a dominant team, and the results of his teammates show that Sawchuk clearly benefitted heavily from the defence in front of him. Sawchuk is another goalie that is clearly very overrated. His longevity, as well as his good fortune to have played for half a decade on a very dominant team, has led to great career numbers. As a whole, however, he was worse than the other goalies on his team, and is certainly not one of the best of all time.

I welcome any comments on the study's methods and results, and I plan to analyze the results in more detail as I continue to try to identify goalies with performances that do not match their reputations, both good and bad.

Monday, August 13, 2007

The "Other Teams Play the Trap" Argument

An oft-repeated argument to defend Martin Brodeur is that it cannot be claimed that the trap is the reason for his success, since if it was, everyone else would just copy it. Other teams have not been able to emulate the Devils success, so it must be someone they have on their team that is the difference-maker. The main guy that has been there the whole time has been Martin Brodeur, so he should get the credit for the Devils' run of success.

This argument is partly correct, but is nevertheless wrong with respect to Martin Brodeur. It is true that if the strategic option of playing the trap dominated all others and did not depend on a team's talent level, then every team would use it and it would even itself out. But what matters is not the strategy alone, but how well that strategy is implemented. That is, how good is the team at playing the trap? Every team in the league plays a very defensive style with a one goal lead late in the third period, because it makes strategic sense, but clearly some teams are better than others at actually pulling it off.

A similar example I can use is the power play. Every team in the league has a power play unit, but some are better than others. Some teams are very good for a number of years on the power play (a recent example is the Detroit Red Wings, top 5 in the league in power play efficiency every year from 2000-01 to 2005-06). By similar logic to that above, it could be argued that other teams should be able to copy what they do and eliminate their advantage, but what is really driving the success is not just power play tactics, but also coaching, teamwork and talent. It wouldn't matter if the Chicago Blackhawks used the exact same power play setup as the Wings, moved the puck around in a similar manner and took the same types of shots, they just wouldn't be as effective.

It is glaringly obvious both on the statsheet and in real life that no matter how many teams used the trap, nobody played it as well as New Jersey. From 2000 to 2004, the worst New Jersey finished in fewest shots against was 2nd, and the most shots they allowed per game was 24.7. This was despite the other teams in the league having a decade to copy New Jersey's style. And it wasn't just Brodeur, since his average was actually 1.5 shots per game higher than his backups (who faced just 22.3 shots per game). Alan Ryder points out that the New Jersey Devils have always led the league in his measurements of shot quality against. This is a team where the immortal Corey Schwab had a 1.27 GAA in 14 games over 2 seasons. The Dead Puck Era New Jersey Devils were absurdly good at defence.

The Devils have generally had a deep defensive unit with a number of excellent defensive defencemen, as well as excellent defensive forwards on their checking line. This is still the case in New Jersey; it is easy to label newcomers like Johnny Oduya as untalented, but Oduya was on the ice for just 37 goals against in 1110 minutes of even-strength ice time this season, a good rate that was better than the rest of the team's when he wasn't playing (source: Behind the Net). Defensive play is hard to judge, but New Jersey seems to keep finding players who are very responsible in their own end, or perhaps developing them through their strong coaching staffs at the NHL or minor league level. The Devils are also very well managed by Lou Lamoriello, who usually brings in players that fit well into the New Jersey system. These are all reasons why New Jersey has been dominant defensively for over a decade, not merely the style of play that they employ.

There are, in fact, some teams that have copied the New Jersey model with some success. The best example is probably the Minnesota Wild, with Jacques Lemaire, although there are a number of other teams playing a tight defensive style, especially in the Western Conference. After the early years after expansion, Minnesota has been quite similar to New Jersey in team defensive statistics. Over the last few seasons, their goaltending has been every bit as good as New Jersey's in terms of save percentage. The Minnesota Wild have basically been the New Jersey Devils in a tougher division and with less overall talent. It is mainly those two factors that explain why the Devils have been at the top of their conference and the Wild have been in a constant struggle just to make the playoff struggle, not merely the choice of defensive style.

No strategy on its own will be successful if executed poorly, and, similarly, there are a number of strategies that will be successful if they are well-executed on a talented team. The Devils have won with a very defensive system mainly because they have done the best job of recruiting and developing talent that complements the system. Regardless of how many copycats there have been, it has only been since the lockout that a number of other teams have begun to rival New Jersey in terms of team defensive play. This coincides with significant talent losses for New Jersey (led of course by Stevens and Niedermayer). Without talent, no system of play is likely to be successful, and it has been primarily team defensive talent that has driven New Jersey's success. Even though some teams have copied New Jersey's tactics, nobody has played it as well as they have, and that is why Brodeur has had a relatively easy job throughout his entire career.

Monday, August 6, 2007

Picking the Wrong Guy, Part 1: 1987-88

This is part one of a series on questionable Vezina voting decisions, where team factors fool NHL GMs into choosing the wrong goaltender as the league's best.

In 1987-88, Grant Fuhr won the only Vezina of his career. He also finished as the runner up to Mario Lemieux for the Hart Trophy. Fuhr outpointed Wayne Gretzky (149 points in 64 games), Steve Yzerman (50 goals, 102 points in 64 games), and Denis Savard (131 pts in 80 games). Four voters even listed Fuhr in first place ahead of Lemieux, who had scored 70 goals and added 98 assists to lead the league in scoring.

Fuhr had not done particularly well in previous years in Vezina voting. His best result was second place in 1981-82. So what was the difference? Was it perhaps an improvement in the team? Or did he merely have a career year?

Edmonton lost Paul Coffey from their blue line after the 1986-87 season, but his strengths were mainly offensive and it is doubtful they lost much on the defensive side. The other main blueliners (Lowe, Muni, Gregg, Smith, Huddy, McSorley, Beukeboom) all returned for 1987-88. The Oiler forward group was very much the same as the year before. As might be expected, then, the team played defence at a similar level. The team shots against totals were almost the same - 28.9 per game in 1986-87, 28.8 in 1987-88.

Given the continuity, therefore, one would expect Fuhr's numbers to be similar. Indeed, his performance rates were almost exactly the same in 1987-88:

1986-87: .618 win %, .881 save %, 3.44 GAA
1987-88: .610 win %, .881 save %, 3.43 GAA

So if his performance didn't improve, how then did Fuhr go from 3rd in Vezina voting with 0 Hart votes to being considered the best goalie in the league and more valuable than Wayne Gretzky?

The very simple answer: Andy Moog left to play on the Olympic team.

Andy Moog and Grant Fuhr spent six years sharing time in the Edmonton net, and were for all intents and purposes the same goalie. Most years their save percentages were very similar, and their records for Edmonton during the dynasty period were almost identical: Fuhr 107-39-14, Moog 104-37-14. No matter who was in net, Edmonton usually won.

Moog's departure in 1987 meant that Edmonton only had 21-year old Bill Ranford and 22-year old Darryl Reaugh on their team. This meant they chose to rely heavily on Fuhr. Fuhr played 75 games, leading the NHL. He also led in minutes played, wins, and shutouts. Martin Brodeur fans will probably recognize that combination. However, 14 other goalies played 2000 minutes or more and finished ahead of Fuhr in save percentage.

Was Fuhr deserving of the Vezina? Let's look at his stats again, compared to his teammates in the Edmonton net:

Fuhr's numbers: .610 win %, 3.43 GAA, .881 save %
Fuhr's backups: .714 win %, 3.95 GAA, .876 save %

Fuhr's backups' numbers are a little misleading because of the influence of one Warren Skorodenski, a rarely used career backup who saw his final bit of NHL action with Edmonton in 1987-88. In 61 minutes of play, Skorodenski was beaten 7 times for a catastrophic 6.89 GAA. Since Fuhr's backups didn't play many games in 1987-88, Skorodenski has a large impact on the stats. Here is the stat line without him included:

Ranford and Reaugh: .714 win %, 3.59 GAA, .890 save %

That compares very favourably to Fuhr, and provides evidence that his season really wasn't that special at all. As had been the case for many years with Moog, Fuhr did no better than his backups had done. Ranford was coming off of a 41 game season with an .891 save percentage in 1986-87, and he improved to .899 in 6 games in 1987-88, so he probably could have played more games at least at the same level as Fuhr. Fuhr was far from the league's best goalie, and he deserved little consideration as league MVP. Rating Fuhr as more valuable to the Oilers than Gretzky has to likely be considered one of the most curious award voting decisions in the history of the NHL.

Let's look at the way the Vezina voters ranked the rest of the top 5 after Fuhr. I have included their performance statistics, as well as that of their backups to get the team context.

2. Tom Barrasso, 2-3-3, 22

Barrasso's numbers: .569 win %, 3.31 GAA, .896 save %
Barrasso's backups: .466 win %, 4.45 GAA, .860 save %

3. Kelly Hrudey, 1-4-0, 17

Hrudey's numbers: .557 win %, 3.34 GAA, .896 save %
Hrudey's backups: .542 win %, 3.22 GAA, .893 save %

4. Brian Hayward, 2-1-1, 14

Hayward's numbers: .667 win %, 2.86 GAA, .896 save %
Hayward's backups: .625 win %, 2.97 GAA, .898 save %

5. Mike Vernon, 0-4-2, 14

Vernon's numbers: .685 win %, 3.53 GAA, .877 save %
Vernon's backups: .556 win %, 4.36 GAA, .858 save %

Barrasso's numbers are very impressive. His backups were Daren Puppa and Jacques Cloutier, both of whom were decent goalies. Barrasso's numbers were very good for any team, much less a weak Sabres team coming off of a sub-.500 campaign in 1986-87. He deserved to have won the Vezina.

Kelly Hrudey's numbers are very good, but the fact that they were pretty well matched by the 38-year old Billy Smith testifies to the Islanders' strong defensive play. It was a similar situation in Montreal, where Brian Hayward and Patrick Roy both played well, but the main reason for their success and Jennings Trophy win was their very strong defensive team.

Vernon doesn't look that great overall, except for his winning percentage. But the fact that his backups could win at a .556 rate despite a 4.36 GAA shows how good the Flames were. The Flames #2 goalie was Doug Dadswell, who played 25 of his 28 career NHL games in 1987-88, explaining the weak performance of the backup goalies. Vernon was no better than average in 1987-88, and only received notice because of his 39 wins as the primary starter on an excellent team.

Coming in sixth place in the voting was the Rangers' John Vanbiesbrouck, who might have been the second best goalie in the league in 1987-88:

Vanbiesbrouck's numbers: .545 win %, 3.38 GAA, .890 save %
Vanbiesbrouck's backups: .438 win %, 3.56 GAA, .876 save %

The backup was Bob Froese, who had a number of very good years in the 1980s, yet was well outplayed by Vanbiesbrouck.

The rest of the top 10 (Lemelin, Roy, Hanlon, and Stefan), all played for very good defensive teams, and had similar stats to the other goalies on their teams.

Were there any goalies that went unnoticed on bad teams? A couple. Daniel Berthiaume of Winnipeg had a .531 winning percentage and .882 save percentage on a weak team, much better than what his backups did, and Darren Pang stopped shots at an .891 rate behind the porous Chicago defence (35.2 shots against per game), although Pang's teammate Bob Mason also did pretty well.

The writers fell into the same trap as the GMs in their All-Star voting. Fuhr again finished first, taking 58 out of 61 first place votes. Patrick Roy was second, followed by Tom Barrasso. The rest of the list went more or less by the strength of the goalie's team, with Malarchuk, Vernon, Lemelin, Hextall, Hrudey, Peeters, Hayward, Liut and Hanlon.

In summary, then, the statistics show that Fuhr was neither outstanding nor especially valuable in 1987-88, and his per-game performance was almost exactly the same as what he had done the previous year. This means that the voters got it completely wrong. Fuhr's Vezina win was the result of the Oilers' loss of Andy Moog, which led to more games played and therefore more wins and shutouts. This attracted the attention of the voters, who, as they often do, overrated durability and gave too much credit for team success. They even went so far as to claim that Fuhr was more valuable than Wayne Gretzky, a completely laughable assertion. The 1987-88 Vezina should have instead gone to Tom Barrasso.

Friday, August 3, 2007

The Critical Importance of Team Context

There are so many synergies in hockey that the team context always needs to be taken into account for all players, not just for goaltenders. Edmonton just acquired Dustin Penner from Anaheim, giving up 3 high draft picks and committing to $21.5 million over 5 years. I doubt that Edmonton's front office staff uses any kind of advanced statistical metrics, or they likely would never have made such a move.

There were probably no more than a handful of players in the entire league who benefitted as much from their linemates and from soft minutes played as much as Dustin Penner. He might be one of the last players in the league I would have rolled the dice on, not because I don't think he is good, just because he is completely and utterly unproven, and I don't care how many goals he scored last season, I still can't be sure that every one of them wasn't because of Corey Perry and Ryan Getzlaf, or because of weak opposition, or both. Not to mention Penner playing almost 3 minutes per game on one of the league's top power play units.

Getzlaf and Perry were #664 and #665 out of #676 players in terms of easiest minutes played according to Behind the Net. Over the course of the season, Penner's minutes were slightly harder (he ranked #572), but a lot of his scoring came together with those two guys playing the same kind of soft minutes. Looking at the scoring rates, the impression you get is that if anything Penner was dragging the other two down. Here are the per game goals for and goals against on the ice stats for Getzlaf and Perry, with and without Penner, from David Johnson's Hockey Analysis site:

Getzlaf with Penner: +2.19, -1.97
Perry with Penner: +2.24, -2.24

Getzlaf without Penner: +3.40, -1.59
Perry without Penner: +3.25, -1.44

Second line scorers on good teams can often be like goalies playing on strong defensive teams - their teammates make them look way better than they actually are. I don't think this signing will end well for Oiler fans, and all because their management (which should have learned this lesson by now, see Lupul, Joffrey) apparently didn't stop to take the time to consider team effects in their decision making.

Wednesday, August 1, 2007

Evaluating Puckhandling

One of Martin Brodeur's strengths is puckhandling. It is difficult to evaluate puckhandling since there are very few stats to go on, meaning we have to mostly rely on subjective judgment. There is no consensus on the subject, as shown in this HF Boards thread, where a number of Brodeur fans claim he is the greatest puckhandler ever, while others disagree, arguing that Brodeur is not even the best in the game today, surpassed by others like Marty Turco and Rick DiPietro. Given that there seems to be substantial agreement on those three as the best in the league, I will focus mainly on those three in the following analysis.

How can we try to analyze puckhandling? Probably the most obvious place to start is with the scoring numbers, goals and assists. Brodeur is famous for scoring a goal in the playoffs, but goalie goals are obviously very rare, as are goalie assists. This year, Turco had 4 assists to lead all goalies. DiPietro had 2, while Brodeur had just one. However, one season can be a bit fluky - some of the goalies who beat out both DiPietro and Brodeur in terms of assists include Vesa Toskala, John Grahame, Dwayne Roloson, Olaf Kolzig, Marc-Andre Fleury, and Johan Holmqvist, none of whom are particularly well-known for their puckhandling.

Looking back a few years to compare Turco and Brodeur can give perhaps a better comparison. In the last 4 seasons, Marty Turco has 10 assists to Brodeur's four. There were a number of goalies who also outscored Brodeur, many of them not known for puckhandling, like, for example, Roberto Luongo. Consistently piling up assists may be evidence of some puckhandling ability, but I think it is a poor measure because of the luck factor. Sometimes a goalie will get a second assist just for setting it up for one of their defencemen, because a teammate makes a superb play. Also, assists do not measure how well a goalie makes the easy plays, just how likely they are to connect on a home-run pass.

A statistic that can perhaps shed more light on the situation would be giveaways. The problem with giveaways, as with all of the NHL's real-time stats, are that scorers around the league vary substantially in how generous they are with awarding them. Therefore, we need to keep that in mind when interpreting the numbers.

Martin Brodeur's numbers came out as pretty solid - 33 giveaways in 78 games. However, his opponents had just 27 giveaways. There could be several reasons for this, such as the Devils having a less aggressive forecheck, but probably the main reason is that the scorers covering Brodeur seem to credit fewer giveaways. In the 4 games involving backup Scott Clemmensen, for example, there were just 2 giveaways for both teams (both charged to Clemmensen). Brodeur almost certainly handled the puck more often than most of his opponents, but his higher giveaway rate means he would have had to have made substantially more good plays just to break even with the goalies playing against him. It's certainly not unreasonable to suggest that he did, but the giveaway numbers don't seem to indicate that Brodeur is head and shoulders above his peers in this regard, as some observers claim.

Marty Turco had 31 giveaways in 63 games. His opponents had 44 giveaways, indicating that the scorers were more generous with the giveaways. This is reinforced by the games involving backup Mike Smith, as Smith had 12 giveaways and his opponents had 15 in 19 games. So although Turco had a slightly higher rate of giveaways per game, he did much better than his opponents, and when taking that into account it is likely that he outperformed Brodeur, at least in terms of taking care of the puck.

DiPietro is a very different story. He had an awful 75 giveaways in 61 games, compared to just 39 for his direct opponents. The scorers seemed to be pretty hard on the Islanders (backup Dunham had 12 giveaways in 15 games), but DiPietro's numbers are still very poor. His giveaways per game rate is almost triple Brodeur's, and he was charged with almost twice as many giveaways as his goaltending opponents. DiPietro may be able to shoot it hard, but he clearly needs to improve in terms of taking care of the puck, and it is doubtful that the positives make up for his very high error rate.

Martin Brodeur has the established reputation, but I believe the best puckhandling goaltender in the NHL is his successor on the competition committee, Marty Turco. This is from a combination of subjective observation and the above number-crunching. Turco is very aggressive with the puck, and his team's defensive style relies heavily on him playing the puck. Nevertheless, he still commits substantially fewer giveaways than his opposing goalies, and he led the NHL in assists. Brodeur was not charged with many giveaways either, but his opponents had even fewer, indicating that the official scorers watching the games were a little trigger-shy in handing out giveaways. Rick DiPietro is not close to as good as either of those two. He makes far too many mistakes, and isn't scoring enough points to make up for it.

This was a very superficial look at puckhandling, but objective analysis is very limited by the numbers available. I would need to collect more numbers to look at other goalies, as well as to get more of a league average benchmark to more properly analyze giveaway rates and possibly adjust for scorer bias in various cities. One other point that I hope to look at some time is whether some goalies reduce the number of hits taken by their teammates when they are in the game. For the time being, combining giveaways and assists with a subjective analysis of the goalie's play will have to suffice.