Monday, December 31, 2007

Mike Keenan Knows Very Little About Goaltending

It is often difficult to advance contrarian positions about hockey players that contradict insider perspectives in hockey. A common retort in many of the comments and emails I receive goes something like, "But (Player X) or (General Manager Y) said Brodeur was great - how can you disagree?" General managers vote on awards like the Vezina Trophy, and their collective wisdom helps shape fan perception of certain players. However, some of these insiders have shown by their actions that they know little more about how to evaluate goalies than an average fan. One of the worst offenders has been Mike Keenan, a celebrated coach who has made his living in hockey despite being clueless about goalies for over 20 years.

In Philadelphia, Keenan built a solid defensive system that made his goalies look good most of the time. However, he chose Pelle Lindbergh as his starter over Bob Froese in 1984-85, and in 1985 until Lindbergh was killed in an accident. Lindbergh wasn't a bad goalie, but he was easily outplayed by Froese. Over the two years, the Flyers were 44-12-3 with Froese in net and 46-19-7 with Lindbergh. Froese had 2.51 GAA and .910 save percentage, Lindbergh just 3.00 and .898. It should have been obvious from the 1983-84 season who was better (Froese 28-13-7, 3.14, .887 at the age of 25; Lindbergh: 16-13-3, 4.05, .860 at the age of 24), but apparently not to Keenan. After the tragedy with Lindbergh, Keenan had to turn to Froese, but he also used the mediocre Darren Jensen and the aging Glenn "Chico" Resch. Keenan still never liked Froese, despite his excellent save percentage numbers, and handed the starting job to rookie Ron Hextall for the 1986-87 season. Midway through the year, Froese was traded to the New York Rangers.

Hextall put up worse numbers than Froese had done in 1985-86 (3.00, .902), but he won 37 games behind behind the Flyers' strong defence and received the Vezina Trophy. After this charmed season, Hextall's numbers then fell off significantly in 1987-88 (3.51, .885).

Keenan moved on to Chicago in 1988. As usual, he spent his first year there tinkering with the goaltending, challenging the incumbent starter Darren Pang with two rookies, Ed Belfour and Jimmy Waite, as well as the newly acquired Alain Chevrier. The goalie shuffle would continue in Chicago for another season, as both Jacques Cloutier and Greg Millen were brought in via trade, but both played poorly and were soon out of the league.

Dominik Hasek arrived in the NHL for the 1990-91 season, but Keenan wasn't a fan and held Hasek to just 5 games that first season, even though Hasek performed well in his first taste of NHL action (3-0-1, 2.46, .914). However, Belfour became the starter ahead of Millen and Cloutier (who was traded mid-season), playing 74 games and winning the Vezina Trophy with a 43-19-7 record, a 2.47 GAA, and a .910 save percentage. Keenan earns partial credit here for at least going with Belfour, even if he probably could have done even better with the Dominator.

Keenan moved upstairs to the GM position for the 1991-92 season, but he still didn't realize what he had with the Dominator. Hasek again put up solid stats (10-4-1, 2.60, .893), but Keenan traded him in the offseason to Buffalo for Stephane Beauregard. Iron Mike decided that he would rather have Jimmy Waite as a backup than Dominik Hasek, even though in 1991-92, Waite played 17 games with a 3.69 GAA and an .844 save percentage.

Jimmy Waite is a great example of a goalie that Keenan completely mis-evaluated. Waite played 53 games for Chicago with Keenan as coach or GM, and nearly all of them were awful - 3.72 GAA and .853 save percentage. Yet Keenan thought he was good enough to stick around, and traded one of the greatest goalies of all-time (Hasek) to move Waite up on the depth chart.

Keenan was fired in Chicago after the 1992-93 season, and found work in New York, where he made his legend by winning the Stanley Cup. With Mike Richter in net, Keenan couldn't possibly screw it up in New York. However, his term on Broadway was very short, and they were still celebrating the Cup win when he decided to move on.

Iron Mike soon found himself brought into the St. Louis Blues organization as both coach & GM. There, Keenan inherited another stellar netminder in Curtis Joseph. However, he decided to sign Shayne Corson away from Edmonton, which cost the Blues two first round picks in compensation. To get the picks back, Keenan traded Joseph and Mike Grier to Edmonton. Corson scored just 20 goals in 88 games in St. Louis, while Curtis Joseph remained one of the league's best goalies for the next decade in Edmonton, Toronto, and Detroit.

To replace Cujo, Keenan signed the aging Grant Fuhr and kept the incumbent Jon Casey, who was also getting up in years, as the backup for the next two seasons. In a low-scoring era, Fuhr was below average (.902 save percentage), and Casey was downright awful (.864). Despite a blue-line that included Al MacInnis and Chris Pronger, the Blues never won anything under Keenan, often because they were let down by their goaltending. Even after Keenan's departure, St. Louis continued to ride Fuhr to disappointing results, until they finally replaced him with Roman Turek and were rewarded with the President's Trophy in 1999-00.

Keenan moved to Vancouver, where he had Arturs Irbe and Kirk McLean. Irbe had a strong season in Vancouver in 1997-98 (2.73, .907), but he wasn't re-signed and moved on to Carolina where he had an even better one (2.22, .923). Pat Quinn was GM in Vancouver, and it is difficult to know how much input Keenan had on some of the goaltending decisions, however it is probably reasonable to assume that Quinn at least consulted with Keenan before making his moves. Midway through the season, Vancouver made a pretty good trade, swapping Martin Gelinas and Kirk McLean for Sean Burke, Geoff Sanderson, and Enrico Ciccone. However, the Canucks managed to mess it up again two months later, getting rid of Burke again in exchange for Garth Snow. Burke was admittedly not particularly outstanding in his brief stint in Vancouver, albeit on a weak team, but trading him for Snow was a blunder. Over the next 5 seasons, Sean Burke established himself as one of the best goalies in the game, posting excellent statistics (2.40, .917). During the same period, Garth Snow continued to be mediocre (2.74, .904), only outlasting Keenan in Vancouver by one season before leaving via free agency. Again, that one is probably more on Quinn than Keenan, but in any event, Iron Mike's choices for backup goalies in Vancouver, Corey Hirsch and Kevin Weekes, were terrible, combining for a 3-16-4 record with awful save statistics.

In 2000-01, Keenan coached the Bruins. He had a solid starter in Byron Dafoe (2.39 and .906 in '00-01), but throughout the course of the season Dafoe missed 35 games to injury. Keenan used four other goalies in backup and replacement duty, but they were mostly terrible, including Peter Skudra (.879), John Grahame (.867), and Kay Whitmore (.809)(!). Having such terrible goalies play big minutes was one of the major reasons the Bruins finished 9th in the East and missed the playoffs.

Keenan was fired as a result of the disappointing finish, but he found a new home in Florida, where he had an elite young goalie in Roberto Luongo. Keenan managed not to screw this situation up for three full seasons until just as he was on his way out the door, when he pulled the trigger on the now infamous trade with Vancouver (Luongo, Lukas Krajicek and a 6th round pick to Vancouver for Alex Auld, Todd Bertuzzi, and Bryan Allen). Shortly after, Keenan resigned as GM.

Keenan is now in Calgary, where he has another star goalie in Miikka Kiprusoff. Kiprusoff has been playing poorly this season, but whether that is influenced by Keenan, is because of poor team defence or just a slump is difficult to determine. It will be interesting to see how Keenan handles the situation. If he was also the GM, I would almost expect him to trade Kiprusoff for pennies on the dollar and bring in a goalie who is way past his prime (Ed Belfour, perhaps?) to take his spot.

Mike Keenan may be a good hockey coach in terms of motivation and team defensive play, but he should never be allowed to make personnel decisions about hockey goalies. His record in that department is absolutely terrible. Nearly every decision or trade he has made involving goalies has turned out bad, and some of them spectacularly so. Keenan traded Hasek, Joseph, and Luongo, and got almost nothing in return for all three of them (all the players acquired combined for just 227 games with their new teams). He also chose to play other goalies ahead of Bob Froese and Sean Burke, which probably contributed to them being traded away for little returns, and he played some brutal backups for far too many games (most notably Jimmy Waite and Jon Casey). Keenan is an extreme case, but the reality is that some scouts, GMs and coaches are not good at evaluating goalies, and whether or not they get paid by an NHL team or have a Vezina vote doesn't change that fact at all.

Friday, December 21, 2007

On Games Played

One of the differences of opinion I have with many other hockey spectators is with respect to the importance of games played for a goaltender. I maintain that in almost all cases there is no significant difference between a goalie who plays 50 games and a goalie who plays 80 games, other than the coaching and team management philosophy that led to those results. Goalies should be durable enough to play the lion's share of their team's games, but extra starts beyond that are mostly just an opportunity for certain goalies to pad their stats, simply because their coaches like to ride their starters, they have established reputations, their teams are up against the cap or because their backups are weak.

I have not seen any evidence that goalie performance deteriorates with more minutes played. Remember the talk about how Brodeur was tired in the playoffs last season? Over the last 6 seasons, Brodeur has the following line in regular season games held in April: 22-3-1, 1.87, .928. I also once did a long post about Grant Fuhr and the 1987-88 season, where he went from 44 games played to 75 games played because of the absence of Andy Moog, and put up an identical save percentage and a virtually identical winning percentage. His performance was exactly the same, just with more of it, and that won him the Vezina. Evgeni Nabokov played 56% of the minutes in San Jose last year, and did pretty well. This year he has played 98% of the minutes and his performance is very similar. Did he work on his durability over the summer? Of course not, his backup went from Vesa Toskala to a guy who put up an .888 save percentage in the AHL last season. Not surprisingly, Nabokov's coach gave him more starts.

Virtually every goalie in the NHL is capable of logging big minutes and playing well. To get to that level, they would have been the starters in minor hockey, junior or college hockey, and in most cases the minor leagues as well. I looked at the 30 current starting goalies in the NHL to see how many of them were experienced goalies (more than 5 full years in the league) without multiple good seasons with 65+ games played. Here is the complete list:

Martin Biron, Rick DiPietro, Johan Hedberg, Manny Legace, Dwayne Roloson, J.S. Giguere

I'd say there is enough evidence that Biron, DiPietro, Roloson and Giguere are capable of playing big minutes. Giguere and DiPietro have both fallen just short of my arbitrary 65 game cutoff a number of times, and Biron and Roloson have both spent most of their careers as backups or platoon goalies, although they have been quite good in the full seasons they have played (Roloson has played in 108 of 136 games since being traded to Edmonton, Biron's career best save percentage was during his 72 game season in 2001-02 and he's been great carrying the load in Philly).

So I'm not sure that Johan Hedberg and Manny Legace can handle big minutes as starting goalies, but other than that I'm pretty sure every current experienced NHL starting goalie is capable of handling a lot of games.

But Brodeur has done it year after year for over a decade, you say. Yes, he has. So have guys like Olaf Kolzig, without nearly the same level of attention. There are two main reasons why guys see their games played reduced: their backups get better, or their performance drops off. Brodeur's games played dropped when New Jersey had Mike Dunham from 1996-98. His playing time increased again when Dunham was replaced by weaker backups, even though Brodeur's performance dropped from .927 in 1997-98 to .906 in 1998-99, and then .910, .906, and .906 over the next 3 seasons. Brodeur's advantage, however, was his low shots against, so his low GAA and high win totals helped mask the fact that his performance wasn't as good. A goalie on a weaker team who had that kind of deterioration in save efficiency might have dropped below .900 in save percentage, and his coach would probably be looking to give somebody else a shot. Brodeur had an established reputation, was on a team that inflated his perceived performance level, faced fewer and easier shots than other goalies, and played with poor backups. And yes, he is very durable. So it is not surprising at all that he ended up playing a lot of games.

But surely it must be valuable to have a goalie log a lot of minutes at an above average level of play? Yes, of course, since any team would rather have above average play than mediocre play, and if the goalie is better than any other option on the team then they are better off with him in the nets as much as possible. But what about a goalie who plays fewer games, but at a higher level of performance? Is that better or worse than a Brodeur-type who is always in the net?

Well, let's take some hypotheticals here. Let's compare a .910 goalie who plays every single game in a season for his team, against a .925 goalie who, because of various personal and injury issues, isn't capable of such a demanding schedule. Let's assume that his backup goalies are terrible (.890), and that his team is about average in terms of shot prevention (30 shots per game). How many games does the second goalie have to play for his team to be ahead of the first team? The answer is just 42 games. And if his team has a quality backup, say one capable of performing at about league average (.905), the goalie needs to play in just 21 games to break even with his more durable counterpart.

Save percentages can often be misleading, because it can look like there isn't much difference between .910 and .920. In fact, a .920 goalie is much better than a .910 goalie. The higher the save percentage, the harder it is to maintain. I am of course speaking about performances over a fairly large sample size, where we can more accurately ascertain the true level of play of the goalie.

Durability is good, because you obviously want your best goaltender on the ice instead of in the press box more often than not. However, whether a goalie plays 50 games or 82 games is really not that meaningful. I'd rather have a truly excellent goalie playing in half of my games than a minutes muncher giving me league average performance in every single one of them, even if the second guy is going to end up with more wins and shutouts and Vezina Trophy votes. Alas, truly excellent goalies are rare, so teams have to make do with what they can find, so from a team management perspective it might make your job easier to find an average/above average goalie and play the heck out of him, and then cheap out on your backup. That does not, however, mean that your goalie magically becomes one of the league's best or most valuable just because he plays every night. The best goalies are the ones with the best rate stats, not the best counting stats.

Friday, December 14, 2007

Building a Defensive System

Do you need a great goalie to build a defensive system? Some would claim that whenever there is a combination of team and goalie that have a run of success that the team was able to build their defensive strength around the goalie. Does an elite defensive system require a great goalie, or is this reasoning false?

To answer this question, let's imagine a team that has a brick wall as a goalie. Yes, a 4' x 6' brick wall in net. The other team cannot score no matter how many shots it takes. What kind of strategy should that team use? Should it play a tight defensive style and try to minimize scoring chances against at the expense of offence? Or should it take every offensive risk possible, since it only needs one goal to win the game? Obviously the latter is the correct strategy, since there is no risk to giving up scoring chances and shots, and a very strong benefit for taking risks to score goals (any goal = guaranteed win). And conversely, what about a team playing with no goalie in the net? What kind of strategy should they use? In real life, teams often use an aggressive strategy with no goalie in because they are trying to tie the game, but what if there was no goalie and the score was tied? Such a team would obviously play very tight defence, because every shot on net becomes a goal. Therefore they would be very defensively focused to try to prevent shots at all costs, using a tight defensive shell and pressuring the puck carrier to block their shots or force a turnover that may allow a chance on the counterattack.

This is why claiming that a goalie is responsible for a defensive system is completely wrong. Grant Fuhr supporters may or may not be right about some things, but one thing they are definitely right about is that a great goalie is valuable for an offensive team because it allows them to take additional risks to try to score goals. If your goalie is better than the goalie on the other team, it makes sense to play a more open game with more shots on each net. If each team gets 10 shots of equal difficulty, a journeyman goalie could easily beat a Hall of Famer with a little luck. If both goalies face 60 shots, on the other hand, it is much more likely that the better goalie is going to win the game for his team.

With Dominik Hasek in net, the defencemen on the 1999 Buffalo Sabres were aggressive on their pinches, knowing that even if they gave up an odd-man rush it was likely that Hasek would make the save. In contrast, last year's Ottawa Senators played a tight defensive system and blocked as many shots as possible to try to protect their inexperienced goaltender Ray Emery. Both were the correct strategic moves, and helped them reach the Stanley Cup Final, since the trade-offs in terms of allowing/preventing scoring chances against were to their team's benefit.

Playing a defensive system does not indicate that there is a strong goalie in the net. On the contrary, a strong defensive system is most needed and most beneficial when there is a weak goalie in the net. With a great goalie, a team should take more risks to try to generate offence, since the star netminder reduces the risk of giving up extra goals from the additional scoring chances allowed.

This is also why it is so important to adjust goalie statistics for team factors. Great defensive systems can hide weak goaltenders, just as poor defensive systems can make excellent goalies look average. With the advantages of a great defensive team, a truly elite goalie should dominate, finishing at or near the top of the league in save percentage and GAA, in addition to getting their wins and shutouts. If a goalie posts an average save percentage on a great defensive team, that is a strong indicator that he is being covered by his teammates. The same thing, in reverse, can apply to goalies with a high save rate on bad teams, as their teammates may be taking a lot of risks to score and leaving their goalies open to high-quality chances against and making their performance more impressive.

Thursday, December 6, 2007

Is Brodeur....Underrated?

Jonah Keri has written an article about the 10 most underappreciated active pro athletes. Coming in first place was Martin Brodeur. Either this was Opposite Day at ESPN, or it shows how clueless ESPN has become about hockey, given that one of their columnists ranked one of the most overrated goalies of all-time as the most unappreciated athlete in sports.

I've already summarized the arguments against Brodeur, so I'm not going to do that here. The comments thread to the article, however, illustrates again how the Brodeur debate essentially falls into two camps, which, despite what Internet message board flame wars would have you believe, aren't actually divided by their personal like or dislike of Brodeur. The difference is simply in their philosophies of evaluating goaltending play, which clash head-on in the case of Martin Brodeur.

The first camp associates team success with the individual goaltender, and the second camp believes goalie play is heavily dependent on the rest of the team. Those in the first group see Brodeur as one of the greatest, if not the greatest, goalie ever, because of his wins, shutouts, and Cups. None of those things mean anything at all to those in the second camp, who look at his save percentages and the strength of his teams and conclude that Brodeur is actually quite ordinary. The reason this perception gap is so large is that Brodeur has been the beneficiary of the most favourable goaltending environment in the NHL since Ken Dryden.

So we have probably an irreconcilable debate about Brodeur, at least until everyone comes to some agreement on the evaluative criteria. Nevertheless, when you take a deeper look at goaltending numbers, it becomes quite clear that the "team dependency" side has much stronger ground to stand on. If you still need convincing, though, check out these 12 examples for starters, and since this is basically the founding philosophy for this blog there is lots more written on this topic available on this site.

Sunday, December 2, 2007

The Dead Puck Era

The period from 1995 to 2004 in the NHL is sometimes referred to as the "Dead Puck Era" in reference to the decreased scoring environment. In this period, the average goals per game decreased and the average save percentage went up. Some of this effect may have been from improved goaltending play, but a lot of it was due to better defensive play, relaxed rules on clutching and grabbing, increased goalie equipment size, and other factors.

How much did this era help the goalies who played in it? Quite a bit, as it became easier to post high save percentages and low GAA numbers. One of the major beneficiaries, of course, was Martin Brodeur. It is easy to look at his career numbers (even his save percentage numbers), and wonder how anyone could claim that several other goalies could have easily matched his performance in New Jersey. Yet that's what I'm going to try to demonstrate right now.

Roberto Luongo, probably the best goalie in the NHL, is in his eighth season in the league. I wanted to look at the goalies who entered the league between Brodeur and Luongo, i.e. roughly somewhere between 1993 and 2000, to see how playing in that particular era impacted their save statistics. For a fairer comparison, I took only their first 8 seasons, starting with the first year they played more than 1000 minutes. Both are somewhat arbitrary cutoffs, but I needed to set the yardsticks somewhere, and for most goalies 8 seasons includes a few prime years and goes up to around 29 or 30 years old. For goalies in their 8th season like Luongo, I included this year's stats. Here are the save percentage results:

1. Roberto Luongo, .919
2. Manny Fernandez, .913
2. Tomas Vokoun, .913
4. Martin Brodeur, .912
4. Mike Dunham, .912
6. Evgeni Nabokov, .911
6. Guy Hebert, .911
6. Martin Biron, .911
6. Olaf Kolzig, .911
10. Jose Theodore, .910
10. Dwayne Roloson, .910

(There are a few others who didn't make the list because they only had 7 seasons played but would also have ranked ahead of Brodeur, including Manny Legace (.916), J.S. Giguere (.916), Marty Turco (.914), and even David Aebischer (.912).)

What conclusions can we draw from this list? Simply that the effect of the era they played in was much stronger than the individual goalie effect. The difference between #2 and #10 on this list is very slight, just .003 in save percentage (1 goal every 333 shots), and that is over an 8 season span. The only guy who distinguishes himself is Luongo. Brodeur is mid-pack, despite playing on the best teams (especially defensively) of any goalie on the list. It is interesting to see him tied with Mike Dunham, who began his career as Brodeur's backup in New Jersey.

Some of the goalies played fewer minutes, because they spent times as backups, which means one or two good seasons could be skewing their results. Those who played over 20,000 minutes in their first 8 seasons include Luongo, Vokoun, Brodeur, Nabokov, Hebert, Kolzig, and Theodore. So there is a good deal of evidence that except for Luongo they were all pretty close to each other in performance over their first 350+ games in the NHL.

Olaf Kolzig and Martin Brodeur had almost the same playing time and the same save statistics (29,432 minutes and .911 for Kolzig, 30,055 and .912 for Brodeur). Pretty much the only difference between the two of them was the strength of their teams. Olaf Kolzig is certainly one of the league's most underrated goalies - I would argue that over their careers, there is little to differentiate him from Martin Brodeur. The only difference between them is quantity, but not quality.

Martin Brodeur has had a very good career, but this has been primarily because of his great defensive teams and his longevity, durability, and favourable deployment by his team (i.e. his coach sending him out to the net virtually every game). For the majority of his career, he has stopped the puck just as well as the other decent goalies in the league, even before considering how relatively easy his shots were. He just came into the league at the right time and on the right team to put up massive career numbers. The low-scoring era Brodeur played in is another major reason for his impressive career statistics, but relatively speaking they are far from great. Brodeur's save percentages are often invoked in arguments involving some of the slightly older goalies like Roy, Belfour, and Joseph, but once you factor in era there isn't much difference at all (except for Roy, who ranks well ahead of Brodeur).

Just as an aside, there was one goalie that did not make the list because his first qualifying season came in 1991-92, but I thought I'd just mention him for comparison's sake: Dominik Hasek. Despite playing in three relatively high-scoring seasons to start his career, in his first 8 seasons Hasek posted a .926 save percentage, which outclasses everybody on the above list. In the Dead Puck Era, the low-scoring environment made it look like everyone was good at stopping the puck, but the record shows that nobody was even close to Hasek.