Monday, July 28, 2008

Why Tony Esposito Lost in the Playoffs

Bring up Tony Esposito with an older hockey fan, and they'll probably think of two things: a few games from the 1972 Series, and Esposito losing in the playoffs. Esposito has the stigma of being a playoff underachiever. But was this really true?

Tony Esposito played in 14 playoff seasons in his NHL career. Guess how many times his team lost in the playoffs to a team with fewer regular season points than his own?

The answer is once. Only one time, to the Montreal Canadiens in 1971, mainly because of the outstanding play of their rookie goaltender Ken Dryden. That was, however, the same year of the single most defining moment of Esposito's career, when he let in a shot by Jacques Lemaire from the red line in game 7 of the Stanley Cup Finals. That helped spark a Canadiens comeback and Chicago ended up losing the Cup.

So Esposito is remembered by many as a bad playoff goalie because he gave up a weak goal in game 7 of the Stanley Cup Finals in the only time in his career that his team lost to a weaker playoff opponent. That hardly seems fair.

There is another major reason for his reputation, as well, and that is that he spent most of his career in an expansion and WHA-diluted NHL full of stacked dynasty teams. During Esposito's time in Chicago, there were several great teams - Orr's Bruins, Clarke's Flyers, Lafleur's Habs, Trottier's Islanders, and Gretzky's Oilers. In the playoffs against those teams combined, Chicago was 7-34 during Esposito's career. Against everyone else they went 49-31.

So the Blackhawks held serve against their equals or inferiors but got stomped by the giants. Esposito certainly had more than his share of run-ins with the elite teams as well. In 9 of his first 11 years in Chicago, Esposito's team lost in the playoffs to the best or second best regular season team. In one stretch, Esposito even went 4 playoff seasons without winning a single game. That's not particularly impressive, however a big reason for this was that his average team had just 75 points, and the average opponent racked up 116. Esposito's teams never scored more than 3 goals in a game over that stretch, and an amazing 10 out of 16 times they scored 1 goal or less, meaning that in over half the games Esposito played he needed a shutout to win.

Yet just playing against a stronger opponent isn't completely an excuse if a goalie played very poorly. How does Esposito's individual performance stack up? This is a more difficult question to answer, especially since we don't have official save percentages from those years. Looking at the overall numbers, his GAA went up in the playoffs compared to the regular season, from 2.92 to 3.07. However, I don't believe his playoff performance was actually any worse. Again, the primary reason for the discrepancy was the relative strength of his teams and opponents.

On teams that finished in the top 5 in the league, Esposito had playoff numbers of 36-25, 2.85. Nothing extraordinary, but that winning percentage is well above his career regular season average and his GAA is also lower. On teams that finished outside of the top 5, Esposito was just 9-28, 3.42.

In addition, when he under the biggest spotlight in his career during the 1972 Summit Series against the Soviet Union, Esposito clearly outplayed his teammate Ken Dryden and was right on par with another Hall of Famer in Vladislav Tretiak (series save percentages found here). Sure it is a small sample, but it goes against the belief that Esposito simply folded his tent in the meaningful games.

I am not claiming that Esposito was a great playoff goaltender, simply that he was not a choker. There isn't much evidence that Esposito was a huge difference-maker in the postseason, but it would have been hard to be in his era. Most of the time he and his teammates simply ran up against a juggernaut in the playoffs and were dispatched in a short series. Perhaps a better goalie could have stolen an extra game here or there, but there seems to be little reason to believe that would have significantly altered Chicago's playoff outcomes, other than in 1971.

So if you want to fault Tony Esposito for letting in 20 goals in 7 games against a high-scoring team with 8 future Hall of Famers in 1971, then go ahead. But to claim that makes him a playoff choker is not supported by the evidence. Playoff team success is borderline irrelevant in evaluating goalies, because it depends on so many different factors. It is probably more fair to exclude it entirely, and Tony Esposito is a good example of why that is the case.

Wednesday, July 23, 2008

The Effect of Playoff Seeding

I ran across an interesting stat recently. Patrick Roy opened the playoffs on home ice in 14 out of his 17 playoff seasons. Dominik Hasek opened the playoffs on home ice in 4 out of his 10 playoff seasons. Given those different situations, should we be surprised that Roy had more playoff success? How much does playoff seeding contribute to the success of a goaltender?

I decided to investigate the impact of playoff seeding a little further. The period I looked at was 1996-2008. This was so that variance from era effects wouldn't be too large, and also I wanted to look at seedings done by conference rather than by division to avoid strength of division as a factor. I took the starting goalie for each team (if there were several goalies sharing minutes, I took whoever played most in the playoffs as the #1 guy), and looked at both regular season and playoff statistics broken down by playoff seeding. This also allows us to see the team effect on goaltending - how do goalies tend to do on top teams as compared to weaker ones in terms of GAA and save percentage?

Looking at the aggregate regular season stats, GAA was very strongly correlated with seeding (0.88), as better teams usually give up fewer shots and fewer goals against. Save percentage was a much weaker correlation (-0.57). Higher seeded teams tended to have slightly better save percentages, but there was just a .006 difference between the highest-performing rank and the lowest one, compared to a 0.29 variance in GAA. This supports what I have been saying that GAA is very team dependent, and although save percentage does depend on the team to some degree it is much more dependent on the individual goaltender.

I also looked at the difference between the playoff stats and the regular season stats. GAA showed a clear correlation, as the top three seeds had average playoff GAA decreases of 0.18, 0.11 and 0.14 respectively, while number 8 seeds were hammered to the tune of a 0.34 increase in GAA. A similar effect was evident for save percentage - it went up .004, .004 and .005 for the top 3 teams, and down .004 for the worst team in each conference. The rest were around average, except for the 7th seeds which were a clear outlier - 0.15 GAA decrease, .010 increase in save percentage. There have been quite a few upsets in the 2 vs. 7 matchups which helps explain this a bit, but probably a major reason for this was J.S. Giguere's outstanding performance in 2003.

This illustrates a problem with the analysis - Giguere played 1,407 minutes at 1.62/.945 in 2003, but someone who did poorly and lost in the first round would not count for so much. For example, Steve Shields was the goalie for the #7 seeded team in the Eastern Conference the same season. Shields split playing time as his team lost in the first round, and as a result only recorded 119 playoff minutes at 3.03/.897. This means that Giguere's outstanding but unique playoff year was weighted 12 times as heavily as Shields' mediocre yet more typical one.

To correct for this error, I took the averages of the yearly GAAs and save percentages for each goalie, and compared them, thus weighting every season exactly the same. These results were very strongly correlated with playoff seeding. Average playoff GAA had a correlation coefficient with seeding of 0.93, which is very high. Average save percentage was also strongly correlated with seeding at -0.82, much higher than the regular season results. The #1 seeded goalies averaged 2.11/.913, while the #8 seeds averaged 3.03/.900. Here is the comparison of average regular season vs. playoff stats:

#1 seeds: 2.18/.914 vs 2.11/.913
#2 seeds: 2.34/.915 vs 2.34/.913
#3 seeds: 2.41/.911 vs 2.49/.911
#4 seeds: 2.34/.914 vs 2.48/.913
#5 seeds: 2.48/.908 vs 2.72/.904
#6 seeds: 2.44/.913 vs 2.70/.909
#7 seeds: 2.47/.911 vs 2.64/.908
#8 seeds: 2.50/.910 vs 3.03/.900

Another very strong correlation was between playoff seeding and playoff shots against per game (0.92). The top seeds saw just 24.4 shots against per game, while the bottom seeds faced 30.7.

A major reason for the discrepancy in the playoff numbers is that top seeds get to play against lower seeded teams in the earlier rounds. Facing weaker shooters and fewer shots per games gives those goalies a great chance to succeed. On the flip side, goalies on lower seeds are usually peppered by shots from top-level shooters. As a result, playoff goalies have large discrepancies in shot quality against. Take, for example, Hockey Numbers' shot quality stats from the 2006-07 playoffs, available on this site. Shot quality factors varied from a low of 0.66 (Dallas) to a high of 1.49 (Nashville). This range was over three times as wide as it was over the 82-game regular season, where the values were all between 0.86 and 1.11.

Looking at the regular season numbers, there is little difference in save percentage between teams at the different rankings, but GAA climbs steadily upwards as the teams get worse. This implies that goalie skill is fairly evenly distributed across the teams, and that a team's ranking is mostly determined by the skaters. I think it is reasonable that top teams would have slightly better goaltending on average, since they tend to prefer veterans and experienced players while weaker teams are probably more likely to take chances with castoffs or young players, but the numbers indicate this effect is not very large.

In conclusion, the effect of seeding on playoff performance is very strong at the team level. Seeding is very highly correlated with GAA, save percentage, and shots against per game in the playoffs. To a lesser extent, these effects are also visible in the regular season results. In both the season and playoffs, the difference between the best and worst teams in GAA is much greater than the difference in save percentage, which indicates that the primary cause of superior goal prevention is fewer shots allowed, rather than better goaltending.

This means that is mostly unfair to fault goalies who played on weaker teams for their lack of postseason team accomplishments. On the other hand, goalies who were routinely on a team that entered the playoffs as a high seed were in a much better climate for postseason success. Playoff team success, therefore, should be pretty much disregarded, and team strength and opposition should be taken into account when comparing playoff performance between goalies. For the most part, evidence suggests that goalie play actually has very little to do with playoff success. The Cup is almost always won by a top team, and the large majority of individual playoff games are won by the team with the better scoring chances. A goalie can occasionally have a large impact on a single game, but it is very difficult to be the single deciding factor in a series and virtually impossible to drag a weak team to a Stanley Cup.

Monday, July 7, 2008

Chicago's Newest High-Priced Goalie

Joining Nikolai Khabibulin ($6.75 million per year) in Chicago is Cristobal Huet ($5.6 million per season) to make the league's highest paid goalie tandem.

I think it is about time to slap Cristobal Huet up on my "underrated" list, especially after Brian Costello of The Hockey News ranked Huet's signing with Chicago as the 10th worst of this free agency. "Huet is not even among the league's top 10 or 15 keepers", he wrote. Of course Costello was probably referring to THN's own list, published in their March 4 issue of this year, which had Huet at #22 in the league, behind such accomplished goaltenders as Martin Gerber, Ty Conklin, and Vesa Toskala. (You can read the list again for entertainment's sake here). Judging from a few threads on HF Boards a lot of fans agreed with Costello, ranking Huet among the worst signings of the day.

The statistics, however, show a quite different picture of Huet. Total save percentage since the lockout:

1. Cristobal Huet, .921
2. Tomas Vokoun, .919
3. Martin Brodeur, .918
4. Henrik Lundqvist, .917
5. Roberto Luongo, .917

Huet is not the best in the league, but I would rank him in the top 10. How Huet at $5.6 million per year can be considered a worse move than Theodore at $4.5 million doesn't make sense to me. For comparison's sake, Theodore's save percentage over the last three years was .895. Over the last three years, Cristobal Huet faced 449 more shots than Theodore, and let in 53 fewer goals. Factor in the expected goals on those extra shots against, and Huet's raw puckstopping performance has been about 100 goals better than Theodore's since the lockout. Now, granted, Theodore had a nice bounce-back season, and was probably even a bit better than Huet last year after considering team factors, but the two of them are only a year apart in age and Huet's much better recent track record makes him a better bet going forward.

Huet had a decent year last year on a Montreal team that allowed a lot of shots, but was traded to make room for prospect Carey Price. In Washington, Huet played behind probably the best defence he has ever played with, and put up the best numbers of his career: 11-2-0, 1.63, .936. Huet is sometimes criticized as not being a big-game goalie, but he is also in the top 10 in playoff save percentage since the lockout.

Huet is one of those goalies that will probably always have his perception lag his statistical performance, just because he is, to use a cliche, solid but not spectacular. In fact he is usually a good deal better than solid, but you don't always get the sense in watching him that you are watching one of the league's best. However, like I always say, at the end of the day the only thing that matters is how many scoring chances did you face and how many pucks went in the net, and on that score Huet does remarkably well.

Saturday, July 5, 2008

They "Saw Him Good"

This post by Dan Tolensky on the Leafs' recent signing of Jeff Finger was an interesting read. Tolensky shows how Leafs GM Cliff Fletcher and coach Ron Wilson were completely wrong in their descriptions of Finger's role on his team. Wilson and Fletcher claimed that the defenceman was matched up against the Western Conference elite, even though Finger actually played the opposition of a bottom-pairing defenceman. And of course, even though the Leaf management was demonstrably wrong in their assessment there were many Leaf fans defending their coach and GM in the comments because they were "experienced" hockey "insiders".

So what went wrong, are Fletcher and Wilson lying about Finger? Trying to cover up for their mistake? I don't think so - I think they genuinely believe what they are saying. They merely, to borrow a term from Lowetide, "saw him good", i.e. saw Finger play a few times when he was doing exceptionally well and assumed that to be his usual standard of play. In the case of Wilson, maybe he remembers Finger rubbing out Thornton once or twice and figures that they must have played against each other the whole game. Worst of all, the Leafs never took the time to look up the numbers and check whether their perception matched the reality.

This isn't just limited to hockey, of course, you can find these types of mistakenly false statements in all professional team sports. Fire Joe Morgan is one of the best at repeatedly catching and pointing them out (such as, for example, this one by Dusty Baker).

Professional GMs, coaches, and scouts see hockey players (and goalies) in small sample sizes, and make conclusions about them. Sometimes they glean valuable insight by watching a player, and sometimes they get it completely wrong. In an example of the latter case, you have a 28-year old forward who scored a career-high of 51 points while playing on the wing of an MVP candidate signing for $31.5 million over 7 years because of a few big goals and some gritty play (read: broken noses) during the past playoff season. Or, to use a goalie example, the spread of the perception that Manny Legace is a backup goalie and not a #1 because of his career record in 11 playoff games.

That is why I do not place a great deal of emphasis on what one individual hockey insider says about a player. For almost any player in the league you can probably find someone who will describe them in glowing terms, just because they saw them play well, or because they have ties to that player as a former teammate or coach. There is a lot more value in the league consensus on a particular player, but even that opinion is also greatly influenced by groupthink and the prevailing popular opinion - e.g. a lot of people haven't seen Brodeur play much over the last couple of seasons, but figure that he must be the best goalie in the league because everyone else says so. I'm not saying professional scouting or expert opinion has no value but merely that it is subject to biases and should be combined with a thorough statistical analysis to make sure that the sample viewed didn't include only the best or the worst of a specific player or goalie.

This also reinforces how unreliable our memories are - if Ron Wilson can misremember who Jeff Finger played against while he was coaching against him just a few months ago, then how can any fan accurately remember how, say, Grant Fuhr performed late in the third period of close games in the playoffs two decades ago? What often happens (as I speculated about Wilson) is that single plays take on disproportionate impact in shaping a player evaluation. So in the case of Fuhr, people remember a few big overtime saves from the 1987 Canada Cup or the 1988 Stanley Cup Finals and as a result always think of Fuhr as a clutch goalie. I have blogged before about how one long shot goal sunk Tony Esposito's reputation as a big-game goalie. Chris Osgood is another goalie who was stigmatized by a few weak goals against in the playoffs, and it is impossible to have an argument with a Patrick Roy hater without them repeatedly bringing up Roy accidentally dropping the puck and giving up a goal against Detroit in the 2002 playoffs.

There are biases in scouting that can lead to big mistakes, even evaluations of relatively recent performances by experienced individuals in charge of running NHL teams. As a result, in the vast majority of cases I prefer to put my trust in the numbers to find out if someone is good or if I've just seen them play well.

Thursday, July 3, 2008

2008 San Jose Sharks - Good Defence or Good Goalie?

A comment on my last post got me interested enough in looking a little deeper at the team context Evgeni Nabokov played in this year. Commentor Rob S claimed I was mistaken that the Sharks were a top defensive unit, and a lot of Shark fans tend to echo that sentiment. I admit I haven't watched the Sharks play extensively during the regular season, but I saw most of their playoff games and thought they had a strong team defence. I would be interested, though, in hearing an explanation of exactly why they are not a top defensive team, because all their team statistics are outstanding:
  • The Sharks allowed the second fewest shots against in the league this year.
  • The Sharks had the best penalty killing percentage in the league at 85.8%.
  • According to Alan Ryder, the Sharks had the 6th best shot quality rating in the league. According to Hockey Numbers, they had the 4th best.
  • A group of 3 probably replacement-level backup goalies combined for 3-2-2, 2.63, .896.
  • Every single player on the San Jose Sharks averaged less than 25 shots against per 60 minutes of even-strength ice time. Twelve of their regular players averaged less than 22 shots against per 60 (source: Behind the Net)

I think the view that the Sharks defence isn't that good is coloured by two things: the no-name defencemen and the fact that the Sharks struggled early in the season. I don't care how famous the defencemen are, just how well they play defence, and the Sharks' defence put up pretty solid numbers. From Behind the Net, four of their regulars (Rivet, Ehrhoff, Vlasic, and Murray) allowed less than 2 goals against per 60 minutes of 5-on-5 play. Also, forwards play a very important role in team defence. The Sharks not only have some very good defensive forwards (7 of them were also sub-2.00 players at 5-on-5), but they also help their defencemen by keeping the puck at the other end of the ice (every single player on the team was on for more shots for than shots against at even strength).

As for their early season struggles, The Forechecker has some defensive stats from mid-December. At that time, he ranked the Sharks 5th in shots against and 11th in shot quality against. Not terrible, but admittedly not among the best in the league either. Back then Nabokov's save percentage was up around .920, so his performance to that point did rank among the league's best. As the season went on, however, the Sharks defensive effort improved substantially while Nabokov's success rate fell dramatically. The reduced level of goaltending was obscured by the fact that the Sharks starting winning games at a much higher rate:

Oct-Dec: .919 save %, 22-12-5
Jan-Apr: .899 save %, 24-9-3

Was Nabokov the Sharks' most valuable player and the key to their defensive results? I think he might very well have been their first half MVP, but in the second half and playoffs the Sharks won because of a great all-around game, and more often than not won despite, not because of, their goaltending.

I think an interesting comparison to Nabokov, and one that shows how perception and team factors often overshadow actual results, is Roberto Luongo. Like Nabokov, Luongo had a much better first half than second half this year. However, unlike Nabokov, Luongo took a lot of heat for his second half decline, and as a result finished 7th in Vezina Trophy voting. The reason was that the Canucks got worse in the second half of the season, and so Luongo's slip became more noticeable. Here are Luongo's splits:

Oct-Dec: 18-12-3, .928, 2.03
Jan-Apr: 17-17-6, .908, 2.69

Just like Nabokov, Luongo was very valuable to his team in the first half of the season. However, the Canucks slipped in the second half because of declining play and defensive injuries, and they starting give up a higher number and difficulty of shots. Luongo's performance got worse as well, and as the team's star player his poor results were magnified in the glare of the spotlight. So Luongo probably outplayed Nabokov in both the first and second halves of the season, but simply because the Canucks were going in the opposite direction of the Sharks Nabokov ended up with the credit while Luongo got the blame.

Were the Sharks an elite defensive unit in the first half of the season? Evidence suggests that they weren't. However, taking everything, including their torrid second half, into account I think it is fair to say that the Sharks were a top defensive team overall, and that they helped produce the GAA and wins numbers that vaulted Nabokov into Vezina consideration.

Wednesday, July 2, 2008

"Goaltender wins is the most useless statistic in hockey"

So says Alan Ryder in his 2007-08 NHL season review, and no arguments here.

Of note, Ryder ranks Vokoun as the best goalie for 2008, with Brodeur #2, although it is close. I think you could have made a decent case for either one. Ryder's shot quality calculations show that New Jersey still has a pretty good defence (8th in shots against, 9th in shot quality against, 6th best in Ryder's combined "defensive index"), whereas Florida's was one of the worst in the league. Tim Thomas, J.S. Giguere, Henrik Lundqvist, and Roberto Luongo round out the top 6.

Ryder agrees with me that Nabokov's 2007-08 season wasn't really all that special. In fact, it had a lot in common with early '00s Brodeur: lots of games played on a top defensive team results in lots of wins, a low GAA, and lots of (probably undeserved) Vezina votes.