Fans are drawn towards statistical hockey analysis with goaltending more so than with skaters, because it seems easier to separate a goalie's contributions from the rest of the team. It seems quite natural to give the goalie the credit for making a save, but is it always appropriate?
Last week we showed how Alan Ryder's Player Contributions statistic can help us evaluate skaters. This week we'll use the same statistic to show how we can separate the truly great goalies from the good goalies who are simply playing behind great defenses.
Toskala, Kiprusoff and Turco
Toronto's Vesa Toskala, Dallas' Marty Turco and Calgary's Miikka Kiprusoff are all NHL workhorses that were punished with GVTs below replacement value of -18.0, -9.3 and -1.7, respectively. And why not? With dismal save percentages of .891, .898 and .903, why shouldn't we conclude that these three teams would have been better off with practically any waiver-wire pick-up instead?
Their PC scores tell a different story, awarding respectable scores of 70.2, 122.9 and 143.5, because their poor save percentages are not entirely their fault. Based on the number of shots allowed, and the quality of those shots -- which is based on the location from which the shot was taken -- Toronto was terrible defensively, 2nd worst to only the Florida Panthers. Maple Leaf netminders like Toskala generally faced 30.3 shots per game, but that's based on a shot quality of 1.124, so they were actually premium shots, and it's as if they were facing over 34. That will definitely make Toskala's numbers look worse than they really are.
Calgary and Dallas were both in the same boat, though they allowed fewer shots than Toronto, the quality was almost as high. Dallas gave up only 28.1 shots per game, but they were of the 2nd highest quality in the league, keeping Turco very busy. Los Angeles is another example of a team that gave up relatively few shots, but of higher quality.
Some teams, like the Atlanta Thrashers and the Philadelphia Flyers, gave up lots of shots, but they were mostly from the outside, making their goalies look better than they actually were if you judge them solely on a save percentage basis. Perhaps we should be judging Martin Biron and Kari Lehtonen more harshly and giving Vesa Toskala at least a little bit of a break.
PC scores can help us explain why Craig Anderson's hot start shouldn't be much of a surprise. Last season Anderson played behind the worst defensive team in the league, as Florida gave up 34.7 shots at a quality of 1.042. Now he's in Colorado, where last season they gave up only 29.0 shots at a quality of just 0.968, and now it's like he's facing only 28 shots a game instead of over 36. That makes a huge difference!
Statistically we can make that adjustment by multiplying Anderson's shots faced last season by 1.042 and working out his new team-neutral save percentage at 0.927. If we multiply his shots this season in Colorado by the quality factor of 0.968 and work out his new team-neutral save percentage, we get almost an exact match: 0.926. So the real surprise is that Anderson actually played a little bit better last season in Florida!
In his excellent 2009 season summary Alan Ryder explained that a goalie that faces few quality shots can't make a very high contribution. The more quality shots a goalie faces, the larger a contribution he can make. That's why a goalie on a great defensive team isn't really in a position to make a big contribution.
Consider the following four sets of goaltenders, each one with a goalie that seems to have a clear statistical edge over the other, and yet with a seemingly contradictory PC score.
GP: Minutes played divided by 60
SV%: Over-all save percentage
ESSV%: Save percentage at even-strength
SA/G: Shots against per game
Goalie Team GP SV% ESSV% SA/G GVT PC
Henrik Lundqvist NY Rangers 69.2 .916 .920 29.0 18.8 248
Niklas Backstrom Minnesota 68.1 .923 .923 30.2 27.7 222
Steve Mason Columbus 61.1 .916 .925 27.2 16.2 169
Marc-Andre Fleury Pittsburgh 60.7 .912 .922 30.5 11.1 180
Dwayne Roloson Edmonton 59.9 .915 .926 32.6 14.2 211
Ryan Miller Buffalo 57.4 .918 .927 30.9 18.2 152
Jonathan Quick Los Angeles 41.6 .914 .926 28.9 9.9 137
Jonas Hiller Anaheim 41.4 .919 .934 29.4 14.9 138
Lundqvist vs Backstrom
On paper, Niklas Backstrom certainly looks like he had the better season, facing more shots per game and yet stopping a higher percentage than Lundqvist, both at even-strength and overall. However, this ignores the fact that Minnesota was far better defensively than the New York Rangers were last season, and we hardly need Ryder's shot quality statistic to tell us that (0.967 vs 1.022 for those that do). If you're still uncertain of that, compare them to their respective backups, Josh Harding's save percentage was actually better than Backstrom's (.929) while Stephen Valiquette's was much worse than Lundqvist's (.907).
Under a new coaching staff, Minnesota's defense has taken a turn for the worse, and as a result Backstrom's save percentage has slid to .909 this season while Lundqvist is holding at .918. PC is showing us statistically what we already knew through observation -- that Henrik Lundqvist is currently contributing more than Niklas Backstrom.
Mason vs Fleury
If we only had the traditional stats upon which to rely, we might conclude that Steve Mason is better than Marc-Andre Fleury, since he posted a better save percentage, both overall and at even-strength last season. Since most people generally consider Fleury to be the better goalie, including the coaches and staff of the Canadian Olympic team, we might get discouraged about the value of statistics for hockey analysis in painting accurate pictures.
The truth is that Fleury was playing behind one of the weaker defensive teams in the league, making his performance look far less effective than it actually was. We can already see that he faced 3.3 more shots per game than Mason, and now we can see that they were also of higher quality, 1.042 to .978. Mason's slow start this year is making the difference between the two a little more obvious, but thanks to PC we already knew about it last season.
Roloson vs Miller
Ryan Miller is a great goalie and he's enjoying his start in Buffalo a lot more than Dwayne Roloson is enjoying his new Islanders uniform. Last season Miller was slightly ahead of Roloson in save percentage, but that ignores the fact that Buffalo had a far better defense than Edmonton. As a Sabre, Miller faced 1.7 fewer shots, and they were of the unbelievably low quality of .913. As a team, Buffalo did a great job keeping the shots to a minimum, and from the outside, while the Oilers let their opponents pummel Roloson mercilessly (shot quality = 1.031). Roloson obviously has a few great seasons left, but it's a pity they're being wasted on non-contenders.
Quick vs Hiller
Jonas Hiller is certainly our golden child at Puck Prospectus, being ambitiously promoted as a possible Vezina contender during our preseason preview. He's off to a great start, clearly much better than Jonathan Quick, but the numbers are perhaps being a little unfair and misleading us about the true value of California's netminders.
While looking at traditional statistics, it certainly looks like Hiller is the better goalie, but we have to remember that he's playing in Anaheim, a team that's at least above average defensively. In contrast, Quick is struggling on a team of young players who haven't yet developed the same consistency defensively. The shot quality difference between the two teams is over 10%: 1.069 in Los Angeles and 0.954 in Anaheim. Stopping pucks for the Ducks isn't the same thing as stopping them as a King.
It's easy to get discouraged with hockey statistics when you're using ones that leave out key pieces of the picture. If you don't consider the defense behind which a goalie is playing, then you'll prefer Chris Mason over Marc-Andre Fleury, and you'll get surprised by a Craig Anderson every year. Whenever the numbers don't line up with what you believe to be true, turn to more detailed sources, like Alan Ryder's Player Contributions for example, to get a more complete picture.
Robert Vollman is an author of Hockey Prospectus.
You can contact Robert by clicking here or click here to see Robert's other articles.