Four Goalies That Handle A Heavy Load, And One That Doesn't
Today's analysis was inspired by a comment left here by Magicpie over at Nucksblog, who wondered whether some NHL goaltenders perform better when they face more shots. That's certainly part of the Common Wisdom of hockey, and when asked, just about every goaltender admits that facing frequent pressure helps them stay focused. But in the final analysis, does this really hold true?
I went back through the game-by-game details for #1 goaltenders over the last two full seasons, to find out which guys fare better facing a ton of shots, and which ones wilt under the pressure. The measurements presented in the graphs below represent Save Percentage for each game, measured against the average Time Between Shots that the goalie faced, with a trendline in bold summarizing that relationship across an entire season. A line that slopes downward to the right shows that as the Time Between Shots becomes shorter, the Save Percentage tends to increase. An upward sloping line indicates a Save Percentage that gets better when more time passes between shots.
For screening purposes, I'm looking at goalies who played at least 41 games in both the 2005-6 and 2006-7 seasons. What we end up with are the following candidates: Roberto Luongo, Miikka Kiprusoff, Martin Brodeur, Marty Turco, Rick DiPietro, Tomas Vokoun, Curtis Joseph, J.S. Giguere, Olaf Kolzig, Manny Fernandez, Henrik Lundqvist, Manny Legace, Nikolai Khabibulin, Marc-Andre Fleury, Ed Belfour, Marc Denis, Ryan Miller, Antero Niittymaki, Evgeni Nabokov, Dominik Hasek, and Dwayne Roloson.
Now remember, this covers 2005-6 and 2006-7, so I'm going to eliminate a few more names from this list because due to trade, injury or age, this year they are clearly no longer considered #1 goalies: Curtis Joseph, Manny Fernandez, Ed Belfour, Antero Niittymaki, Marc Denis, and Marc-Andre Fleury. This leaves us with 15 netminders to consider. There are certainly players who have ascended to the role of #1 (Chris Mason, Nicklas Backstrom, etc.) but I don't have the two full seasons worth of history to go through there, so I'm leaving them out as well. I'm also not digging into 2007-8 data, as the NHL changed their Game Summary format mid-stream, so further work is needed to derive that information.
The first interesting result is that for most goaltenders, there wasn't much of a link between frequency of shots and save percentage. Among the goaltenders that failed to show any strong trend in either direction were Miikka Kiprusoff, Martin Brodeur, Rick DiPietro, Tomas Vokoun, J.S. Giguere, Henrik Lundqvist, Manny Legace, Ryan Miller, Evgeni Nabokov, and Dominik Hasek. Also, when I looked at the combined work of all goaltenders across the league, there was no relationship found. Among the #1's, however, there were a few noteworthy expections...
THANK YOU SIR, MAY I HAVE ANOTHER?
SLOW IT DOWN, ALREADY!
Labels: goaltending, statistics
7 Comments:
I guess it isn't an overall trend. Maybe in Luongo's case it's the team in front of him that causes this. Our backup last year showed the same pattern: 0.931 in games where he faced over 25 shots, 0.872 in games where he faced less than 25. This year's backup, same story. 0.927 in games facing more than 25 shots, 0.819 in games facing less than 25. Both of those sets of stats are only for about 10 games played so I guess that's not the greatest sample size in the world. But anyway, it looks like there's something about the Canucks themselves that makes goalies play better when they face more shots.
Anyway, interesting stuff, thanks a lot for taking a look at the issue.
I hate to bitch but your responce variable goes on the y axis.
You're right, mogen_david, and as an approximation of such feel free to lay down on your side while reading this post.
Ah, that helps. I like the stripes in the data. 0, 1, 2, 3 goals allowed?
Yeah, those stripes do correspond to goals allowed, it makes for an interesting visual pattern. Tomorrow I'll throw up the graph for all goaltenders, all games across both seasons, and you'll see it even more clearly.
When you have as big a spread in your data as you do, fitting a line to it is nearly meaningless. I imagine the residuals (R^2) are huge.
I am not sure you can have any confidence in the conclusions that come from that.
I'm not sure it's just a visual effect. Average time between events in a finite time period has an upper limit (your breaking a 60 minute game into time periods if their are 2 shots the maximum time between goals is 60 minutes while 3 shots can be at most 20 minutes apart). Since Save percentage is goals per shot per game the two variable are not mathmatically independent. The "limit" I mentioned above should create a natural data hollow in the upper right of your graph. I would expect your trend lines to project downward (intresting that so few don't). I'm not sure how strong this trend should be since the relationship isn't simple. It might be intresting to look at the problem as a logistic regression with goals being your responce, shots your oservation and time between shots as your predictor. Trouble is I don't expect a linear relationship near zero (think rebounds and time for the goalie to reset). I'll just sit here and critisize while you do all the good work.
Post a Comment
Links to this post:
Create a Link
<< Home