Thursday, June 24, 2010

Goals Against Per 30 Shots

I'm always looking out for different ways to express goalie performance, and ran across an interesting one that I'd never seen before while going through the archives at Tom Tango's Inside the Book blog. He wrote a brief Wall Street Journal article that expressed goals against average in terms of goals against per 30 shots, rather than goals against per 60 minutes of play. That gives a metric that is pretty much just a translation of save percentage (GA/30 = (1 - Sv%) * 30), but the advantage is that it results in figures that look like goals against averages. GAAs are more familiar and intuitive for most people to understand than save percentages, and allow a better sense of the actual difference between goalie performance on a per-game basis.

There is one slight tweak I would suggest to Tango's number, and that is to adjust for special teams. I'd suggest calculating even strength goals against per 24 shots and goals against per 6 shots on the penalty kill, which reflects the typical 80/20 split between non-PK and PK shots, and then adding those two numbers together to get a special teams adjusted goals against per 30 shots number.

Here are the league's top 20 goalies last season based on this metric (min. 35 starts):

1. Tuukka Rask, Boston: 2.17
2. Ryan Miller, Buffalo: 2.21
3. Jaroslav Halak, Montreal: 2.26
4. Evgeni Nabokov, San Jose: 2.34
5. Tomas Vokoun, Florida: 2.36
5. Ilya Bryzgalov, Phoenix: 2.36
7. Jimmy Howard, Detroit: 2.37
8. Henrik Lundqvist, N.Y. Rangers: 2.41
9. Jonas Hiller, Anaheim: 2.44
10. Miikka Kiprusoff, Calgary: 2.45
11. Craig Anderson, Colorado: 2.50
12. Cam Ward, Carolina: 2.53
13. Tim Thomas, Boston: 2.57
14. Chris Mason, St. Louis: 2.60
15. Johan Hedberg, Atlanta: 2.61
16. Martin Brodeur, New Jersey: 2.63
17. Roberto Luongo, Vancouver: 2.64
18. Marty Turco, Dallas: 2.65
19. Antti Niemi, Chicago: 2.67
20. Jose Theodore, Washington: 2.68

I think by the majority of measures Miller was a deserving Vezina winner and First Team All-Star, so kudos to the voters this year for getting it right.

6 comments:

Derick said...

This is pretty cool.

I've often imagined evaluating a goalie's season with a point system where, say, 1 point is awarded for a save and 8 points are taken away for a goal. It will roughly correlate to save percentage, but goalies who have played more will have a higher score. Perhaps it could be adjusted for even strength and penalty kill shots as well.

Have you ever researched what the optimal range to go back is when estimating a goalie (or skater's) performance the next year?

For instance I imagine a three year sample size is more accurate than a one year sample size, but that a ten year sample size would be worse than either because of all that's changed.

The Puck Stops Here said...

I think it would be interesting to hear your thoughts on Ilya Bryzgalov making second team all star and coming fifth in the Hart Trophy voting.

Matteau the Magic Wrap-Around said...

Great little piece. Best part is how far down Brodeur was.

neilshyminsky said...

I'm wondering, is it fair to scale shots like that?

For instance - what's the correlation between the # of shots faced and goals? Or to reword - do goalies who face more shots than average tend to also allow more goals per shot than average? And do goalies who face more time on the power-play tend to allow more goals per shot than the average?

Because if these correlations exist, then this GAA/30 would need to account for them, right? It's not really fair to say that this figure tells us how the goalie would do with 24 even-strength shots and 6 power-play shots against, since it's just scaling their existing numbers - numbers that don't control the 'shots-over-average' penalty, for instance. (If one exists, that is.)

The Contrarian Goaltender said...

I think it is fair to scale shots, because I don't think there is any significant relationship between shots and save percentage, and I've looked at that problem is a bunch of different ways (just search the archives here if you want to read more).

I don't know whether there are any specific effects on the power play alone. That might be something to look at, but it is a lot more difficult problem data-wise than looking at the overall issue.

Anonymous said...

I just made up this statistic my self. Needless to say, I kind of like it better than save percentage, even though it's the same thing.

It would be even better to combine the two.

A GA/60m of 2.20 and a GAA/30s of 2.10 would be 2.15.

PK time, obviously, is tougher but nobody accounts for that with GAA or Sv% either. Life isn't fair. If you're team is on the PK a lot, you're just not going to get the Vezina easily.