New Puck Prospectus piece: A basic goaltending model

by Hawerchuk on Oct 7, 2009 4:27 PM EDT Comment 9 comments

I wanted to start from the basics on goaltending and estimate how many wins a goaltender can be worth to his team. I looked at 5-on-5 and 4-on-5 save percentage only (no shot "quality" adjustments), relative to replacement, and assumed every goalie was average at 5-on-4 (given that they face so few shots). I'll make the appropriate adjustments as I expand on this topic.

The key insight is that the absolute best goaltender in the league in terms of performance over the last two seasons - Tim Thomas - was worth about 8 wins above replacement per 82 games. Goalies never play 82 games, so their real values are slightly lower than that, but this puts us in the ballpark. You may disagree with where I put replacement level, but if it was any higher, Miikka Kiprusoff would be right there, and I don't think that's quite right.

Here's the piece:

http://www.puckprospectus.com/article.php?articleid=298.

0 recs | Comment 9 comments |

Story-email Email Printer Print

Comments

I’m a little confused at where you’ve set replacement level. When I checked out the numbers on nhl.com there was only one goalie in the league whose EV save percentage exceeded the “replacement level” of .938. Admittedly, this also includes 4v4 hockey. Is the 5v5 save percentage really that much higher than total EV save percentage?

by Scott Reynolds on Oct 7, 2009 7:32 PM EDT reply actions

I am using the European save percentage = GA/(GA + SOGA + MISSA). It’s a more accurate way to look at shooting – we already do it for shootouts (the well-known 33% shooting percentage includes missed shots.) So I want to be consistent.

by Hawerchuk on Oct 7, 2009 7:39 PM EDT up reply actions

Ah, that makes sense. It seems like it’s a bit curious to use for save percentage though. Are there certain goalies that benefit much more than others with the inclusion of missed shots?

by Scott Reynolds on Oct 7, 2009 10:37 PM EDT up reply actions

I think save and shot percentage need to be symmetric. Plus we don’t want to ignore crossbars and goalposts. Most goaltenders are clustered between 25 and 31% missed shots over the last two seasons, though Khabibulin was at 21% (and so benefits the least.)

by Hawerchuk on Oct 8, 2009 2:05 AM EDT up reply actions

cool article

the numbers in that list… are the values determined by using each goalie’s sv% and shot against per game and then assuming every goalie played the same number of games? OR are they determined by using each goalie’s sv% and then assuming they all faced the same number of shots per game and played the same number of games? or is it some different combination of that?

by sunnymehta.com on Oct 7, 2009 8:02 PM EDT reply actions

It is just a measure of how they actually performed: DELTA_GOALS = (REPL_55SVPCT – ACTUAL_55SVPCT)55_SHOTS + (REPL_45SVPCT – ACTUAL_45SVPCT)45_SHOTS

Then multiply by GP/82, and divide by 6 to get Wins. Very coarse metric – this was one for the “Welcome to Hockey Analysis” tag.

by Hawerchuk on Oct 7, 2009 8:11 PM EDT up reply actions

Damn stars (*) turn into bold!

by Hawerchuk on Oct 7, 2009 8:11 PM EDT up reply actions

You may disagree with where I put replacement level, but if it was any higher, Miikka Kiprusoff would be right there, and I don’t think that’s quite right.

Kiprusoff was, I believe, 31st in the NHL last year in GAA and SVP: that sounds pretty close to replacement-level to me. Not saying I disagree with where your line is, per se, just that this isn’t a good reason in and of itself.

SNN Sports - A theoretical Oilers blog (i.e. theoretically, I write stuff there). Link now 100% less broken.

by Doogie2K on Oct 8, 2009 3:12 PM EDT reply actions