View Single Post
04-01-2013, 09:00 PM
Doctor No
Registered User
Doctor No's Avatar
Join Date: Oct 2005
Posts: 6,857
vCash: 500
Goaltender Game-to-game Consistency

Something that I've been wanting to do for awhile, but is now feasible with all of my information databased...

It's reasonably obvious that, given two goaltenders with equal save percentage, the one with more shutouts is also going to be the one with more stinkers. How can we best quantify the variability of goaltender performance with numbers?

Quality Start Percentage is a good start, but it's a binary metric - and there's a very thin line between "quality" and "non-quality". If you have six straight starts that just cross the line, then you're a solid quality goaltender. Here's the definition from Hockey Prospectus:

An idea blatantly lifted from baseball, Quality Starts were one of Hockey Prospectus' first contributions back in 2009 as a measure of whether a goaltender "gave his team a chance to win". In order to record a Quality Start, the starting goalie must stop at least a league average number of shots (typically 91.3% prior to 2009-10, and 91.7% since), or play at least as well as a replacement-level goalie (88.5%) while allowing two goals or fewer.

It's a good definition, but two things that bother me with it: the baseline save percentage is the same whether you're facing the Pittsburgh Penguins or the Colorado Avalanche (I can say that as an Avs fan - what I'm having a hard time saying lately is why am I an Avs fan?). And, like I said above, I wanted something with more reasonable degrees of distinction (and made sense intuitively).

I also wanted something that would translate back to all of the seasons for which I have game-level data (so no even-strength save percentage, although it might be great for this).

Doctor No is offline   Reply With Quote