Crashing the Crease: How Many Wins Does a Low Save Percentage Cost?

Ilya Kovalchuk #17 of the New Jersey Devils scores a shorthanded goal against Ilya Bryzgalov #30 of the Philadelphia Flyers at 17:18 of the first period at the Prudential Center on March 13, 2013 in Newark, New Jersey.The Devils defeated the Flyers 5-2.
Image Courtesy of Bruce Bennett/Getty Images North America

At the time of my last article, I had a thought about save percentage. We often harp on it, but very rarely do we think of it in practical, dollars-and-cents terms. If a goalie A has a .900 save percentage and goalie B has a .910, for example, what does that tell you aside from goalie B is stopping more shots? A simple way to look at it is that goalie A will let up ten out of every 100 shots he sees, while goalie B only lets up nine.

Big deal, right?

Prompted by Eric T.’s article over at Broad Street Hockey, I decided to try and quantify what these goals mean and why you really want to maximize your starter(s)’s efficacy over a season. To do this, we’ll need to look at some more numbers (apologies in advance) in two parts…

 

Part One


Assuming a starting goalie sees anywhere from 1,500 to 2,000 shots over a normal season, that .010 difference in save percentage amounts to 15 to 20 goals over a season. If one could ideally take those 15 to 20 recovered goals and apply them strictly to losses, which I fully admit to be an ideal assumption, we have an interesting little exercise on our hands.

Given that, let’s carry out our math exercise using real numbers from this year. Since we have a tendency to be a little Ilya Bryzgalov-centric, I’ll look at the 36 games he’s appeared in so far. Of those 36 games, 19 were losses (ouch). The breakdown for those losses can be seen below:

You’ll see that 32% of the losses were by one goal, including overtime. In other words, assuming you could magically take away goals, those could’ve been wins. In addition, you have two-goal losses that came as a result of an empty net goal. These will count as one-goal losses.

One caveat here is that, if the Flyers lost in a shutout, no amount of additional saves would provide an outright win in that scenario. Furthermore, an overtime loss (OTL) really only needs to count as one goal, because one additional save in regulation would’ve been the difference. With that understanding, let’s adjust the loss totals to the following, which will now represent our actual lost opportunities, so to speak:

Okay, so there are our lost opportunities. How much does it “cost” (i.e. how many additional saves would we need) to turn each one of these into a win? Well, see the following:

I’m going to ignore the cost associated with “fixing” a three-or-more goal loss, because that would require an unrealistic overall save percentage adjustment. Part of the premise here is that we’re talking about a clutch save or two in a tight game, not just changing history altogether.

Part one is now complete. We basically know how to apply any additional saves we get. On to part two, which is the fun part of looking at the material impact of an improved save percentage.

 

Part Two


As I sit here, eating a pretzel and drinking some Wawa lemonade, Bryz has a save percentage of .896 (I’ll relent and just write it as a decimal) and has faced 953 shots. If you’re a calculator, you already know that this means he’s surrendered 99 goals. There are a few different tiers of performance I’d like to look at using this number set:

  • Good, representing .905
  • Better, which we’ll call .910
  • God-like, which we’ll define as .920 given the Flyers’ “defense”

Taking the same shot count and applying the above hypothetical performance tiers, we get the following table, replete with additional saves to sprinkle over the mounting losses (note: I let Excel round up or down, as I can’t account for half saves…):

Even over the course of a shortened season, the slightest increase in save percentage has a noticeable effect on goals allowed. Apologies for jumping around, but allow me to list out all of the losses involving Bryz this year in order to better explain where we’re going:

Looking at the above, you can see each game (including ones he was pulled from), whether it was a shutout loss and/or an overtime loss, whether an empty net (EN) goal was scored (which reduces the cost by 1), as well as the total “cost” to recover this loss. Using our original cost assumptions, let’s start taking back some goals to increase the win count. Below, you’ll see numbers highlighted in green. These are recovered saves being paid back to reduce the goals against for a given game, or, put simply, these losses are now wins and it “cost” the number in the cell.

Since I know someone will say, “Nice idea, but it’s optimistic to assume that all of the additional saves will apply to losses,” I’ve run the numbers with 100% of the the “recovered” saves, as shown above, as well as 50% of the “recovered” saves going to losses, shown below:

What you see here on the right-hand side is that, even with the “worst case” scenario (out of those run) of 50% of the additional saves going to losses and a mere .009 increase in save percentage from .896 to .905, the Flyers are suddenly sitting at 41 points. Looking at the current standings below, this puts them one point back of the damn Winnipeg Jets and one up on the New Jersey Devils, with two games in hand on the Jets and one on the Devils. In other words, they’re in the thick of the playoff race.

Ignoring whether you think a barely-made eighth seed is worth it relative to this year’s draft class, you can see why goaltender is such a money position in sports. Using that “worst case” scenario again, if your keeper makes those four saves in critical overtime losses or one-goal losses, you suddenly have a guaranteed additional two home playoff games, minimum. I don’t even need to point to 2009-10 for further support. Gates, concessions, parking… everything riding on four saves out of 953 shots (to date).

There’s an inherent understanding that save percentage is likely the best non-advanced stat to account for goaltender performance. We look to it constantly to evaluate goalies, and what I was ultimately hoping to explain with this simulation is how it impacts games, teams and entire organizations. This analysis is not meant to be a condemnation of Bryz. I maintain he’s been hung out to dry more than an old bath towel this year. The intent, however, is to show, at a slightly deeper level, why that .896 just feels bad, and, ultimately, how badly it really impacts everything.

Feel free to reach me via Twitter (@HeyItsBrenno) with any questions, comments, etc.

  • wheineman

    Could you break down how many penalties have cost the team a win or at least a point this season? I know that the power play has been great, but how many one goal games did they lose where they gave up a short handed goal? Who committed the penalty the most in those cases?