Score effects are a well known and well understood observation in hockey analytics. Essentially what score effects tell us is teams play differently depending on the score and in turn the resulting statistics are altered because of it. To keep this simple, in general teams that are leading give up more shots, but a smaller percentage of them end up as goals (they also take fewer shots but a higher percentage of them end up as goals).
Let me reiterate the main point here. When a team has a lead they effectively give up more shots but those shots are, on average, of lower quality and thus a lower percentage of them end up being goals. This effectively means when playing with a lead teams play in such a way that they boost their goalies save percentage.
This brings me to Brandon Sutter. Whenever I suggest that Brandon Sutter has an ability to boost his goalies save percentage there is always a backlash from a portion of the hockey analytics community or from those that believe in and use hockey analytics.
So bottom line: if your justification for the Sutter signing is that he suppresses GA via awesome on-ice sv%, good luck. You’ll need it. –@67sound
There are even guys that are so adamantly against the idea that players can influence save percentage they have to completely twist what I am saying into a claim that I am not making just to prove their point that the whole idea that players can influence on-ice save percentage is just an insane concept.
Mark your calendars. Today is the day Sutter was declared a better defensive forward than Kopitar and Bergeron https://t.co/rfpedIJFJk
— Lyle Kossis (@LyleKossis) August 6, 2015
The interesting thing is I am almost certain that the same people who dismiss the idea that a player can influence on-ice save percentage fully believe in the concept of score effects, including that teams generally have a higher save percentage when defending a lead.
So, my question is, if a team (and even individual players) can post higher on-ice save percentages when they are protecting a lead, why is it so unbelievable that an individual player can accomplish the same during all situational play. For example, Brandon Sutter.
|Season||Team||Sutter GA60||Team GA60||Sutter-Team GA60||Sv%RelTM|
Now, a lot of players when they are on the ice outside of specialized situations (such as protecting a lead later in the game) don’t primarily focus on defense. For these players it is understandable that they don’t have a signfiicant impact on the oppositions shooting percentage. But some players, like Sutter, do and this is why he has consistently posted better on-ice save percentages than his teammates. There are in fact very few players that play this role almost exclusively over several seasons but those that do often exhibit the ability to boost their goalies save percentage when they are on the ice.
One thing I like to do to see if a statistic exhibits pure randomness is to look at the top players and the bottom players and see if there are any commonalities in the players within the two groups. If I can identify commonalities within the best or worst of a particular stat it tells me that that stat is probably not purely random.
So, what I did was take the top and bottom 15 players in Sv%RelTM over the past 6 seasons and compare them to their ZSO%Rel (from War on Ice). The theory is a higher ZSO%Rel would indicate they are given more offensive roles and a lower ZSO%Rel would indicate they played a lesser offensive role and thus presumably a more defensive role (ZSD%Rel would have been interesting to look at as well to explicitely identify defensive roles). Here are the results.
|Rank||Player Name||Sv% RelTM||ZSO%Rel|
|Top 15 Average||1.5||-2.57|
|Bottom 15 Average||-1.4||6.14|
(The above are 5v5close stats for forwards with minimum 2500 minutes over last 6 years)
What we clearly find is that the players that have poor Sv%RelTM are generally offensive players. Only two of the 15 worst Sv%RelTM players (Chimera and Erat) had a negative ZSO%Rel indicating the majority of them played more offensive roles. Of the top Sv%RelTM players there are a number that have a positive ZSO%Rel but on average these players are getting fewer offensive assignments than the bottom 15 group. Only Stepan and Pouliot really stand out as having significantly more offensive zone starts.
Even just looking at the players names you would identify the majority of the top 15 as either solid 2-way players or more defensive specialists while the bottom 15 are more offensive oriented players.
Just as teams that are playing defensive hockey protecting a lead can boost their goalies save percentages so can players that are playing defensive hockey during play in 5v5close situations. Why this is such a surprise to people is beyond me.
Conversely, just as teams that are playing offensive hockey when trailing typically see their save percentages drop, players who are given primarily offensive assignments also see their on-ice save percentages lower than their less offensive oriented team mates.
In the end it really dumbfounds me that people can fully accept that score effects indicate that different playing styles can impact save percentage but can’t accept that players have an ability to impact save percentages if they are given primarily offensive or defensive roles.
This is not unlike the debates I have had with people about zone starts and face off winning percentage. There are many who have believed that zone starts had a significant impact on a players statistics while at the same time claiming face off win percentage had little to no value. This really makes zero sense to me. You simply can’t on the one hand claim that being on the ice for more (or fewer) offensive faceoffs than defensive faceoffs will have a significant impact on a players stats while on the other hand claim that winning those offensive or defensive faceoffs doesn’t really matter at all. It makes no logical sense to me and yet many leading members of the hockey analytics community believed that for years. Happily it is slowly changing and people are starting to realize zone starts don’t dramatically impact a players stats.
Similarly, there was a time where the hockey analytics community largely dismissed the notion of shot quality. Happily this is changing too and more and more people are accepting that shot quality exists and more people are researching shot quality and how to best quantify it, even if the results are still far from perfect.
It is time that we get past the old school hockey analytics belief that players can’t impact save percentage and start asking who does and how they do it.