Analyzing player usage

[BoredBrandonPridham]

There's a ton of discussion that floats around about player's production and the statistics that go along with that. Typically when we have 2 people that get into a debate about some player being "better" than another, we bring up usage and context, and rightfully so: 5v5, QoT, QoC, Zone Starts, etc...

For 5v5, it seems we slice the data -- I personally haven't seen anyone do something like an "adjustment" for 5v5 v.s. 5v4.

For Zone Starts, I've heard of Corsi adjustments there that essentially remove any Corsi events that happen some time period after the face-off (10s?), as the impact of ZS tends to have little impact after that. I'm not sure I've come across that adjustment in some time now.

And when it comes to Q0T and Q0C, I'm not sure I've seen either slicing nor adjustment for those aspects. In most sites I've seen, they give you production and some will also provide this as a separate tidbit of information, but I think many (including myself) are confused on how this actually factors into comparison.

Is lack of available adjustment to usage a result of a limit of my (or our collective) understanding on player usage and its effect? When trying to compare 2 players that have substantially different usage, say one that gets 60% DZS high QoC and one that gets 10% DZS low QoC, should we just bucket them separately rather than try and debate an apple and an orange, similar to 5v5 v.s. 5v4 production?

Are there any other stats or research available that I'm missing that do a good job in evaluating players that take into consideration their usage?

Click to read more...

 

[NHL WAR]

In the long run QoC has little impact on an individual's stats. While OZS% has a huge range from 30%-70%, QoT usually ranges from around 5% below average to 5% above average and QoC even less than that. Since the differences between players are insignificant, I wouldn't lose any sleep over QoC stats such as CF. QoC or xGF.QoC. Quickest and easiest way to get a little context is to check out how they perform relative to their team and how much they start in the o-zone. When working on the Corsi/ xGF portion of W.A.R, I have had the weights ranked: how they did compared to average> compared to QoT>> compared to ZS>> compared to QoC. I currently am not even using QoC as the data from Evolving Hockey doesn't list it and at this point it isn't worth the time of formatting multiple data sets together. Hope at least some of this helped, but there currently is no perfectly adjusted stat I could point you towards.

 

[BoredBrandonPridham]

In the long run QoC has little impact on an individual's stats. While OZS% has a huge range from 30%-70%, QoT usually ranges from around 5% below average to 5% above average and QoC even less than that. Since the differences between players are insignificant, I wouldn't lose any sleep over QoC stats such as CF. QoC or xGF.QoC. Quickest and easiest way to get a little context is to check out how they perform relative to their team and how much they start in the o-zone. When working on the Corsi/ xGF portion of W.A.R, I have had the weights ranked: how they did compared to average> compared to QoT>> compared to ZS>> compared to QoC. I currently am not even using QoC as the data from Evolving Hockey doesn't list it and at this point it isn't worth the time of formatting multiple data sets together. Hope at least some of this helped, but there currently is no perfectly adjusted stat I could point you towards.

Thanks, that is helpful. I’m seeing a trend that the most relied upon contextual factors are the ones that the players’ own coach has the most control over:

OZ starts vs DZ: A coach always has control over the choice of player for OZ but DZ is muddied by icings?

QoT vs QoC: A coach has most control over your teammates on the ice, but QoC is muddied by opponent’s coaching choice.

I think I’m a bit surprised that if coaching is playing such a big part in how statistical models are built, that I don’t see more analysis around what objectives a player may actually be getting tasked (it’s not always just taking shots and scoring goals is it? Could it not just be getting a weak shot into the chest of the goalie to stop play so your first line can come on?) and if they are successful at those objectives.

 

[NHL WAR]

Thanks, that is helpful. I’m seeing a trend that the most relied upon contextual factors are the ones that the players’ own coach has the most control over:

OZ starts vs DZ: A coach always has control over the choice of player for OZ but DZ is muddied by icings?

QoT vs QoC: A coach has most control over your teammates on the ice, but QoC is muddied by opponent’s coaching choice.

I think I’m a bit surprised that if coaching is playing such a big part in how statistical models are built, that I don’t see more analysis around what objectives a player may actually be getting tasked (it’s not always just taking shots and scoring goals is it? Could it not just be getting a weak shot into the chest of the goalie to stop play so your first line can come on?) and if they are successful at those objectives.

You raise good points. As for the objectives, I think a big factor in most of the analysis being towards scoring and shots is that it is just easier. You can just copy and paste some stats into a spreadsheet and do some analysis. Whereas, to use your example of a bottom 6 player throwing it on net to get a whistle and a line change, I guess you would:
- search through play by play data for occurances of a player on the end of a long shift or with ATOI <12 that took a shot resulting in a whistle
-see if it was low percentage shot
-see if a new line was out for the o-zone draw
Or, watching video and manually tracking this.
I'm sure unique analysis like this could be done to an extent though, and it would be interesting.

 

[Golden Puppers]

Thanks, that is helpful. I’m seeing a trend that the most relied upon contextual factors are the ones that the players’ own coach has the most control over:

OZ starts vs DZ: A coach always has control over the choice of player for OZ but DZ is muddied by icings?

QoT vs QoC: A coach has most control over your teammates on the ice, but QoC is muddied by opponent’s coaching choice.

I think I’m a bit surprised that if coaching is playing such a big part in how statistical models are built, that I don’t see more analysis around what objectives a player may actually be getting tasked (it’s not always just taking shots and scoring goals is it? Could it not just be getting a weak shot into the chest of the goalie to stop play so your first line can come on?) and if they are successful at those objectives.

You can't start in the defensive zone on an icing, because the defensive team can't make line changes after an icing.

As for the QoC. the home team controls match ups via last change so you could argue for 41 games you can control the opposition (to an extent).

 

[BoredBrandonPridham]

You can't start in the defensive zone on an icing, because the defensive team can't make line changes after an icing.

Are you implying that the ZS stats we use day to day don’t include face offs in the defensive zone as a result of an icing a “start”? Or in general, if the play stops for some reason, and a player who was already on the ice stays, that’s not added up as a “start” for that player?

 

[Golden Puppers]

Are you implying that the ZS stats we use day to day don’t include face offs in the defensive zone as a result of an icing a “start”? Or in general, if the play stops for some reason, and a player who was already on the ice stays, that’s not added up as a “start” for that player?

Right on the Bolded. Starts only count that first faceoff a player comes onto the ice for. Therefore, you can't 'start' a shift in the defensive zone off an icing because you can't change.

Zone starts are meant to describe a players usage. If it counted all faceoffs, it would be flawed because good players would get more offensive zone faceoffs because they are driving play and getting shots.

You can go to Natural Stat Trick and see how many total faceoffs a player had in a particular zone as opposed to just the ones they started in.