Kukla's Korner

The Puck Stops Here

Top 20 Players By Team Adjusted Corsi

In my summer sabermetrics and hockey posts, I am looking at making sense of individual Corsi ratings.  These are the difference between a team's attempted shots and their opponents while a player is on the ice in five on five situations.  I posted the top 20 players (due to a tie it's actually a 21 player list) last season and observed that this group of players come from only seven teams and all of them are good puck possession teams.  The top raw Corsi players play on good puck possession teams.  That probably isn't a surprise.  In order to make it more of an individual rating we must adjust for the team in which a player plays.  That will not be the final step in making Corsi a meaningful individual rating because not all players play under the same circumstances.  Even on the same team some players play in more offensive roles and some play in more defensive roles and that should affect puck possession while they are on the ice. 

In order to make a team adjustment, we first must know team Corsi ratings.  These can be used to find a baseline for a player on a given team.  The baseline is the team Corsi divided by five (since five position players are on the ice at a given time).  This baseline is subtracted from each player.  In order to ensure that the adjustment is meaningful only players who played at least 30 games in the lockout shortened 48 game season are included.

Here are the top 20 players by team adjusted Corsi in 2013:

Top 20 Players In 2013 By Team Adjusted Corsi

1 Daniel Sedin Vancouver +258.8
2 Henrik Sedin Vancouver +248.8
3 Justin Williams Los Angeles +240.0
4 Jake Muzzin Los Angeles +214.0
5 Anze Kopitar Los Angeles +209.0
6 Tyler Seguin Boston +204.6
7 Max Pacioretty Montreal +188.8
8 Patrice Bergeron Boston +185.6
9 Sidney Crosby Pittsburgh +167.6
10 David Clarkson New Jersey +166.6
11 Mikko Koivu Minnesota +162.8
12 Alexandre Burrows Vancouver +156.8
13 Brad Marchand Boston +156.6
14 Chris Kunitz Pittsburgh +155.6
15 Dustin Brown Los Angeles +155.0
15 Lubomir Visnovsky NY Islanders +155.0
17 Anton Stralman NY Rangers +152.2
18 Zach Parise Minnesota +147.8
19 Dan Boyle San Jose +147.0
20 Drew Shore Florida +142.2

This list of 20 players comes from 11 teams.  This is more inclusive than the 7 teams for 21 players in the top 20 raw Corsis list.  It isn't reasonable to ever expect a list with 20 teams represented in 20 players because maybe groups of teammates on this list were often linemates and thus if one makes the list then the other likely will too.  The players on this list and not on the raw Corsi list include Sidney Crosby, Mikko Koivu, Zach Parise and Chris Kunitz.  Those dropped include Marek Zidlicky, David Desharnais, Patrik Elias and Nathan Horton.  I think it is clear that this group of players is a stronger group - although not necessarily by a huge margin.  This is a step in the right direction but not a final step.

At the top of the list we have Daniel and Henrik Sedin.  These are two players who base their game around cycling the puck deep in the opponent's offensive zone and play on a team that maximizes their offensive zone starts.  They probably are not surprising players to be at the top of the list.  It is evidence that they are good players, but it is not reasonable to suggest they are the best players in the league because of their offensive roles - this will increase their Corsi.

Adjusting individual Corsi ratings for the team the player plays is the first step toward making this rating an individual value.
 

Filed in: | The Puck Stops Here | Permalink
 

Comments

Nathan's avatar

(These are not rhetorical…)

Is dividing by 5 the best way to determine the baseline for a player on a given team? It seems like there needs to be a way to weight by position (at least forward vs. defense) and/or adjust to a rate stat rather than a cumulative one.

How valuable is subtracting the team baseline? Does this accidentally victimize players on strong puck possession teams? Perhaps there is an issue in that it seems the baseline simply determines what the average player on a given team’s rating would be, and what we really need to derive a more meaningful number is to know what replacement level is. Measuring a player compared to his team average doesn’t give that player proper credit for his contributions to that average, and it also doesn’t weight ice time properly. If we had a replacement level Corsi, that would help adjustments make more sense.

We could take another page out of the book of the baseball world, and given that the CBA defines minimum salaries, use that to establish a pool of players that are considered replacement level, and derive replacement level numbers from there.

Posted by Nathan from the scoresheet! on 08/20/13 at 09:21 AM ET

PuckStopsHere's avatar

Nathan these are good questions.

Is dividing by 5 the best way to determine the baseline for a player on a given team?

Probably it isn’t the best method.  It is a method that has long been used going back to the Klein and Reif Hockey Compendium.  It is a pretty good first attempt.  I think there is a problem in sabermetrics that many people make complicated higher order corrections which go beyond the quality of the data (the statistics) and nothing is learned by this except for some math.

It seems like there needs to be a way to weight by position (at least forward vs. defense) and/or adjust to a rate stat rather than a cumulative one.

I don’t like the tendency to weight by position when it is unnecessary - is it here?  It can significantly change a player’s value merely by defining him to be another position in cases where a player’s position is uncertain (there are always players who play more than one position).  All players are valuable in terms of puck possession aren’t they?

As for a rate stat - that can be useful - it also can be misleading.  It is easier to get a good rate stat by playing sheltered minutes.  It is harder to get a good counting stat because you must play in all circumstances including tough situations.  The best players lead counting stats.  That cannot be said with rate stats.

How valuable is subtracting the team baseline? Does this accidentally victimize players on strong puck possession teams?

No it doesn’t victimize players on good teams.  Players on good teams will have good Corsi because they play on good teams.  It removes that effect.  In fact even after adjustment there are more players on good teams in a top list and more players on bad teams on a bottom list.  This is probably expected.  Good players are what makes a good team good and vice versa.

If we had a replacement level Corsi, that would help adjustments make more sense.quote]

We have a replacement level Corsi for a given team.  That is the idea here.  The Corsi that a replacement player will have depends on the team he plays with.  That is unavoidable but you seem to want to avoid it.

Posted by PuckStopsHere on 08/20/13 at 12:03 PM ET

Nathan's avatar

I don’t like the tendency to weight by position when it is unnecessary - is it here?  It can significantly change a player’s value merely by defining him to be another position in cases where a player’s position is uncertain (there are always players who play more than one position).  All players are valuable in terms of puck possession aren’t they?

Good point, I agree with that. My thinking was that perhaps positional adjustment could allow us to better judge how a player is performing in terms of puck possession given the role that he’s being deployed in. So perhaps there is some combination of this team adjustment and the zone start adjustment that makes more sense than positional. Obviously you acknowledge that in your first paragraph, so I’m not pointing the finger at all, just throwing an idea out there. But yes, I didn’t really think it through very well when I threw position out there. Something like zone starts seems more useful for that type of adjustment, on second thought.

As for a rate stat - that can be useful - it also can be misleading.  It is easier to get a good rate stat by playing sheltered minutes.  It is harder to get a good counting stat because you must play in all circumstances including tough situations.  The best players lead counting stats.  That cannot be said with rate stats.

So, I guess this goes with what you and I are both saying in the above portion that I quoted. I see what you’re saying, but perhaps the issue isn’t that a rate stat hides the type of minutes you play, but that an adjustment for something like zone starts, that at least roughly indicates the types of minutes you play, is the way to go.

Certainly, there is value in the counting stat as well. Big time. But I think if we can get some adjustment based on a number that helps us indicate the type of minutes being played, the rate stat then becomes useful to see what players may be utilized in the correct role, but are possibly being under-utilized in terms of ice time. Certainly, this won’t help much for small sample sizes. I’m thinking more along the lines of baseball UZR and UZR/150. Still working with a rate that is a large sample size, but that attempts to normalize the players’ contributions as a rate that can be compared player-to-player.

No it doesn’t victimize players on good teams.  Players on good teams will have good Corsi because they play on good teams.  It removes that effect.

I guess it depends what you’re saying. Removing the team average from the given player’s score gives us an ability to make some better judgments within the context of a team that one player is a bigger contributor to puck possession than another. But it does not give us a good way to see how a player on the Oilers or Leafs compares to one on the Hawks or Kings, or how a given player season in ‘11-‘12 compares to another given player season in ‘12-‘13.

We have a replacement level Corsi for a given team.  That is the idea here.  The Corsi that a replacement player will have depends on the team he plays with.  That is unavoidable but you seem to want to avoid it.

Then I believe the idea is wrong. To me, the idea of the replacement player is that it allows you to compare across teams (and even across years and eras). An average puck possession player is not necessarily the same thing as a team-agnostic replacement level player.

Posted by Nathan from the scoresheet! on 08/20/13 at 05:50 PM ET

Add a Comment

Please limit embedded image or media size to 575 pixels wide.

Add your own avatar by joining Kukla's Korner, or logging in and uploading one in your member control panel.

Captchas bug you? Join KK or log in and you won't have to bother.

Smileys

Notify me of follow-up comments?

Feed

Most Recent Blog Posts

About The Puck Stops Here

imageThe Puck Stops Here was founded during the 2004/05 lockout as a place to rant about hockey. The original site contains over 1000 posts, some of which were also published on FoxSports.com.

Who am I? A diehard hockey fan.

Why am I blogging? I want to.

Why are you reading it? ???

Email: y2kfhl@hotmail.com