Wins Above Replacement

[NHL WAR]

I have created a Wins Above Replacement stat for the NHL. Here is a link to a video I made going over the basics. Further results of the stat will be uploaded in the days and weeks to come.

Click to read more...

 

[Doctor No]

I'd prefer to not watch a video. What makes this more useful than the other WAR calculations out there?

 

[NHL WAR]

A) There was a R2 value of 0.703 between team WAR and team wins last season, showing the components of W.A.R do translate to wins
B) It incorporates a players even strength, power play and penalty kill contributions; nothing is left out
C) It credits players for what they have done while also accounting for the role of luck ( ex: William Karlsson did not get the same Shooting Goals AA value as a 43 goal scorer with a shooting % of 12 would have.)
D) It is still W.A.R 1.0. I am very much open to constructive criticism for potential adjustments to make this stat as good as it can be.

 

[Doctor No]

I'm familiar with WAR. What makes your WAR estimate better than the others?

 

[NHL WAR]

What others are you referring to? I am aware of the GAR model and know that corsica hockey has tinkered with WAR before and of the previous existence of the website war on ice. Do you mean one of these? I have not seen any other public WAR models and even those I mentioned before don't seem to have leaderboards or anything of that sort.

 

[Doctor No]

Corsica has a model published, yes, as does Dawson Springs (DTM About Heart). Andrew Thomas and Sam Ventura developed a model a few years back.

Those are the ones I'm aware of, and I'd consider GAR models to be essentially WAR models (since one can convert goals to wins in a straightforward fashion for a given season).

Anyhow, looking forward to seeing some results so that people can see how they smell. Thanks for posting.

 

[NHL WAR]

I would agree with you that GAR and WAR are interchangeable. In the video I showed the 2017-18 leaders in each component of W.A.R: Ovi for shooting, Barzal for passing, Bozak for shootouts, O'Reilly for faceoffs, Gaudreau for penalty differential, Giordano for defensive goals saved and Kopitar for replacement goals.

Further results will be posted in a couple days, here and on YouTube.

Cheers to the start of a new season and good luck to your favourite team (unless it is the Bruins.)

 

[Doctor No]

It's the Avalanche, unfortunately. Good luck to you and yours as well!

 

[Trilliann]

What others are you referring to? I am aware of the GAR model and know that corsica hockey has tinkered with WAR before and of the previous existence of the website war on ice. Do you mean one of these? I have not seen any other public WAR models and even those I mentioned before don't seem to have leaderboards or anything of that sort.

Corsica | WAR

Evolving-Hockey

And yes you can look at the leaders of each component if you want.
DTM About Heart's model is no longer public.

 

[NHL WAR]

Corsica | WAR

Evolving-Hockey

And yes you can look at the leaders of each component if you want.
DTM About Heart's model is no longer public.

Nice, I'll check those out sometime, see how they compare.

 

[Hivemind]

What makes this more accurate that GVT ?

 

[NHL WAR]

What makes this more accurate that GVT ?

The advantages of this stat to GVT is that it gives more credit for primary assists than secondary and adds in faceoffs and penalty differential. It also incorporates Corsi numbers rather than plus/minus. As well, if I understand it correctly, the threshold used in GVT is often a specific team, which doesn't make as neutral a baseline as league average or replacement.

 

[Michael Farkas]

From a coaching perspective, I disagree with secondary assists not being viewed as valuable...and contrary to popular belief, roughly one-third of faceoffs don't matter. In-zone draws are important - a lost defensive zone draw is a big no-no. A NZ draw is not nearly as important...

Corsi is, to put it lightly, not en vogue...

 

[Canadiens1958]

From a coaching perspective, I disagree with secondary assists not being viewed as valuable...and contrary to popular belief, roughly one-third of faceoffs don't matter. In-zone draws are important - a lost defensive zone draw is a big no-no. A NZ draw is not nearly as important...

Corsi is, to put it lightly, not en vogue...

Secondary assists probably more important as the importance of "first pass" defencemen increases.

NZ draws, ES true, PP/PK disagree.

In-zone. Should distinguish between change and no-change(post icing defensive team)

 

[NHL WAR]

From a coaching perspective, I disagree with secondary assists not being viewed as valuable...and contrary to popular belief, roughly one-third of faceoffs don't matter. In-zone draws are important - a lost defensive zone draw is a big no-no. A NZ draw is not nearly as important...

Corsi is, to put it lightly, not en vogue...

Assists value is calculated by the percentage of goals that include that type of assists ( there are more goals that require A1 than require A1 and A2.) Good point on NZ faceoffs.
Corsi is by no means perfect, but it has been heavily adjusted for WAR and is without a doubt better than plus minus.

 

[Michael Farkas]

Secondary assists need work. Ideally, adjusted for style of both team and player. Center-carry vs. defense-carry teams, etc.

Good point re: special teams faceoffs and no-change situations. Become more important if future rule change comes about where power play team can pick which side the faceoff is on. I see that being toyed with at lower levels already.

Corsi is biased towards shoot-from-anywhere wingers and defensemen. Biased against strong defensive centers.

Plus/minus has warts as well, but just because you have more data points with corsi doesn't make it more reliable. Again, ought to be adjusted for style of play and team tactics if at all possible. Otherwise, depending on its weight, I'd scrap all together...you don't want heroes like Tyler Kennedy and Cody Franson getting to the top of the heap because they have tunnel vision.

I doubt there's enough public data to make a terribly reliable stat. Then again, I didn't view the formula because it's in video form and I can't process that...

 

[Canadiens1958]

Assists value is calculated by the percentage of goals that include that type of assists ( there are more goals that require A1 than require A1 and A2.) Good point on NZ faceoffs.
Corsi is by no means perfect, but it has been heavily adjusted for WAR and is without a doubt better than plus minus.

True if game film analysis is not part of the process. Otherwise you slide down the slope of rewarding failure. A1 being a rebound assist after a transition pass A2 gives more value to failure - not scoring initially than to the important transition pass.

Plus/minus has withstood the test of time. Corsi is your basic flavour of the era stat like SV%.

 

[NHL WAR]

I will post best and worst players in all Corsi based components. That way everyone can draw their own conclusions on whether it is the right man for the job.

 

[NHL WAR]

Here are the best and worst players by each Corsi based component. This is not raw corsi, it is based off how much they outperformed the Corsi totals that would be expected based off of quality of teammates, competition and zone starts.

Shooting:
Top 10 in iCF above expected “goals”:

Ovechkin ,Alex

Tarasenko ,Vladimir

Gallagher ,Brendan

Skinner ,Jeff

MacKinnon ,Nathan

Burns, Brent

Toffoli ,Tyler

Arvidsson ,Viktor

Kane ,Evander

Kane ,Patrick

[TBODY] [/TBODY]

Bottom 10:

Schmaltz ,Nick

Tierney ,Chris

Kerfoot ,Alexander

Larsson ,Johan

Johansen ,Ryan

Desharnais ,David

Stephenson ,Chandler

Wennberg ,Alexander

Filppula ,Valtteri

Sedin ,Henrik

[TBODY] [/TBODY]

Passing:
Top 10 in CF-iCF above expected “goals”:

Stastny ,Paul

Ekholm, Mattias

Johansen ,Ryan

Toews ,Jonathan

Wennberg ,Alexander

Couturier ,Sean

Barzal ,Mathew

Giroux ,Claude

Doughty, Drew

Giordano, Mark

[TBODY] [/TBODY]

Bottom 10:

Kane ,Patrick

Laine ,Patrik

Cullen ,Matt

Boychuk, Johnny

Burns, Brent

Braun, Justin

Tarasenko ,Vladimir

Barrie, Tyson

Ceci, Cody

Ovechkin ,Alex

[TBODY] [/TBODY]

Defense:
Top 10 in CA below expected “goals”:

Hamilton, Dougie

Giordano, Mark

Couturier ,Sean

Ekholm, Mattias

Slavin, Jaccob

Letang, Kris

Backlund ,Mikael

Tkachuk ,Matthew

Miller, Colin

Giroux ,Claude

[TBODY] [/TBODY]

Bottom 10:

Skjei, Brady

Barrie, Tyson

Martinez, Alec

Kempe ,Adrian

Filppula ,Valtteri

Pyatt ,Tom

Cullen ,Matt

Orpik, Brooks

Phaneuf, Dion

Ceci, Cody

[TBODY] [/TBODY]

Top 10 in shooting,passing,defense combined:

Giordano, Mark

Hamilton, Dougie

Ekholm, Mattias

Tkachuk ,Matthew

Couturier ,Sean

Panarin ,Artemi

Toews ,Jonathan

Backlund ,Mikael

Giroux ,Claude

Slavin, Jaccob

[TBODY] [/TBODY]

Bottom 10:

Desharnais ,David

Barrie, Tyson

Pyatt ,Tom

Braun, Justin

Beagle ,Jay

Orpik, Brooks

Filppula ,Valtteri

Cullen ,Matt

Phaneuf, Dion

Ceci, Cody

[TBODY] [/TBODY]

Discuss

 

[Michael Farkas]

Quick notes at a scan...

First category (best):
As I suspected, all wingers (except converted winger Burns, former winger MacKinnon).

Next category (worst):
Defensive minded centers: Tierney, Larsson, Stephenson, Filppula...Schmaltz and Wennberg young centers that had to serve as defensive conscience to some degree because of wing deployment...no wingers on the list at all, except young Kerfoot gets used at LW some/most times.

Defense (best):
Hamilton, Letang, Tkachuk, Miller all questionable defensive players. Letang was horrific last season, the worst of his pro career.
Giordano, Slavin, Backlund all very good defensively. Giroux had to the biggest reduction in his defensive responsibility in his career...moved to LW, Couturier covered for him. Skating limitations allowed him to thrive at wing offensively.

Defense (worst):
Cullen and Orpik ought not to belong. Orpik was much improved from the previous year. The rest of the list isn't bad. Skjei, Barrie, Phaneuf and, surprisingly, Ceci had been poor defensively.

 

[NHL WAR]

Quick notes at a scan...

First category (best):
As I suspected, all wingers (except converted winger Burns, former winger MacKinnon).

Next category (worst):
Defensive minded centers: Tierney, Larsson, Stephenson, Filppula...Schmaltz and Wennberg young centers that had to serve as defensive conscience to some degree because of wing deployment...no wingers on the list at all, except young Kerfoot gets used at LW some/most times.

Defense (best):
Hamilton, Letang, Tkachuk, Miller all questionable defensive players. Letang was horrific last season, the worst of his pro career.
Giordano, Slavin, Backlund all very good defensively. Giroux had to the biggest reduction in his defensive responsibility in his career...moved to LW, Couturier covered for him. Skating limitations allowed him to thrive at wing offensively.

Defense (worst):
Cullen and Orpik ought not to belong. Orpik was much improved from the previous year. The rest of the list isn't bad. Skjei, Barrie, Phaneuf and, surprisingly, Ceci had been poor defensively.

U have a point with the center/ winger responsibility difference. On the one hand, those centers who are punished in shooting grade far better than the wingers in the passing category, which balances it out. On the other, anyone will tell you a center has more responsibility than a winger and all else being equal they would take a center. I have wondered whether a positional adjustment would fix this...either via comparing centers to centers and wingers to wingers or by adding positional "goals" to the WAR formula ( adding a couple "goals" for centers, taking away a couple from wingers.)

 

[Michael Farkas]

Team tactics matter, so do player styles.

Instances of wingers taking F1 defensive responsibilities (typically center responsibilities) are not terribly uncommon: Jari Kurri taking F1 responsibilities for Gretzky to allow Gretzky to man the pointmen and better facilitate transition chances...sometimes Pascal Dupuis would do the same for Sidney Crosby...now Crosby plays with two young'ins in Conor Sheary (since departed) and Jake Guentzel who are disreputable defensive players. This year expected to see time with rookie Daniel Sprong, known for his lack of defensive conscience. As a coach, certain players can be trusted bolster and shelter young players (Crosby) and some players can't (Malkin almost never gets young players...he gets Kunitz, Kessel, Neal, etc.; Ovechkin in Washington is another example of this, always someone to support him, not the other way around). Then defensive deployment applies as well...

Naturally, you're not going to come up with every variation that will ever happen. That's why the eye test is the gold standard. But there's a lot of stats that have tried to be a one-stop-shop that fall way short because there's only a loose connection to the game itself. Target audience shouldn't be fan polls, it should be coaches/scouts. Come from a place of authority with good input, don't try to placate mass appeal - which seems to be what a lot of these types of catch-all stats attempt to do...

 

[NHL WAR]

Team tactics matter, so do player styles.

Instances of wingers taking F1 defensive responsibilities (typically center responsibilities) are not terribly uncommon: Jari Kurri taking F1 responsibilities for Gretzky to allow Gretzky to man the pointmen and better facilitate transition chances...sometimes Pascal Dupuis would do the same for Sidney Crosby...now Crosby plays with two young'ins in Conor Sheary (since departed) and Jake Guentzel who are disreputable defensive players. This year expected to see time with rookie Daniel Sprong, known for his lack of defensive conscience. As a coach, certain players can be trusted bolster and shelter young players (Crosby) and some players can't (Malkin almost never gets young players...he gets Kunitz, Kessel, Neal, etc.; Ovechkin in Washington is another example of this, always someone to support him, not the other way around). Then defensive deployment applies as well...

Naturally, you're not going to come up with every variation that will ever happen. That's why the eye test is the gold standard. But there's a lot of stats that have tried to be a one-stop-shop that fall way short because there's only a loose connection to the game itself. Target audience shouldn't be fan polls, it should be coaches/scouts. Come from a place of authority with good input, don't try to placate mass appeal - which seems to be what a lot of these types of catch-all stats attempt to do...

I agree. The defensive aspect is so much harder to capture than the offensive. While the current formula for defense doesnt capture everything ( every coach having different roles and strategies for their players makes this impossible) I think it serves it's purpose. Good points and advice.

 

[Canadiens1958]

Center/winger issue.

Compare Claude Giroux, Corsi numbers at each position. Perhaps Jonathan Drouin.

 

[NHL WAR]

Center/winger issue.

Compare Claude Giroux, Corsi numbers at each position. Perhaps Jonathan Drouin.

Claude Giroux's Corsi For % each season since 2010-11 (excluding lockout shortened season) has ranged from 53.2% to 54.4%. Jonathan Drouin experienced a slight drop going from winger to center, but one that can be explained easily when you factor him leaving Tampa for Montreal. These examples agree with my thinking that position doesn't tilt Corsi ratings one way or the other. On the plus side, wingers aren't getting extra credit...the question remains should centers?