Post-Game Talk: Player Performance Tracking: Contributing to Goals For/Against

[Hivemind]

http://www.japersrink.com/2013/10/10/4818726/japers-rink-mailbag-john-erskine-possession-and-kuznetsov

Speaking of metrics/GF/GA tracking...how the heck is this possible???

Erskine, by advanced stats, is having a strong start.

Last year he was, by advanced stats, "lucky" at best.

To my eyes its been the opposite. He was outstanding last year and putrid this year.

Simple. Three games is not an adequate sample size to do much real statistical analysis.

Also...as a team...how are our possession number stacking up so far if anyone knows?

Poorly.

 

[Ajax1995]

Poorly.

But of course three games is not an adequate sample size to do much real statistical analysis so...

 

[Langway]

Game 4, 10/10/13 V CAR:
Chimera ES: Chimera +0.65, Fehr +0.15 (soft goal so...)
Lindholm ES: Holtby -0.5, Carlson -0.2, Wilson -0.15, Fehr -0.15
Ovechkin ES: Ovechkin +0.7, Oleksy +0.15, Alzner +0.1, Backstrom +0.05
Semin PP: Holtby -0.5, Backstrom -0.35, Carlson -0.15
Gerbe ES: Holtby -0.4, Carlson -0.3, Erskine -0.3

ES:
Ovechkin +0.7
Chimera +0.65
Oleksy +0.15
Alzner +0.1
Backstrom +0.05
Wilson -0.15
Erskine -0.3
Carlson -0.5
Holtby -0.9

PK:
Carlson -0.15
Backstrom -0.35
Holtby -0.5

Same ol' same ol' for Erskine & Holtby. By this metric at least Holtby's performance to date has been as bad as Ovechkin's has been good. Fehr & Erskine really stick out as failing in their current roles among skaters.

 

[KevinM]

Can someone explain to me why Fehr has been shifted to Center? Hasn't GMGM himself explicitly said you can move a Center to wing but you can't move a wing to center? Meanwhile Laich can play 3C no problem and Erat could fill in for him at 2LW without really degrading the quality of our top 6 and any ******* can plug the hole on the fourth line for less than 4.5 million in cap.

I just don't understand the decision at all, what am I missing?

 

[Langway]

Game 5, 10/12/13 V COL:

Tanguay ES: Carlson -0.7, Backstrom -0.2, Neuvirth -0.1
Duchene ES: Alzner -0.7, Neuvirth -0.3
MacKinnon PP: Oleksy -0.35, Erat -0.25, Alzner -0.2, Neuvirth -0.2
Tanguay SH: Neuvirth -1
Fehr ES: Fehr +0.5, Chimera +0.25, Ward +0.15, Carlson +0.1
McGinn ES: Oleksy -0.35, Ovechkin -0.3, Neuvirth -0.25, Schmidt -0.2

ES:
Fehr +0.5
Chimera +0.25
Ward +0.15
Backstrom -0.2
Schmidt -0.2
Ovechkin -0.3
Oleksy -0.35
Carlson -0.6
Neuvirth -0.65
Alzner -0.7

PP:
Neuvirth -1

PK:
Alzner -0.2
Neuvirth -0.2
Erat -0.25
Oleksy -0.35

Chimera ranked second among forwards at ES through five games.

Oleksy was overmatched by Colorado's pace and goaltending continues to smell. At least Urbom looks capable enough.

 

[malyk]

Gotta play catch up!

Game 4: Carolina at Washington

Carolina Goals

Lindholm
-1 Fehr - gave up on the play thinking it was offsides
-1 Wilson - slow to pressure the puck
-1 Carlson - trying to block an easy shot unnecessarily

Semin
-1 Backstrom for losing the draw too cleanly

Gerbe
-1 Carlson getting striped in the corner
-1 Carlson again for the weak pass back to holtby
-1 Erskine for flailing away at the puck

Washington Goals

Chimera
+1 for chimmer

Ovechkin
+1 Alzner with a good pinch
+1 Oleksy putting the puck on net
+1 Ovi with the deflection

+1 Alzner
+1 Chimera
+1 Oleksy
+1 Ovechkin
-1 Backstrom
-3 Carlson
-1 Erskine
-1 Fehr
-1 wilson

 

[malyk]

Game 5: Colorado at Washington

Colorado Goals

Tanguay
-1 Ovechkin horrible backcheck
-1 Backstrom slow backcheck
-1 Mojo biting on the weak deke

Duchesne
-1 Alzner...yikes.

MacKinnon
-1 Alzner gets too tied up in the corner
-1 Oleksy loses mckinnon
-1 Erat doesn't pick up mackinnon

Tanguay
-1 Backstrom with the weak stick swing/check
-1 Neuvirth..gotta have that one.

McGinn
-1 Schmidt with a bad pass off grabo's legs
-1 Ovechkin for giving up on the backcheck too soon



Washington Goal

Fehr
+1 Ward with good board work
+1 Chimera with a nice early look and great backhand pass
+1 Fehr burying it



Totals

+1 Chimera
+1 Fehr
+1 Ward
-2 Alzner
-2 Backstrom
-1 Erat
-1 Johansson
-1 Neuvirth
-1 Oleksy
-1 Ovechkin
-1 Schmidt

 

[Langway]

Game 6, 10/14/13 V EDM:
Gordon ES: Green -0.7, Ovechkin -0.3
Laich ES: Laich +0.6, Oleksy +0.3, Grabovski +0.1
Ward PP: Ward +0.6, Backstrom +0.4
Ovechkin ES: Ovechkin +0.6, Johansson +0.3, Backstrom +0.1
Brouwer PP: Brouwer +0.6, Grabovski +0.25, Backstrom +0.15
Acton ES: Schmidt -0.4, Ovechkin -0.3, Carlson -0.3

ES:
Laich +0.6
Johansson +0.3
Oleksy +0.3
Grabovski +0.1
Backstrom +0.1
Carlson -0.3
Schmidt -0.4
Green -0.7

PP:
Ward +0.6
Brouwer +0.6
Backstrom +0.55
Grabovski +0.25

 

[BobRouse]

Simple. Three games is not an adequate sample size to do much real statistical analysis.


Poorly.

Well yeah it goes both ways. Now the sample has doubled.

I saw the games vs Carolina and Edmonton and thought we had the better of the chances in both of those games.

I didn't see Colorado but based on some comments Oates made it appeared we were in there zone 3x longer than they were in ours. Varlamov also had 41 saves for them.

The first period vs the Oilers was brutal but then I thought we took over from there.

I'm glad we finally got some breaks vs the Oilers. We were brutally unlucky before then

 

[Hivemind]

Do keep in mind that three (or any quantity "X") games is a much bigger sample size for a team than it is for an individual player. A lot more events happen with a much broader mix of variables when looking at an entire team's set of data compared to a player (who's zone starts, match-ups, and linemates all bias his data). While three games certainly isn't a large sample size for a team, I feel much more comfortable doing team analysis on that sample than individual player.

So it doesn't really cut both ways all that much.

 

[BobRouse]

Do keep in mind that three (or any quantity "X") games is a much bigger sample size for a team than it is for an individual player. A lot more events happen with a much broader mix of variables when looking at an entire team's set of data compared to a player (who's zone starts, match-ups, and linemates all bias his data). While three games certainly isn't a large sample size for a team, I feel much more comfortable doing team analysis on that sample than individual player.

So it doesn't really cut both ways all that much.

Well how have we been thru 6 games so far? I see some mixed results and a ton of bad puck luck (someone mentioned we were bad in another post so it confirms what I felt)

 

[Ajax1995]

Well how have we been thru 6 games so far? I see some mixed results and a ton of bad puck luck (someone mentioned we were bad in another post so it confirms what I felt)

Give a man a fish or teach a man to fish...?

Enjoy

http://www.extraskater.com

 

[Hivemind]

Well how have we been thru 6 games so far? I see some mixed results and a ton of bad puck luck (someone mentioned we were bad in another post so it confirms what I felt)

You're right in terms of the PDO being terrible. Hockeyanalysis lists our 5v5 PDO as 951, good for 29th in the league.

The shots for at even strength has improved overall, with the Caps now registering 50.2% of 5v5 shots. However the 5v5 GF% is only 34.8%, and the shots statistics in close (44.5%) and tied (40.9%) scenarios range from bad to absolutely awful. So the Caps are riding score effects (and the fact they trail a lot) to better shot statistics. This is mirrored by their Corsi and Fenwick stats as well.

This would seem to be the case with the dominating possession/shot effort against Colorado, but I do have some question about when Colorado decided to change up their strategy. I didn't see a shift in their gameplan until the 3rd period, so I think part of that may have actually been better possession play from Washington rather than purely driven by score effects. Granted, Colorado is a below average Corsi/Fenwick team in 5v5 close scenarios, and is below average is SF%, CF%, and FF% at overall 5v5 scenarios. They're more of a counter-attacking team than a possession team, so beating them in the possession stats probably isn't all that telling.

 

[BobRouse]

I reviewed that site a bit and will try and follow it.

Basically I am trying to compare the advanced stats to what I see taking place and see how well they align.

While I don't think we've been great I don't think we've been terrible either and if our goaltending was a bit more solid (especially some of the downright bad goals I've seen) and offense a bit more opportunistic (outside of Ovechkin) then our record would be easily reversed.

 

[Liberati0n*]

You're right in terms of the PDO being terrible. Hockeyanalysis lists our 5v5 PDO as 951, good for 29th in the league.

The shots for at even strength has improved overall, with the Caps now registering 50.2% of 5v5 shots. However the 5v5 GF% is only 34.8%, and the shots statistics in close (44.5%) and tied (40.9%) scenarios range from bad to absolutely awful. So the Caps are riding score effects (and the fact they trail a lot) to better shot statistics. This is mirrored by their Corsi and Fenwick stats as well.

This would seem to be the case with the dominating possession/shot effort against Colorado, but I do have some question about when Colorado decided to change up their strategy. I didn't see a shift in their gameplan until the 3rd period, so I think part of that may have actually been better possession play from Washington rather than purely driven by score effects. Granted, Colorado is a below average Corsi/Fenwick team in 5v5 close scenarios, and is below average is SF%, CF%, and FF% at overall 5v5 scenarios. They're more of a counter-attacking team than a possession team, so beating them in the possession stats probably isn't all that telling.

How much time have they actually played tied and close? They've given up early leads in a lot of games, so the sample can't be that big yet (not that I'd really expect it to be much better anyway).

 

[Hivemind]

How much time have they actually played tied and close? They've given up early leads in a lot of games, so the sample can't be that big yet (not that I'd really expect it to be much better anyway).

302 corsi events in 5v5 close scenarios. League mean is 323.67, with the median at 321.5. Lowest is St. Louis with 199 events, highest is Vancouver with 441. So the spread is pretty large right now, with Washington having the 19th most events.

156 corsi events in 5v5 tied scenarios. Leage mean is 188.33, with the median at 189. Lowest is the Rangers with 87 events, highest is Buffalo with 296. Once again, a pretty wide spread, but Washington is proportionally further away from the mean and median, but is roughly in the same position league wide (20th most events).

Not a tremendous sample size, but still a decent peek into what's happening. For reference, Washington has had a total of 515 5v5 corsi events, with the league mean at 529.

 

[RandyHolt]

Look at that, Erat and Wilson leading the CF Rel charge

Critics. Tiny sample size. Supporters. Dominating.

http://www.extraskater.com/players/dashboard?team=wsh

And look at that, Laich... and I am sorry I am trying not to laugh as I type this, and Brouwer taking up the rear

 

[Hivemind]

Randy, you may also want to check out these websites to figure out which you like best.

http://stats.hockeyanalysis.com/index.php
http://behindthenet.ca/

 

[Liberati0n*]

Could someone explain what goes into the HAR stats?

 

[RandyHolt]

I am quickly overwhelmed using just one site.

iCorsi, Ovi FTW, or Points/60 Oleksy. I can't lose.

Is there a single goto stat, to measure 5 on 5 play.

Corsi On? Carlson, dominating.

 

[Hivemind]

There really isn't a good, "all-in-one" hockey stat yet. A few people have tried, most notably HockeyProspectus' GVT (Goals Versus Threshold) stat, but the results have been met with pretty poor reactions.

There are pros and cons to Corsi, Fenwick, and relative Corsi in terms of reasonably estimating a player's impact on possession. Those are generally the most accepted, but it still takes a lot of context to put those numbers to any good use.

 

[Xaroc]

There really isn't a good, "all-in-one" hockey stat yet. A few people have tried, most notably HockeyProspectus' GVT (Goals Versus Threshold) stat, but the results have been met with pretty poor reactions.

There are pros and cons to Corsi, Fenwick, and relative Corsi in terms of reasonably estimating a player's impact on possession. Those are generally the most accepted, but it still takes a lot of context to put those numbers to any good use.

This isn't too bad a high level guide: http://www.broadstreethockey.com/2013/3/14/4104952/hockey-advanced-stats-corsi-intro-primer

But keys to look for are Corsi 5v5 close (or Fenwick 5v5 close), zone starts, quality of competition, and quality of teammates. That gives you a pretty good picture of how guys are being used and against who and who they are playing with.

 

[malyk]

Game 6: Edmonton at Washington

Edmonton Goals

Gordon
-1 Green what was that pass?
-1 Ovechkin slow on the backcheck

Acton
-1 Schmidt getting beat 1v1
-1 Ovechkin for losing his point man
-1 Carlson with a slow recovery and ignoring acton

Washington Goals

Laich
+1 Oleksy with a nice quick pass
+1 Laich with a quick release

Ward
+1 Backstrom with a beauty of a pass
+1 Ward for the laser

Ovechkin
+1 Johansson with a nice feed
+1 Backstrom for getting to the front of the net and disrupting the goalie
+1 Ovechkin

Brower
+1 Backstrom patience
+1 Grabovski touch pass
+1 Brower one timer

Totals

-1 Carlson
-1 Green
-1 Ovechkin
-1 Schmidt
+1 Brower
+1 Grabovski
+1 Johansson
+1 Laich
+1 Oleksy
+1 Ward
+3 Backstrom

 

[usiel]

My job for the past 13 years is data/spatial analytics. That said the eye test, imho, just seems like bs when this fancy stats stuff when its less then quarter season/half season/season in review stuff.

 

[Hivemind]

With only two goals last night, it'll be any easy first track for me. I'll get back to the first six games eventually (hopefully):

Game 7 - New York Rangers @ Washington Capitals

5-on-5:

• Callahan goal:
  Carlson -0.5 (Allowed Callahan to get in front of the net)
  Ovechkin -0.3 (Didn't contest Callahan's shot)
  Urbom -0.2 (Moved out from in front of the net, but wasn't able to deflect/contest Richards' pass)
  
  4-on-4:
• Moore goal:
  Carlson -0.4 (Ditched Moore)
  Fehr -0.3 (Didn't pick up Moore when Carlson ditched him)
  Holtby -0.3 (That was a savable shot)