HFBoards

Go Back   HFBoards > General Hockey Discussion > By The Numbers
Mobile Hockey's Future Become a Sponsor Site Rules Support Forum vBookie Page 2
By The Numbers Hockey Analytics... the Final Frontier. Explore strange new worlds, to seek out new algorithms, to boldly go where no one has gone before.

Evaluating Penalty Killing (What's Useful and How Can We Make it Better?)

Reply
 
Thread Tools
Old
07-28-2014, 08:09 PM
  #1
Mystlyfe
We're Touched
 
Join Date: Oct 2010
Location: Philadelphia
Posts: 12,837
vCash: 500
Evaluating Penalty Killing (What's Useful and How Can We Make it Better?)

There aren't any great metrics for evaluating individual performance as a penalty killer (or powerplay presence, for that matter), and I'm not aware of many studies regarding this area. Inspired by the premises in this thread, I decided to see how well various possession and non-possession events stand up to PK% for teams in 2013-14. For now, this is operating under the assumption that PK% is a good metric to evaluate a penalty killing unit. It's not perfect, but it's still likely the best we have given the goal of surviving a penalty without surrendering a goal trumps all else when it comes to killing penalties.


Predictably, GA/2 stacks up rather well with PK%. No surprises there.


This is where things start to get interesting. Shots Against/2 has a pretty darn weak correlation with actually killing penalties. While preventing shots as a whole is a good idea, there are some poor PK units that don't give up a ton of shots.


Corsi-event Against/2 fairs even worse than SA/2, and has the weakest correlation of any metric. So weak that it doesn't seem particularly useful at all.


Fenwick-event Against/2 does slightly better than Corsi. All things considered, this makes sense. Possession metric followers tend not to favor blocked shots not because it's a bad idea to block shots, but rather because high blocked shot totals mean the other team has control of the puck more. Given that it's pretty much assumed that the team on the powerplay is going to have the vast majority of possession, blocked shots are obviously a positive factor to killing penalties. Still, the FA/2 has a pretty weak correlation to PK%, and is worse than SA/2 (potentially related: the impact of goalies on missed shots).


Perhaps most surprising is how weak the correlation between sv% and PK% is. Based on shot metrics doing poorly, I was assuming that scoring chance quality would have to be the primary mover here, and that sv% could play out that hypothesis. While it certainly has a substantially larger correlation than anything aside of GA/2, it's nothing to write home about.

All this is pretty quick and dirty (I would have used more than one year's data if I had more time or it was quicker to combine the sources I needed), but was more intended to get the discussion rolling. What metrics are actually useful for measuring a penalty killing unit and individual penalty killers? How can they be improved?

Mystlyfe is online now   Reply With Quote
Old
07-28-2014, 08:21 PM
  #2
BroadwayJay
Corsi Hawk
 
BroadwayJay's Avatar
 
Join Date: Jun 2014
Location: New York
Country: United States
Posts: 1,942
vCash: 500
This is fascinating and, yet, wholly disappointing!

Where do we go from here?

I'm open to suggestions.

BroadwayJay is online now   Reply With Quote
Old
07-29-2014, 12:09 AM
  #3
NHLPaul
Registered User
 
NHLPaul's Avatar
 
Join Date: Jun 2014
Location: Philly
Posts: 341
vCash: 500
For an individual player, I think CA and CF are good indicators. As far as for a team though it's tough as it relies on sv%, where the shots are coming from (a team may only get two shots but both are back door or a team may get 7 and 6 are long range or bad angle drives), and obviously deflections and bounces and stuff.

NHLPaul is offline   Reply With Quote
Old
07-29-2014, 08:38 AM
  #4
BroadwayJay
Corsi Hawk
 
BroadwayJay's Avatar
 
Join Date: Jun 2014
Location: New York
Country: United States
Posts: 1,942
vCash: 500
Maybe take bigger samples. Try a three-year distribution.

BroadwayJay is online now   Reply With Quote
Old
07-29-2014, 10:48 AM
  #5
charlie1
Registered User
 
charlie1's Avatar
 
Join Date: Dec 2013
Location: Massachusetts
Posts: 472
vCash: 500
Quote:
Originally Posted by Mystlyfe View Post


Perhaps most surprising is how weak the correlation between sv% and PK% is. Based on shot metrics doing poorly, I was assuming that scoring chance quality would have to be the primary mover here, and that sv% could play out that hypothesis. While it certainly has a substantially larger correlation than anything aside of GA/2, it's nothing to write home about.
Nice work.

Regarding the bolded-- Sv% explains 42% of the variance in PK%. That's pretty darn good.

So this is going to boil down to shot quality, and maybe goalie quality. A spatial analysis is the next step. Someone had blogged about a location-adjusted shot quality statistic, but I can't seem to find it now.

By the way, my Islanders are the lonely dot to the far left.

charlie1 is offline   Reply With Quote
Reply

Forum Jump


Bookmarks

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off



All times are GMT -5. The time now is 11:14 AM.

monitoring_string = "e4251c93e2ba248d29da988d93bf5144"
Contact Us - HFBoards - Archive - Privacy Statement - Terms of Use - Advertise - Top - AdChoices

vBulletin Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
HFBoards.com is a property of CraveOnline Media, LLC, an Evolve Media, LLC company. 2014 All Rights Reserved.