Improving Defensive Statistics and Quantizing Power Forwards

[Wesleyy]

I finally got around to writing on new hockey metrics again. As many of you many know, possession is inversely correlated with hits and blocked shots. As a player has possession of the puck more when he is on the ice, his opportunities to hit will decrease. I looked at the regression between hits ~ laegap, and blocked shots ~ laegap and used the residual of the regressions to develop the stat "Expected Hits Difference" and "Expected Blocked Shots Difference". By looking at the residuals, we can better understand a player's physicality without biases caused by different levels of possessions between different players. Here is the full article with more details: http://hockeymetrics.net/improving-defensive-statistics-and-quantizing-power-forwards/

Click to read more...

 

[Micklebot]

I finally got around to writing on new hockey metrics again. As many of you many know, possession is inversely correlated with hits and blocked shots. As a player has possession of the puck more when he is on the ice, his opportunities to hit will decrease. I looked at the regression between hits ~ laegap, and blocked shots ~ laegap and used the residual of the regressions to develop the stat "Expected Hits Difference" and "Expected Blocked Shots Difference". By looking at the residuals, we can better understand a player's physicality without biases caused by different levels of possessions between different players. Here is the full article with more details: http://hockeymetrics.net/improving-defensive-statistics-and-quantizing-power-forwards/

Interesting read, but why use a location adjusted metric as your proxy for possession instead of Corsi? Have you found it correlates better than Corsi does to possession times? If the goal is who blocks shots and hits more per min without the puck, then why include location data?

 

[Wesleyy]

Interesting read, but why use a location adjusted metric as your proxy for possession instead of Corsi? Have you found it correlates better than Corsi does to possession times? If the goal is who blocks shots and hits more per min without the puck, then why include location data?

There wasn't a specific reason I used laegap over corsi for this study in particular. I am just trying to push laegap as a complete replacement for corsi because I believe laegap can give us a better picture, by design.