One of the first posts on this blog was about looking at how much players play together and how well they do together. My post called: "introduction to shift analysis" explains the details of this method with examples. A quick summary: to tell how good a player is, look down a given player's column and if it's mostly green they are helping others do better and if it is red, they're making others worse. If it's white then I lack information to say either or (played less than an hour together).
I've attempted to simplify these extremely complicated, but interesting, tables. By providing the players weighted average score on the bottom row (labeled average). If the '+R' column is green it means the player has helped the team offensively (and hurt the team if it's red). If the '-R' column is green it means the player has helped the team defensively (and hurt the team if it's red). If both averages are green (greater than 1 for +, less than 1 for -) then the player has had more goals scored then expected while on the ice and allowed fewer than expected while on the ice (a very good thing). Often a player hurts the team offensively (+R=0.68) and helps the team defensively (-R=0.68) (or vice versa). One can think of the net contribution above expectations to be equal to (+R) - (-R).
Example:
From Vancouver:
Row: Cooke [55139 seconds], Column: Bieksa
- Cooke spent 38% of his ice time with Bieksa
- Cooke scored 39% of his goals with Bieksa
- Cooke had 31% of his goals against with Bieksa
Bieksa:
Another way of saying this: Cooke should have had 38% of his goals against with Bieksa based on ice time, however Bieksa was somehow (either by luck or skill) able to lower the number of goals against while Cooke was on the ice to 31%. So basically I'll reward Bieksa for lowering Cooke's goals against score and give Bieksa a green defensive score (-R) of 31%/38% = 0.81. In general if you look at Bieksa defensive score (-R) he actually makes most players worse. His team average is 1.13, but he makes up for his defensive skills with his offensive skills: +R=1.13, making him basically an even defenseman (1.13 - 1.13).
- Vancouver Canucks
- Minnesota Wild
- Edmonton Oilers
- Colorado Avalanche
- Calgary Flames
- St. Louis Blues
- Nashville Predators
- Detroit Red Wings
- Columbus Blue Jackets
- Chicago Blackhawks
- San Jose Sharks
- Phoenix Coyotes
- Los Angeles Kings
- Dallas Stars
- Anaheim Ducks
- New Jersey Devils
- New York Islanders
- New York Rangers
- Philadelphia Flyers
- Pittsburgh Penguins
- Boston Bruins
- Buffalo Sabres
- Montreal Canadiens
- Ottawa Senators
- Toronto Maple Leafs
- Atlanta Thrashers
- Carolina Hurricanes
- Florida Panthers
- Tampa Bay Lightning
- Washington Capitals
- Time% - this represents the percentage of the row player's ice time that column player played with the row player. [(row and column time)/(row time)]
- +% - this represents the percentage of the row player's goals for that column player had with the row player [(row and column goals for)/(row goals for)]
- -% - this represents the percentage of the row player's goals against that column player had with the row player [(row and column goals against)/(row goals against)]
- +R = +%/Time%
- -R = -%/Time%
- Time measured in seconds.
7 comments:
Kick ass stuff, JG. I like these a lot.
Thanks for the quick turnaround, too. Glad it was easier this time around.
Analysis is nice presentation needs some work. It's a bit siezure inducing
I would like to see some sort of error. To be exact I think that those values that cann't be distinguished from zero by say a 95% CI shouldn't be colored at all.
mogen:
Simply stated: There is too much error for such a technique, because the board would be white. But I might consider choosing a different cutoff. The current method is simple.
Is this analysis of all time spent on the ice or just even-strength time?
Thanks,
Scott.
Anonymous:
I should have mentioned that in the post.
This is just even strength
Post a Comment