Ryan Stimson's excellent tracking of passing, zone exits, "shot attempts generated" and his derivative "Corsi Contribution," (as well as some discussion regarding the usefulness of hits as a statistic following the Columbus blowout) got me thinking about how the interest in tracking all of these various statistics is generally to gauge how an individual player performs in various ways. But ultimately the end result of player performance should be to contribute in some manner to offensive goal scoring and defensive goal suppression.
A shot attempt, a hit thrown, a pass attempt, a successful zone exit are all intended to lead to two outcomes: (1) scoring a goal, and (2) suppressing a goal against. Why is there no clear measurement of how the score changes when a given player is on the ice? Hockey Analysis tracks goals for and goals against when players are on the ice. So I looked to see if I could use these numbers to determine how the team's likelihood to score goals and suppress goals against is effected when a given player is on the ice.
This I did by using Brodeur's and Schneider's 5-on-5 WOWY (Without & With-You) charts which have the team's GF20 (goals scored by the team per 20 minutes) and GA20 (goals scored against the team per 20 minutes) when each player is on the ice and off the ice. These charts are limited to 5-on-5 situations.
I added each player's GF20 and GA20 from each Goaltender's WOWY pages to get a total Team WOWY. Then I subtracted the GF20 without player from the GF20 with player to determine "Relative GF20." The same was done with GA20. This generates a number where positive Relative GF20 (team scores more with player) is good and positive GA20 (team is scored upon more with player) is bad. Relative GF20 gauges how many more 5-on-5 goals the team scores per 20 minutes when a player is on the ice versus off the ice. Relative GA20 is the same for goals against. Lastly, I subtracted GA20 from GF20 to generate a summary statistic that expresses the relative total goal differential of the team per 20 minutes when a given player is on the ice versus off the ice. This can also be expressed as how the score changes when a given player is on the ice versus when he is off the ice. This is what I have lovingly (and tongue-in-cheekingly) termed "goalsie." Goalsie = (GF20WY-GF20WO)-(GA20WY-GA20WO).
I then ran the averages and standard deviation to determine where a player was within the top and bottom 16% of anticipated results based on a normal distribution, and where a player was within the top and bottom 25% of anticipated results based on a normal distribution. To represent this, I color coded each number. Dark Red = Bad, Orange = Below Typical, Uncolored = Typical; Light Green = Above Typical; Dark Green = Great.
Here is how the team turned out after 34 games (players with less than 138 minutes are separated and defensemen are bolded:
Based on "goalsie" there were only a few surprises for me. I can sum up those three surprises as: Bernier, Fayne, and Volchenkov. Now remember, this is only 5-on-5 and this doesn't account for quality of competition, but I was surprised at how much the score tilts in favor of the opposing team when Bernier has been on the ice versus when he has been off of it. Likewise, I was surprised at how much the score tilts in favor of the Devils when Volchenkov and Fayne are on the ice, and even more surprised that this improvement is fueled by better offense as opposed to defensive goal suppression.
Now since almost every single decision and action by the team and plenty of external influences culminates in the score, there are countless reasons, why deploying a given player can effect goal scoring. Quality of teammates, Quality of Competition, when within a game (leading, tied, trailing) a player is deployed, when in the season a player has played (early before chemistry had time to develop or late when the team found its stride). Still looking at the 24 games this season and how each player has effected goal scoring on 5-on-5 does seem to me to be worthwhile.
As a sidenote, whether or not there is an opportunity cost to having Janssen and Sestito on the fourth line, one cannot argue that they have well executed their very limited role based on goal suppression.
What do you think? Have I made some grand mathematical and statistical error in my calculation method. Do you find this type of data useful? Are you surprised at how a particular player has effected scoring?