Friday, July 10, 2009

A Note On Expected Goals

Expected Goals has popped up around here lately - it's something devised in-house and can be brutally wrong, but it gives at least a guideline of what to expect of new players on new teams.

Expected Goals is basically:

(Expected Shooting Percentage) * (Expected Shots On Goal)

Expected Shooting Percentage is (Average Shooting Percentage Since Lockout) * (% Difference Between Shooting Percentage Of Old Team and New Team in 2008-09 Season)

Expected Shots on Goal is (Average SOG/Game Since Lockout) * (% Difference Between Shots On Goal of Old Team and New Team in 2008-09 Season) * (% Increase/Decrease Expected As A Result of Aging)

There's lots of adjustments that can and should be made here, but they are quite difficult - they include time on ice adjustments, power play time on ice adjustments, power play strength adjustments, penalty killing time on ice adjustments, strength of expected linemates, and so forth; not adjusting for these things likely skews things between 1-3 goals either way (but could vary things as much as 5-10 goals in extreme cases). Regardless, it's a reasonable ballpark metric, and is certainly more instructional than looking simply at the past.


  1. Are you familiar with any studies that look at these adjustments?

    SOG for example seems like it could go both ways. Teams that control play tend to get more shots so that implies the direction you are assuming. On the other hand, a team that shoots a lot might do so because they have guys that fire the puck at the net if they can see the goalie, instead of passing. In this case, I may go from an average-shots team to a high-shots team and find that I'm passing to teammates who are shooting a hell of a lot more than vice-versa and my shots could go down and not up.

    As an example, I would think a guy would get more shots on Sidney Crosby's wing than on the one opposite Ovechkin, but Washington, largely because of Ovechkin's line, got a lot more shots than Pittsburgh.

    I don't have a solution to this problem and agree with you that it's difficult, I'm just wondering what research has been done on this.

  2. admittedly, no.

    interestingly enough, guerin and kunitz's shots/game didn't really change when moved to the penguins and with crosby. but if you look at dainius zubrus's career, he definitely gets more S/G when playing with ovechkin than at any other time, and i don't think it's just era and TOI too.

    i wouldn't think someone like crosby would increase a player's SOG tremendously - it would increase his shooting percentage.

    but yeah pittsburgh is a bad example, the devils basically outshot them 2 to 1 3 times during the year when therrien was coaching - 'expected goals' is going to downplay guys going there this year.