seventieslord
Student Of The Game
Is there a place where I can see stats for empty net goals for and against by teams? Including for this season?
Yes, you can!
http://morehockeystats.com/teams/en
Unless I am missing something, your stats don't match these:
https://www.sportingcharts.com/nhl/stats/team-empty-net-goal-percentage/2016/
Makes me wonder.
I get the EN goals from the play by play HTML reports from the NHL.com (checking the players on ice column). It's possible that these reports are not correct. I should convert to parsing the JSON, just haven't the time to do that.
Pardon my error.... I was incorrectly viewing your site. The difficulty comes in the drop down menu.
Playing with empty net.... Obviously means 'trying to score extra attacker goal'. Thus, Minnesota's column there is.....EAG.....ENG against.
One has to change the drop down menu to 'playing against empty net' to find ENGs scored.
Again, my apologies....
Wow, so Calgary, LA and Nashville have 7, 8, 9 goals against wrecking their differentials in games that were actually razor thin.
As far as I’m concerned, losing a goal by 1+EN is the same as losing by 1.
I know goal differential is a better predictor than W/L, but has anyone taken out ENGF and ENGA to see if what you’re left with is an even better predictor? I think it has to be.
Well, maybe.. But then you get the 10-0 between CBJ and MTL which is large enough to skew any prediction.
You might be interested in this blog entry I wrote a few weeks ago.
When I looked at it from an SRS perspective, the most predictive modifications were to remove empty-net goals but also collapse large margins of victory (basically what MHS said above).
I also found that it's better to do neither than it is to just do one or the other.
I should look at that again (I don't have access to my grad school accounts any longer, sadly).
Almost exactly - I don't truncate them, but make each additional goal in a victory worth less than the previous ones (so a 6-2 win is better than a 5-2 win, but not as much as a 5-2 win is better than a 4-2 win).
(I'll add that this is what I do when I want to maximize predictive power instead of measure things like strength of schedule - the difference between concurrent modeling and prospective modeling).