Quote:
Originally Posted by DeluxeMoustache
Yes, with the volume of data upon which the model is built, it should trend such that teams who are poor defensively will have higher xGA.
My response was to a guy talking about breakdowns, though. Single events that get under represented or muted in the model.
Let’s get specific before we generalize
Consider the Malkin goal. Hanifin puts it to a Penguin at the blue line and disappears into thin air. Pen picks it up, passes to Malkin who walks in under no pressure and picks his spot.
Things like the shot position, rebound or rush (preceding events), and shot type may be factored in. The lack of pressure and absence of defence are not.
You put a defenseman in position getting stick on puck or forcing Malkin to make a quicker shot and the shot has the same xG contribution
An xG model is intended to reflect the value of the average shot, based on those measured parameters, from an average shooter, more so than reflecting the actual probability of that specific shot becoming a goal
|
No argument from me.
But teams with high expected goals against are generally bad defensively. If you have a low event game against things tend to go well.
There are many more parameters that could make the numbers better once they're added, but as it stands you aren't likely to succeed if you're giving up too many expected goals as an average.