herman
Well-known member
Significantly Insignificant said:
There's descriptive results and predictive metrics. This one is, as Heroic Shrimp explained very well, the descriptive result, a couple steps more complex than a W-L record or Sv%. A high GSAx correlates strongly to Sv% and Ws though.
If you want a predictive measure, you'd have to compare this to his historical GSAx, but it will be a high variance estimate given his rather small sample playing on this team. Vasilevskiy is often near the top of these charts, for example. If his current year numbers are out of whack with his past performance, there's a significant element of luck involved (or playing through injury, etc.).
Goalie evaluation requires a good deal more input to develop a foundation of context to anchor results reliably. Public models will struggle with this because there just isn't enough data available in the NHL scoresheets for things like rebound control, screened shots, tipped shots, rush chances, and how all of those figures are skewed by the team playing in front of the goalie.
The term Expected here refers to an NHL average shooter vs NHL average goaltender (sampling all the shots from that spot in previous data and seeing how often it went in).
