doesn't the final xgf% number get spit out by a regression that accounts for zone starts and quality of opponent, among many other things?
My main issue with its use here is that its predictive value is at its best once the sample size has reached ~half a season, the number means half as much after 8 games as it will in january
I could be wrong on either of these things, i last read papers on xgf% in like 2018
McLeod is 5x the player Cozens is right now for us
Isn’t xGF% supposed to account for deployment and linemates, QoC, that type of thing?
xGF/xGA is much simpler than that.
It assigns a value to every unblocked shot attempt based on the distance and angle to the net with the assumption of average shooting and goalie performance.
For example, a shot 5 feet from the net is worth more than a shot 20 feet from the net.
A shot from the goal line 10 ft from the net with a 1% angle to the net is worth less than a shot 10 ft from the net with a 25% angle.
Is xGF good at predicting the actual number of goals scored in any individual game? Not really. There's way too much variance in shooting skill, goalie skill and just dumb luck, especially with stuff like deflections and bounces. There's also the issue of NHL data quality. The shot attempt locations aren't always accurate from the score keepers.
Even with all of it's faults xGF/xGA is good at telling you who was better at generating/suppressing shot attempts from dangerous areas of the ice.