Quote:
Originally Posted by The Cobra
I find some inconsistency in your report.
You spend a lot of time praising Zary, then mention his xGF% is in the 30%.
You conclude that Vladar had a good game , but his expected goals against was about 3, so he was actually -3 on the night.
I think on goaltenders we tend not to blame them when a goal is scored on a high danger chance, especially a breakaway. But in reality, even a breakaway should be stopped about 70% of the time looking at shoot out goals.
Both goaltenders were very poor last night, I think the difference was simply the skill of the AV players.
|
Here’s the deal with xG
xG is basically looking at every shot, taking its location, shot type, and depending on the model, some other events.
Then it calculates an xG for that shot, by basically saying that shot has the same probability of going in as the average shot, from the average shooter, that meets the same descriptive criteria, within the data set
You correctly note that breakaways are stopped ~ say 70 percent of the time, based on shootouts (ignoring that there is no back pressure on shootouts). But just think about when you watch a guy in a shootout put the puck into the goalie’s pad because he can’t elevate, or just puts it in to the goalie’s logo. Both of those shots have the same ~30% xG then, as the shootout attempts that are elevated over the pad, or sniped top corner. They get lumped in to that bucket of shots that fit the description.
Goalies stop about ~90 percent of all shots, and say 70 percent of breakaways.
xG models simply try to divide those shots into basic groups of higher and lower probability shots, based on some pretty basic differences. So breakaways are more likely to go in than the average shot
Obviously any single breakaway shot where the shooter puts it in a logo has no real life expectation of going in, but (using your 70 percent number) still has a xG of ~0.3. So the goalie’s GSAx changes by +0.3 on that poorly shot puck. If the guy undresses him, his GSAx changes by -0.7
The goalie covers about what, 75-85 percent of the net if they’re in good position? Lots of shooters simply hit the goalie with their shot. Goalie doesn’t have to do anything other than be there, and it’s a data point. xG ranges are based on large data sets, and have zero to do with where the puck is aimed
So Makar’s one timer blast placed perfectly inside the post has xG based on many shots, and the xG for that shot is tempered by the numerous other one timers that are blasted into the goalie, because they fit the same basic criteria.
MacKinnon’s breakaway dinged the GSAx by probably ~-0.7
Understand its applicability and its limitations and then you will be able to evaluate it appropriately