Quote:
Originally Posted by Bingo
An outside shot is low danger (a corsi event that doesn't register as a scoring chance or high danger), and then shot attempts inside the home plate as a scoring chance, ... same chance elevates to a high danger if it comes on a tip, rebound or direct pass.
|
Passing data is only incorporated by private data providers, such as SportsLogIQ. Unless you are paying tens of thousands of dollars for access, xG is not including passing information.
Quote:
Originally Posted by Bingo
They're all counted.
|
No, they aren't. Corsi includes blocked shots whereas no xG model includes blocked shots because the NHL records blocked shots where the block is made, not where the shot is taken. Do you mean Fenwick?
Quote:
Originally Posted by Bingo
Over the course of time each type of shot is graded by chance of going in and then builds expected goals.
|
Important nuance is publicly available xG assumes league average shooter by position and league average goaltender. For instance this is why Ovechkin has outperformed xG in 18 of 20 seasons. I suspect, anyway. Could be a hot streak.
Quote:
Originally Posted by Bingo
So in a game a player that is on the ice when the opposition is able to shoot more from in close, and on dangerous measures (tips, rebounds, wide open passes) then his xGA goes up and it generally matches the eye test.
|
Passing data is not publicly available.
Quote:
Originally Posted by Bingo
Yeah I meant all shot attempts are counted.
|
Not all shot attempts are counted. Blocked shots are not counted. As about 30% of shot attempts are blocked, xG excludes ~30% of data in a given game hence should never be used on a single-game basis. This is why Corsi is more positively correlated to goal differential in small samples but xG surpasses Corsi in larger samples (20 games or so).
Quote:
Originally Posted by Bingo
I've never heard an argument for why a measure that counts bad and good things would ever be useless or irrelevant in determining if a player had a good game.
|
Because the sample is too small to be representative as 30% of shot attempts are disregarded in xG. Corsi performs better in small samples, though all measures by nature perform poorly in single-game analysis.
I love stats, obviously, and the fact they are publicly available and free has been such a huge development for hockey. But, unfortunately, a small downside is the ease of misapplication when the numbers are copy-pasted rather than truly understood.