Quote:
Originally Posted by Red John
It makes tons of sense actually. If you watch a game and your observation is "Joe Shmuck has a great scoring touch" then at the end of the night you look down and see he had 2 goals, that's great it corroborates your observation. But if you skip the game and see the boxscore, you are missing the context. What if one goal was an empty netter? What if the defenseman fell on a routine play allowing an easy tap in? What if he scored both goals being matched against the 4th line? By looking at stats only you only see the result - not what actually happened on the ice.
|
What you're describing is exactly the point of these so-called "advanced" stats. You're describing internal, random events within a game. What possession numbers are designed to do is to track much larger sample sizes (an entire season) of much larger sets of events (shots instead of goals). And yes, many of these stats do take into account things like playing against the fourth line - quality of competition is a very frequently used stat, and will often be combined with possession to adjust a player's possession statistics to account for how good the players he plays against are.
These guys aren't stupid, if you've thought of a flaw in the methodology that seems obvious to you, they've probably also thought of it and either tested it and found that it doesn't actually impact the results, or they've tried to find a way to account for it.
Quote:
Originally Posted by PeteMoss
Here's the thing... how do you build a team with high corsi? In theory would you not build the Oilers? A bunch of puck possession guys who can dangle and hold the puck?
|
No. The Oilers are a terrible possession team. They're 27th in the league. Why would it lead you to build that team? If I'm taking over as GM and want to improve the team I've just inherited from the perspective of puck possession, I'd probably sign guys who are a) good at it, and b) see their teams generate more shots when they're on the ice as opposed to when they're on the bench. There are lots of ways to look at this; pure fenwick, WOWY (measuring whether a player does better or worse with or without a particular teammate), Corsi Rel, etc. In other words, I'd probably sign guys like Justin Williams and Logan Couture.