That's the word, "Context". I get the basic gist of the idea of "possession metrics" the longer you have the puck, the better, or the more you shoot, the higher your odds... but I also understand the context of someone firing pucks into a chest vs cycling around a goalie and taking shots into an empty net.
You're looking at it from a primarily offensive perspective. Why not flip that around? You're the team facing the offensively potent team. If you're a team that regularily gets shelled, you're at the mercy of the other team. A shot-metric stat can't differentiate between Jamie Benn and Mikael Backlund? Sure. Maybe. But on any given night you're going to face good players. Jamie Benn doesn't care who is guarding him - chip him the puck in his corner and he will go to work. The only way to actually shut him down is to get the puck, and take care of the puck. Then the other team is at the mercy of Johnny Gaudreau instead.
Something I've been curious about, is whether a certain coach will have certain types of stats purely based on play style? ie: Whether Hartley's system traded possession for high danger shots (collapse and explode) or if someone else's system trades high danger shots for possession (cycle?)
He presumed he did. But when you dig into it the Flames gave up more dangerous opportunites than they received throughout Hartley's tenure. And yes, coaching is fundamental to these stats in three ways:
1) How you defend as a team
2) How you break out as a team
3) Who is doing what and how often.
I have wondered about how this fits with players play styles as well as whether a player in different systems could see different play stats.
Skaters might see some change in their success as their roles on different teams change. Some of the Leafs skaters are good examples from last year to this. In general though, since individual stats should be used relatively to teammates, any boost one player gets, others who are relatively better should experience the same boost.
I mean, the above might sound like I get it, but I really don't. I can calculate it, but I can't understand how useful it is until I see reasonable applications. We aren't talking about a calculation of Celsius to Fahrenheit which a rule of thumb is close. There are some significant outliers. I can't wrap my mind that way because I don't have a mathematics or scientific way of thinking. Perhaps I will never know.
Of shot based metrics? Treat them as you would a "defense/dependability/tangible details" stat, if point based metrics are your "offense" stat. And since they are independant of luck, apply them as a likely-to-repeat element of a veteran player's impact.
However, as mentioned, I wouldn't mind a cheat sheet explaining when not to use a stat (for instance).
Don't use Corsi/Fenwick when the sole audience is an obnoxious, whiny peanut gallery. See posts above by certain poster.
Don't use raw Corsi stats for individuals, use CorsiRel and verify linemate effects with WOWY.
Don't use all-situations Corsi, ever. Compare across like situations.
Score-Adjusted Stats are preferred, as corsi rises when a team is trailing in score.
A player usage chart (QoC vs ZS vs Corsi) would a preferred way of comparing over a single stat.
Never use iCorsi or iFenwick unless you're actively examining an individual's specific shot generation. It is a stat that favours shooters over playmakers for obvious reasons.
For instance, I'd assume you'd be laughed out of a building for suggesting the idea of using Corsi/Fenwick on a goaltender, no?
Corsi was actually named after a goalie coach who measured shot attempts rather than shots, because even misses and blocked shots require reaction and physical exertion. It's a good way of comparing goalies' workloads and is to me preferable over comparing basic shot against because sometimes deflections off of intentionally wide shots are what score goals on NHL tenders and the best way to get deflections is with volume.
Fenwick Save Percenrage is a real stat (percentage of Fenwick attempts that did not go into the net) and in many ways when overlayed to something like xFSV% via adj.FSV% is a more accurate to a goalie's puck stopping ability. Goalies don't always stop misses, but since xFSV℅ is an analysis of what types of shots are taken in front of a goalie, missed attempts (like crossbars) still have a danger factor of going in. In many ways there exist missed shot sttempts that are more dangerous than "shots into dead centre of a goalie" and vice versa, so it's just more data.
For instance, my attempt at an easier way to understand Corsi would be in more layman terms... (correct if wrong)
Corsi = Shots your team takes vs shots the other teams takes
How about:
Events in which your team possessed the puck in the offensive zone and found an opportunity to direct pucks toward the net vs events in which your opponent possessed the puck in your zone and found an opportunity to direct pucks towards your net.
Gist: The more you shoot vs other team, the higher chances of more goals. Whoever shoots more is likely in the other teams zone more often.
How about "the more often you find an opportunity to create events directing pucks towards the net, the higher chance of your skill players finding the puck on their stick to make a play" and "the more often you suppress opportunities for the other team to create events directed towards the net, the lower the chance of their skill players finding the puck on their stick to make a play"?
Lastly I don't like to think of pure offensive zone time as the end goal. Hockey is a three zone game and most of us would agree that winning the neutral zone is huge. A team like Nashville is unspectacular in the offensive zone, and probably overrated in the defensive zone off the puck. But they win the neutral zone and by extension corsi. No, not by trapping '95 Devils style but by pressuring the puck and retrieving the puck and quickly getting it back up the ice. I don't think of the Preds as a cycling team - they just have the puck often because they don't turn it over and they don't waste possessions in their zone that they can nip in the bud in neutral ice.
Weakness: Doesn't take into account quality of shots,
Is a Giordano floater from the left point dangerous? It's not a high quality shot, but it could be double screened by Ferland and Bennett and go post-and-in, it could be deflected in by Tkachuk, it could bounce off a shot blocking Dallas Stars #2 right towards Monahan's stick, it could bounce off a stancion and end up on the right side of the ice on Gaudreau's stick with a wide open net. Shot quality has its value, and there are models that better correlate it to goal scoring, but at the end of the day NHL players are too skilled, too sharp, too coordinated, too quick, and too accurate to take your chances. If you can give yourself a 10% edge in shot attempts (55-45) as the L.A. Kings do, you are giving yourself a significant edge. Even if the shot quality is not top-tier.
the numbers can be messed up by quality of team mates. Great for forwards, but perhaps lacking for defensive dmen or goalies.
It's not "lacking" for defensive dmen of the glass-and-out nature so much as pointing out savagely that they are simply not very good defensively to add to the fact we already know (they are not very good offensively). Good defensive Dmen are not guys like Engelland, they are guys like (for a bottom pair) David Schlemko who may have more visually unappealling miscues but far fewer giveaways in general and and can help the team on its break out. It also points out that those roles in general are out-of-whack - why push a player who can't escape the defensive zone into the defensive zone? So he can shove his glove into the other team's scrub's face? Why not push a player who won't escape the offensive zone into the offensive zone instead? This is where some (not all) context based arguments fall apart.
"Engelland failed spectacularily by the stats - but it's okay because he was used in a shut down role" - they say while not identifying the fact that there are better approaches that better teams employ and better, if difficult to cope with, options for a shut down role.
"Justin Schultz had good advanced stats, but he was sheltered in Pitt and so it doesn't really matter what the stats say" - they say while not identifying that sheltered role as a successful one on a championship team that had no "Engelland" (ironically enough, his former team)
Fenwick is based on corsi and thus similar. However, it attempts to correct the weakness of the quality of shot issue in Corsi which presumably allows it to more accurately measure scoring chances?
Fenwick accounts for blocked shots. It doesn't account for shot quality any more than Corsi. Scoring Chances stats and Expected Goal stats account for shot quality - but the difference is not stark. Bad corsi teams allow more high quality chances and get less high quality chances - because they have the puck less.