I have some questions. Not trying to tear anyone down or make any larger arguments, I would just really like to know how these things work out.
I apologize if this has been answered somewhere in a C&B or Hockey Prospectus or BtN article or something, and if it has been please throw me a link and I will be more than sated. I've looked and not found anything, though. I also apologize that it's kind of a wall of text. It is the offseason, though. What else are you going to do? Work? Anyway..
The problem I have is this: what do advanced stats do to compensate for the fact that they can be recursive or circular enough to seem useless as an explanatory tool in some cases? The biggest example from this season that bugs me is CBJ's SoS stat. I have seen it argued that, "oh man, the BJs got the short end of the stick for sure, they had a super tough schedule in an amazing division, so they're not as bad as their W-L and points would make them seem (they are still pretty bad though)." However, the "strength" of the other teams in SoS is derived from their records.. and every team's (and Victor Stalberg's) record in the central division was bouyed by just straight up tearing Columbus a new asshole every chance we got. So which is it? Were they a half-decent team that could've done something in another division, but got the short end of the stick and were constantly outmatched... or is our "strength" stat overrated because we regularly got to devour an awful, awful team?
I also wonder the flipside about the Pacific Division: were they all bad, or were they all just pretty good and pretty evenly matched? Playoffs being crazy aside, the WCF teams were both from that division. We're all too happy to point out how feeding at the trough fattens up the Nucks' President Trophy hopes, but I only rarely saw someone bring up the fact that maybe everyone in the pacific had such low point totals because they were, essentially, sharing the points more evenly than in a division where one team feeds on the rest or the rest feed on one team, as in the Northwest and Central respectively.
A separate but similar problem applies to player stats, also. Let's say VAN is playing at home against a hypothetical team that runs 2 scoring lines (the first of which is also stellar defensively), a shutdown defense 3rd, and a 4th line that is never played because John Scott is inexplicably dressed on it tonight. How these lines are ranked for the purposes of Qualcomp by Corsi, Corsi Rel, or +/-, meaning the first line is the highest, the second is next, and the third looks like pretty crappy competition since they spend all their time playing against the best lines in the NHL, and while they ultimately cause those lines to score less than they would, the guys on the 3rd still end up being minus players. Even corsi rel only compares them to other members on their team, so they're going to look crappy out there against the scorers and barely holding their own vs the first line, who then gets to go against the less talented players and light shit up.
It would seem to me that it would behoove AV to get the Sedins as far away as possible from that shutdown line so they can score on the more defensively weaker second line. However, this move (which I would consider to be 'sheltering,' since it's getting the sedins away from the crazy-eyed axe-murderer-and-twin-telepathy-jammer on that hypothetical team), would mean they would have a higher qualcomp for that game, and our later statistical analysis could lead us to say "oh man, I guess AV trusts them more in defensive situations." Which would not really be the case, he just would rather run-and-gun against another scoring line than watch them get punched in the face and score less.
So those are my questions. They might be dumb, and they might be answered somewhere else already? I'm not sure. I'm not sure if my nail-biting can just be resolved by "well, over the course of the whole season (since not every team plays shutdown lines and teams do play outside their division), the stats that would be totally misleading for a few games game become reasonable metrics for how "good" a player or team is (though the problem of statistically differentiating offensive and defensive "goodness" of players still eludes me (and the NHL organization, if the Selke is any indication) in that case)."
Please let me know if there are good answers, bad answers, or if I'm just dumb (if so, how). Please refrain from having a stat person/eyetest shitfest in here, we've all read and/or participated in plenty of those and they're never productive.