sure, but that's an different issue.
I don't' think they have quite the same issues, and my argument isn't necessarily against PDO per se, just advanced stats in general. And I fully realize this is a matter of opinion that's rather polarized, and others might not agree with it. I think that stats suffers a bit from what other advanced stats suffer from in terms of complicating it with other measured stats that may or may not have been heavily influenced by unaccounted for variables. There's just so much potential for variance, and situational factors at play.
I just don't think a regression towards the mean in most situations in hockey can be as helpful as when it's used in other settings, or other types of statistics. I think it has the potential to do more harm than good.
Take baseball for example. There's much fewer factors on each eventual play that you're measuring. The pitcher makes one play when he throws the ball. The hitter makes another when he swings, and then one or sometimes two outfielders have an influence after the ball is hit. The entire at bat or game events could have had a small influence, but no where near as much as the actual "plays" made during the at bat, and those inning/game factors come into play with hockey during a period or game as well.
In hockey, during a shift where a goal is scored, there's infinitely more "plays" that had influence on the the result. Did the goalie just make a big save allowing the play to keep going rather than a goal against happening. What kind of plays did the defenseman have to make to keep the goal out of the net. What kind of plays did he have to make to get the puck out of the D zone. What kind of plays did the three forwards have to make to set up the play. What kind of goalie were they shooting on. What penalties were possibly let go during that shift. What kind of coaching decisions like line match ups or combos did the coach make. That's just off the top of my head things that heavily influenced the play, and not even adding in whether a big hit or fight or something happened prior to that shift that gave them a tremendous amount of momentum, or x factors like battling through emotional play and pressure to make a clutch play, or confidence and chemistry between linemates.
These things just aren't accounted for in advanced hockey stats, and don't have nearly the influence they do in a game like baseball where the actions are much more a direct result of a few "plays."
I think in theory many of these stats have a chance to be fairly accurate in what they're trying to tell, because over the long haul a regression towards the mean can do this, but in hockey it requires a much bigger sample size than people realize, to weed out all the variance, and unaccounted for variables. Much bigger than one season's worth in many cases, and that's not the measurement most people use.
All our top guys are interesting this year vs last year. Duchene is missing the middle, Landy is missing a whole side. So weird. I know sample isn't exact yet but we are getting closer.
