I liked PLD's response to analytics. It more or less summarizes my view. The game happens too fast to be broken down correctly in analytics.
I just remember how many posters here thought they had a read on Neal Pionk, because of analytics. And how guys here had Jacob Trouba as the best defensemen in Winnipeg Jets 2.0 history, because his Corsi was high. How many wanted to break the bank for him? Because he shot a lot? Were they quality shots, or quantity shots? And did he make key mistakes during games, that negated his high shot volume? He sure as hell did, which is something that matter. Because you can make a difference in shot quantity, but you only have to make one defensive mistake, and the puck is in your net.
Turns out all those things, like kill plays, like passing efficiency, are important too.
I know there is a loyal following to what Garret posts, and I respect that. I'm not buying it though. Not fully. And I get quite irritated when people think they are smarter because they have mathematized the game into a science, without having a good read for what is going on on the ice. You certainly can't explain chemistry in numbers. That would be positioning, which isn't quantifiable data. If you play hockey you know the game is ebb and flow. The difference between one side and the other is marginal, most of the time, and the end results are one part talent, one part mistakes, one part execution, and one part luck. Certainly no one can predict how a goalie is going to determine the outcome. So while many stats are informative they are less than predictive.