I have a different argument for weighting powerplay production less: volatility. I haven't done the regression but it seems that season-to-season variance in even strength production is much more stable than power play production. For one, even strength is just the default state of play: most of the game is played in this situation. Teams don't get even amounts of power play time (by extension neither do they have even strength time, but the relative difference in samples is much smaller), and individual power play production rates can fluctuate by integer multiples. There's value in repeatability.
I'm a bit of a heretic when it comes to faceoffs. They've been said to not be important because their regression coefficient is small when regressed against goals. But the same can be said for just about everything. Even shots and goals have a 10:1 relationship which is starting to get into the ''rare event'' regime.
What hasn't been mentioned in the hockey analytics literature, that I've seen, is that logistic regression is biased in the rare-event regime. Linear regression is not. One way that researchers have used to get around this bias is ''dependent variable selection.'' Essentially, eliminating some zeros so that the 1's and 0's in your dependent variable are in roughly even proportion (thus eliminating bias). This is essentially equivalent to reversing the conditioning on your probability: instead of asking what's the probability that you'll score a goal given that you won a faceoff, we ask given that we've scored a goal, what's the probability that we just won a faceoff. We can all name times off the top of our head where our team lost a defensive zone faceoff and it resulted in a goal against, or a big power play goal scored to send things to overtime.
That's essentially where the overemphasis on faceoffs comes from. The small regression coefficient just says that trying to increase goals scored on a team level by maximizing faceoff wins is an inefficient strategy. It doesn't mean they're not important though.
I haven't really thought of penalty killing yet. It's my hunch that I don't really want my top line players killing penalties.
To me, a #1C means that when he's on the ice, which should be a lot, your team has to score a lot. That separates the pretenders from the contenders fairly well. What separates them even better is scoring chances. If, when your #1C is on the ice, your team is generated 35 scoring chances per hour, your center's name is Matthews, or Crosby, or McDavid. I specifically value centers who are directly responsible for those scoring chances. Your center will often occupy the highest danger ice on an even strength shift, and therefore, having a pass only guy there seems stupid. The best centers in the game can all put the puck in the net, and try to do so regularly.
To tie this all into Kotkaniemi, Jesperi shoot a lot. About 3 shots per game, about 9 shots per hour, in a pro league.
