As a stats undergrad who's quite interested in this and has an active project going in the space, I think the main question here is what exactly are we trying to predict? Is the sole purpose of analytics to "rate this player relative to others"?<p>I think one of the things slowing down hockey analytics is the focus on finding a "new Corsi". In Baseball, given the game is segmented and the people who score points do so in a single action by themselves, it makes sense to have one metric that measures how much a players scores/their value. In hockey, maybe other types of analysis would be more accurate. What about predicting scoring rates (for/against)? Or the probability of winning particular games? Or simulation models of game outcomes? None of these things will give you a +/- type rating of a particular player - but they could be more accurate and potentially just as useful.