Again, you can only do this for some of the things you're reviewing, but since you understand math, keep in mind that there are elements that cannot be untangled easily from other players on the ice, plus that hockey is basically a game governed by randomness. Over time, it fits Poisson distributions quite nicely.
The game is nowhere close to being something you can boil down to a bunch of stats.
I think we are in agreement. I like the idea of what he trying to do, but I am not confident that his process will ever replace "hockey smarts".
My understanding is that he would be very precisely reviewing video, in order to quantify actions and conditions and then begin to compile those data to look for patterns. Those patterns will begin to allow for insight into better practices on the ice.
Take a power play, for example. I can easily see an enormous amount of numeric data being generated in one 2-minute duration. Geospatial techniques will allow for tracking of each player - who is where at what time - this could likely be generated by video tracking techniques and presumably, this would be recorded for all players on the ice, probably even the officials (because their bodies sometimes get in the way). Speed and direction of all player body movement would be measured and quantified. So too would stick placement, head placement/direction, and any other physical characteristic of a player at any given time. Each "body" becomes an object that is tracked (and we'd also have object oriented data, like each player's weight, height, even age.) Now, track the puck. Where is it, where is going, who hit it, how did they hit it, how fast it is going in what direction, etc. Also keep track of time left on penalty or time in period.
Analysis packages exist that can then take those observational data and begin to compile them into sets of actions, reactions, expectations, etc. and then begin to deliberately parse out specific situations and results. (Traffic study helped to developed these sorts of geospatial analytical packages.) The analyst will then begin to ask questions, like what happened this time when three forwards hit the blue line at the same time? What happened that time when they hit the blue line staggered? What happened when the right defenceman laid back or maybe pushed ahead? When the puck is shot in on the right side, and the left forward presses against the boards to cut off the rebound, what happens? Does he intercept the puck and gain control, or does it scoot past him. The idea here is that eventually, patterns will emerge. And it will become possible to observed how various actions might more frequently result in preferred outcomes. The end result of the analysis is then to suggest those actions ... ie., pinch in on the left when the puck is shot in on the right.
Now, again, ALL of this is instinct. But the benefit of the analytical process is that it may possibly identify certain positive (or negative) results that may not necessarily have been apparent. Simple stats like penalty minutes, # goals and assists, time on ice, etc are completely meaningless to this sort of analysis. No, this sort of analysis requires a MUCH deeper, precise and purposive approach. That's my understanding of this body of analytics. I could be way off base, I dunno.
Im reminded of my dog. When we go to the park and I run away, she runs towards where I am going to be, not to where I am. Gretzky was a master at this. Skate or shoot to where someone will be, not to where they are now. Deek me out once, good for you. Deek me out twice, that's my fault. There are mathematical and statistical techniques that allow for us to understand and predict these sorts of behaviours, and results.
Is it worth doing? Absolutely it is. It is a replacement for a good hockey mind, absolutely not.