Finding a group of prospective engineers who want to attend a top-flight engineering school who also happen to be wicked-good hockey players, all at the same time....and then molding them into a cohesive unit instead of a group of talented individuals.
In this day and age, I wonder if creating a neural net software program and using detailed game analytics might give us an advantage over others.
For example, Houston Astros found a pitcher with an exceptional "spin rate" on his curveball, and then turned him into a great pitcher by using detailed analytics to help him understand when to use his pitch (e.g., through the curve more frequently to left-handed batters).
There are all sorts of technical details that baseball is using* that would translate really really well to ice hockey. As I envision it, it would be two-fold: individual skills development (what technical factors are key to winning face-offs? is it positioning? is it stick angle? is it observing when a particular muscle in the referee's forearm twitches, right before he releases the puck?) and also team positioning and movement (what is optimum spacing for a three-on-two rush? for a two-on-one rush? do you skate parallel to each other?, or should there be an offset? how far apart? etc etc etc).
It seems to me we will have to win as much on guile as well as talent because the "best" talent overall will go to the powerhouses that give them the exposure they need for the next step in their pro career. As far as I can tell, the best coaches are "intuitive geniuses" because, through observation and experience, they have internalized the kinds of "rules" I mentioned based on empiricism. We can take it one step further by applying detailed quant analytics to refine and target these detail areas, moving it beyond empiricism into actual metadata analysis.
* I mentioned "spin rate" for breaking balls; there is "launch angle" for batters. just a few examples.