There's two distinct things here. First, is the underlying technology SportVU. Second, is the analysis of it. There are hurdles with adopting both in the NHL.
First on the technology:
SportVU uses specialized cameras to automatically track the position of all 10 players on the court, as well as the ball. I assume it also tracks who has possession of the ball. In hockey the players aren't that much different to track (12 instead of 10, but that's a trivial increase) and they occupy about the same percentage of the playing surface, however hockey players have something that basketball players don't. Large extensions on their body in the way of a stick.
Stick position clearly affects how plays unfold on the ice, not only on offense, but especially on defense. Even if you don't touch the puck, a stick in a passing lane can prevent a player from even attempting a pass through the slot. Although that blends into the analysis area, it's clear that it's vital to track stick positioning, at least at a basic level, to understand defensive impacts.
Another challenge is in tracking the puck. A basketball is a ~29.5 inch sphere. A puck is a 1 inch thick, 3 inch diameter cylinder. Clearly that makes it much harder to track. As we all know from watching the war room try to figure out if a puck crossed the goal line, it can sometimes be really hard to tell where a puck is exactly. Granted, we may not need to know *exactly* where a puck is and might be able to settle for "It's somewhere in this foot or so" although that does introduce more unknowns.
Second regarding the analysis:
Assuming we can actually solve the issues above (I see no reason why it can't eventually be done), we have to analyze the data. In the NFL and MLB, there are defined plays. One team is on offense, one team is on defense. A play has a very defined start and a very defined end. In the NBA, although it's more flowing, there's still a clear differentiation between the majority of "plays" and for a given play (possession if you will) one team is on offense and one team is on defense. In addition, for any given play all players remain the same. The NHL is a different animal. Although you have stretches in the game where one team is clearly offense and the other is defense (power plays or other times of zone cycling), a majority of the game is the back and forth neutral zone play and teams can switch from offense to defense in a split second.
Although the NBA also has these switches with rebounds and turnovers, they're a lot more infrequent. You can still basically define a "play" as a clear possession of the ball and your number of play "types" are much smaller. Fast break, transition, half court offense/defense. 99% of a Basketball game can be classified as one of those 3 play types. In the NHL you could try the same method and call each play a clear possession of the puck, but you have much larger stretches of questionable possession during scrums along the boards, or bouncing pucks. You also have a much larger variety of play types. I'd say on offense you'd generally have defensive zone possession and breakout, neutral zone play, offensive zone entry and offensive zone possession. On defense you have forechecking, neutral zone play, zone entries, defensive zone play and zone exists (when just clearing the puck).
For the 3 play types in the NBA they basically line up with one another. If one team is in half court offense, one is in half court defense. If one is in fast break offense, the other is in half court defense. Hockey has a much more fluid matchup. Is forechecking a defensive play or an offensive play? Does chipping the puck in and going to get it count as forechecking or offensive zone entry? If one team is in defensive zone exit/breakout offensive formation (lets say defenseman behind the net) the other team might be in forechecking mode, or might already be in neutral zone defense. If you turn the puck over in the offensive zone, trying to prevent a zone exit by the other team is just as much an offensive play as a defensive play.
Mix into all this the fact that you have players switch out at "random" (Yes, I'm fully aware it's not random) and it changes if a play may be considered a success or failure. After a long shift, simply getting to center ice and dumping the puck in for a change would be generally considered a successful result. However, to dump it in and lose possession at the beginning of the shift would be generally a failed play.
That's not even to mention that hockey is the only one of the Big 4 where the players involved in the play is truly variable. NFL is a static 11 on 11. NBA is 5 on 5. MLB is 9 on 1 (1 on 1 and then a defensive response by 9 players, many of whom aren't involved in the play at all) with the minor addition of baserunners. In the NHL you have the "basic" 5 on 5 as well as 6 on 5, 6 on 4, 6 on 3, 5 on 4, 5 on 3, 4 on 4, 4 on 3 and 3 on 3. Obviously some are very rare (6 on 3, and 3 on 3 probably only have a handful of minutes through the entire league in a give season) but you have non-trivial amounts of time spent at other strengths, namely 5 on 4 and 4 on 4.
Again, none of this means the NHL *can't* use stuff like this. It just means that it's harder to do. We use corsi as a proxy for possession purely because we don't actually know who had the puck and for how long. We don't really know, on a macro/quantifiable level, if certain teams or players have strong possession numbers but poor corsi numbers because they cycle forever before taking a show. This type of data would change that.
I'd be interested to see how much of that is automatic and how much involved manual data manipulation.
