There’s a similar concept in soccer called non-shot xG (non-shot expected goals).
The models I’ve seen for it are pretty simple, based on the location where the ball is possessed. Advancing the ball to a more dangerous location (further up the pitch and closer to the goal, basically) from where you started earns NSxG, even if there is no immediate or subsequent shot attempt on that possession. The concept is similar to EPA (expected points added) in NFL football, which is based almost entirely on field position, down and distance and time remaining.
Here are a couple of overviews of NSxG in soccer:
The Power of Goals.: Non-Shot xG Models
The Next Level of xG: Expected Possession Goals — American Soccer Analysis
Applying the NSxG approach to hockey would reward offensive zone time, not just shot attempts, which would tend to even out some of the differences in attacking styles you suspect underlie the team-wide trends in G/xG across the last few seasons in the NHL. A classic Scheifele line shift that results in 30+ seconds of possession often in dangerous areas but in the end no good shot attempts (after the last incisive pass bounces over a stick in the slot, or whatever) would score highly by this metric.
But even that wouldn’t account entirely for more or less dangerous pass attempts or other offensive moves, if they weren’t completed and didn’t result in possession in a more dangerous area. A 2-on-1 that results in a shot into the pads would grade the same as one where a pass attempt from that same spot just misses the stick of its targeted teammate, who had an open net if he’d been able to reach the pass.
And a purely location-based NSxG model wouldn’t account for the value in hockey (which IMO is significant) of cycling the puck from one area of the offensive zone to another location that’s “only” equally (or maybe even less) dangerous, but where the quick switch in possession from side to side, or high/low or low/high, opens up a shooting angle and space to shoot. For instance Stastny’s pass on the PP last game from the front of the net to Wheeler on the side would probably lose NSxG (because Wheeler was further from the goal), even though that quick switch in play created a much higher danger scoring chance than if Stastny had shot it himself.
The positioning of the skaters and especially the goalie is so dynamic, and the details of their movements (not just where they are, but which direction they’re moving, and how fast) are so important to the actual danger of each shot or possession that it’s hard to believe a NSxG model will give an full account of the value of possession in hockey until those factors can be included.
