First of all, as has been pointed out in numerous forums, the quality of the data being recorded (particularly when it comes to the location and type of Shots taken) is less than you'd expect from what is supposed to be the world's greatest hockey league. For many statistics, the level of activity found in a game has more to do with where the game is played (in front of which official scorers), rather than the players on the ice. Secondly, however, there seems to me a fundamental gap in the design of these stats that would still hamper their utility even if accuracy wasn't an issue. Let's take a look at a random game that I pulled the Play-by-Play file for, a contest from November 25, 2005, wherein the Colorado Avalanche punished the Columbus Blue Jackets 5-0:
# Per Time Event Team Type DescriptionLet's start right off the top - we see that Columbus wins the opening faceoff, but then how does Colorado get a shot 33 seconds later? Was there a Giveaway by Columbus, or did an Avalanche player land a Hit or Takeaway? Are we to presume that after Brunette's shot the Avalanche picked up the puck again and held it for Laperriere's slapper at 00:49? It looks like the Blue Jackets grab the rebound off that shot and Fedorov takes a shot that the goalie holds for a faceoff. But after Sakic wins the next draw in his defensive zone, the Blue Jackets take a shot only 9 seconds later. How'd they get the puck to do that?
----- --- ----- --------------- ---- ---- --------------------------------------------------------
1 1 00:00 FACE-OFF N/A - CBJ won - neutral zone. COL 14 LAPERRIERE vs CBJ 17 BRULE
2 1 00:33 SHOT COL EV 15 BRUNETTE, Wrist, 21 ft
3 1 00:49 SHOT COL EV 14 LAPERRIERE, Slap, 39 ft
4 1 01:02 SHOT CBJ EV 91 FEDOROV, Slap, 35 ft
5 1 01:03 STOPPAGE N/A - Goalie Stopped
6 1 01:03 FACE-OFF N/A - COL won - defensive zone. COL 19 SAKIC vs CBJ 91 FEDOROV
7 1 01:12 SHOT CBJ EV 13 ZHERDEV, Slap, 24 ft
The ideal scenario would have details recorded for each event that occurs when a player controls the puck. For example, they might pass it (successfully or unsuccessfully), fire a shot on goal, bobble it off their stick (a Giveaway), or dump it into the zone. Given the speed of the action and the effort already shown in gathering accurate data, I wouldn't expect such a dream state to develop any time soon. It would be nice to approach something like Dean Oliver's Possession Scoring System for basketball. Since he's ascended into actually working for an NBA team, I doubt he's available to extend that model to hockey at the present time.
An acceptable stepping-stone towards that destination would be to at least commit to a team-level recording of puck possession results, and the current statistics come close to theoretically meeting that goal. For instance, we know the results of Faceoffs which start any given play, and the zone in which that faceoff occurred. We also know if the puck changes team possession due to Giveaways and Takeaways, or whether a Stoppage results (for instance, due to the puck going over the glass). I haven't found a true definition of a Hit, however; does it refer solely to a hard check dished out on the ice, or is a change of possession required as well?
When a shot is taken, however, our knowledge of which team next possesses the puck enters a fog. We know if there was a Blocked Shot, but who picks up the puck afterward? If the goalie makes the save, we know if a Stoppage occurs because he froze it, but if the puck squirts loose, what happens next?
I would propose that in order to close that loop, as well as provide a useful measure for defensive performance, the Rebound should be added as an additional statistic. It could be credited to the first player to take possession of the puck (making any sort of play with it, including simply firing it out of the zone) after a Shot, Missed Shot, or Blocked Shot. Every hockey player, coach and fan knows of the importance of battling for loose pucks, and those battles should be recorded in the official stats.
What we need to know is when and how the puck changes possession from one team to another, in order to reach a more thorough understanding of exactly how the game flows and scoring opportunities are created. For the defensive aspect of the game (which is largely devoid of useful measurements) the Rebound could provide a means of assigning some individual credit to team figures such as reduced shots against. The current state of affairs leaves us with incomplete and inaccurate information, and when I tried deriving a Puck Possession +/- last year, the results were disappointing.
It looks to me like basketball is light-years ahead of hockey in terms of the sophistication of its statistical analysis, and that the main obstacle standing in our way is the availability of high-quality data. Whatever the NHL is paying to have the Play-by-Play files done is 100% too much right now. If I had my five minutes with Gary Bettman, I'd just have to ask him, "why are you even bothering to publish this info"? It simply appears amateurish and hackneyed, as if the league felt compelled to put something out there, regardless of whether it was backed by any sense of reality.