Dallas Stars GM vs. the reliability of NHL stats

 (Photo by Glenn James/NHLI via Getty Images)

There isn’t much common ground between those at the forefront of the analytics movement and those who bemoan their value, outside of a mutual disrespect. But they’ve always agreed on one aspect of advanced stats in the NHL, which is that the data produced is only going to be as good as the data analyzed.

Hockey is a subjective sport, from the League determining supplemental discipline to referees defining penalties to official scorers defining what consitutes a shot. The greatest example of this, of course, is “hits,” which are handed out so idiosyncratically and inconsistently that they might as well be points assigned by Hardwick on “@Midnight.”

Corsi and fenwick are the most widely used metrics to measure puck possession, and rely on the stats reported by the NHL, which has four people tracking stats at every game. As technology advances, in theory so do the accuracy of those stats: The Dallas Stars, for example, use a computer program that does that work for them.

But the Stars’ coaches don't trust those numbers completely, and not just because Lindy Ruff is probably a single-finger typer. They go back over the game tape themselves to track stats, and according to the Dallas Morning News, they always find discrepancies:

The numbers flag certain elements the Stars staff might miss. From there, the staff examines the aspects further and comes to a decision that could determine a player’s future.

“If it’s a number you get out of right field, I think it’s great for analyzing that,” Nill said. “Why such a difference? If the numbers are closer, now you’ve got more of a comfort level to say, ‘OK, yes. There’s something here. We made this decision as a staff and the numbers back it up.’ It’s a lot easier to make a decision then.”

But it’s more than just following the numbers. Half a decade later, reliability is still a problem. The coaches’ numbers never coincide with the program’s, according to Nill. The focus instead is trying to find a common level of discrepancy. If the program’s scoring chances are always 20-percent higher than the coaches’ scoring chances, they know they’re on to something.

That’s why Nill is a proponent for the next evolution in hockey stats: Real-time video player tracking, from companies like SporsVu.

As he told Travis Yost:

“I want to see SportVu implemented league-wide. It’s something we’ve talked about at the GM meetings and the Board of Governor meetings. They’re going to start playing with a little bit — it’s in a lot of the NBA arenas, and we were following it closely during the World Cup as well. The league’s working on it and it’s something that’s going to come in to play down the road. You know, our sport — it’s so unique from other sports. You talk about these other sports, no one else has the number of changes we have during our game. There are changes every 30 or 40-seconds. The important part is getting the technology right when we do it, we want to do it right. And I look forward to that being implemented moving forward.”

(Check out this Lambert column for an in-depth look at real-time player tracking.)

The NHL is on board with video player tracking technology. COO John Collins sees SportVu (owned by Stats, Inc.) and Sportvision (makers of the NFL’s first-down stripe and, incredibly enough, the creators of the glow puck) as another way to open up the game to fans. The NHL is expected to test it this season and implement something league-wide by 2015-16, after giving the Board of Governors a sneak peak earlier this year. (We imagine the word “newfangled” was dropped once or twice.)

There are other innovations, too. The NFL just announced that its players will wear small sensors in their pads that track their movements during games, ones that might go as far as predicting the result of a pass had it been completed.

(No word if it will administer an electric shock to any player that does a choreographed touchdown dance.)

Advanced stats have made us all reconsider how to evaluate players and teams. The more data generated, the more ingenuity there’ll be in analyzing those numbers; and the more accurate that data is, the smaller the gap might become between the true believers and the puck luddites.