I watch a fair bit of hockey and am of two minds about the rise of statistical analysis.
Analytics can offer insight when comparing players but comes with its own set of biases.
Let me explain.
My favourite stat is one that analytics specialists scoff at: the Neilson number.
The Neilson number traces back to the late coach Roger Neilson, aka Captain Video, who in his day guided eight NHL teams including the New York Rangers and Toronto Maple Leafs.
It works like so: any time a player deserves some credit for a “scoring chance for” he gets a plus mark. Any time he bears some responsibility for a “scoring chance against” he gets a minus.
Neilson’s idea was that scoring chances are what count, at both ends of the ice, regardless of whether they actually result in a goal.
This is following the same logic as the analytics workhorse stat Corsi – but whereas Corsi uses shot attempts as a proxy for scoring chances, the Neilson number uses no proxy.
In order to cut out the middleman, it has to do one important thing: it has to be evaluative.
You need to watch a scoring chance over and over to figure out what went wrong or right.
The long-time Edmonton Journal hockey writer David Staples, a champion of the Neilson number, says he sometimes has to watch a chance as many as 20 times to parse it properly. (That is my experience as well.)
However, the evaluative aspect means the Neilson number is poison for analytics specialists.
It is not a proper statistic.
It is not inter-subjective.
That means, two people can look at the same scoring chance and describe it two different ways.
Observations that cannot be replicated regardless of viewer are of limited utility to statisticians.
Here we run into the core of the issue.
The rise of analytics is secure. It is now part of the mainstream, and many teams keep stats experts on staff or on call.
The old-school view, one that valued grit and character and the eye test, no longer stands alone.
But just as the old school valued only those measures that derived from human judgment, so, too, the new school is limited in that they value only those measures that derive from numerical analysis.
That is a flaw.
The measure that is truest is not necessarily the measure that is most statistically useful.
One can draw a parallel with the history of economics.
For many decades, from men like Adam Smith through to Thorstein Veblen, economic thought was far more a matter of philosophy than of number-crunching.
Then in the mid-20th century, as data became better and more widely available, the field went through a half-century in which numbers were everything.
More recently came a backlash, as behavioural economists forced the human element back into the equation.
My hunch is that something similar will happen in hockey. It cannot be all numbers, and it cannot be all impressions.
Someday maybe the Neilson number, which really is just a series of yes or no choices, such as did so-and-so cover his man?
Or did the player fire his shot from the danger zone near the net?
Did “X” win the battle along the boards that got the play started?
All those could become a replicable statistic.
This probably cannot happen until the league puts microchips in the puck and in players’ sticks or jerseys.
And that is something the league began experimenting with only this year.
rmckenzie@thenational.ae
Follow us on twitter at @NatSportUAE