Which analytics offer best insight into hockey?
Teams can compile great numbers and still be losers, writes William Hanson.
Edmonton is lucky. Lucky because local TV, print and radio personalities discuss hockey analytics, and they do so a lot. As an academic, former NCAA athlete, and round-the-clock sports enthusiast, I particularly like articles by David Staples, who’s cautiously optimistic, yet appropriately skeptical, about analytics and their rightful place. Over the weekend, his online article “Shots Fired! Ex-Oiler Winger Blasts Analytics Guys for Undervaluing Intimidation” inspired me to join the fray. This article, like so many others, questions the role of analytics. It also questions the role of toughness and physical play. Staples says analytics adds “spice to hockey debates and perhaps some amount of insight” and I agree. He also says, “I like numbers as much as the next hockey nerd (even as I don’t agree with many other hockey nerds about which numbers are best (italics added) …” This is the core issue: identifying the best, most meaningful numbers.
GOOD, BAD, AND UGLY NUMBERS
To paraphrase Dr. Anne Anastasi, one of the all-time great assessment experts, it’s not whether analytics are good or bad, it’s how they’re used. When used properly, they’re extraordinarily good, as statistical/actuarial prediction outperforms professional judgment or, in sports parlance, scouting. But you can’t just gather stats for stats’ sake. You have to gather meaningful stats. Otherwise, it’s bad. By meaningful, I mean stats associated with desired processes and outcomes, like grade-A scoring chances, wins, etc. You can’t pick stats willy nilly. If you do, you’re merely “fishing,” which can be misleading, even ugly.
TAKE FACEOFFS, FOR EXAMPLE
Faceoff win percentages, shots on goal, etc., are popular water cooler topics, and they’ve garnered considerable media attention. Regarding faceoffs, it’s probably because the Oilers finished dead last in 2016-2017, winning 47 per cent according to puckbase.com. Based on my quick-and-dirty analysis, this year’s regular-season NHL team average was exactly 50 per cent, with a standard deviation of two per cent. So, the Oilers were significantly worse than most teams, statistically speaking. Seems like a big-time problem, right? Turns out, there’s no meaningful relationship between team faceoff wins and regular-season points or standings. If anything, it was inversely related this past year (correlation coefficient = -0.2). To underscore the point, consider this: the Stanley Cup Champion Pittsburgh Penguins finished 28th in regular-season faceoff wins and the hapless Colorado Avalanche finished second. Granted, it’s a small sample size; essentially an n of 1. Five, 10, or 20 years of data might suggest otherwise, and, in the end, show a positive relationship. Alas, it’s food for thought and brings us full circle.
WHAT NUMBERS ARE INDEED BEST?
This is an empirical question — one hockey analysts should answer resoundingly, with rock-solid data. Maybe they have. My sense, though, is they haven’t. Otherwise, we wouldn’t still be debating sports analytics. By analogy, in psychotherapy, hundreds of theoretically based factors could conceivably affect positive treatment outcomes. After 75 years of research, though, we know two factors standout: therapist empathy and therapistclient collaboration, as they affect change the most, making them two of the best, most meaningful numbers to assess. I suspect it’s the same in hockey. Certain factors — or clusters of factors — affect outcomes the most. This leads me to the take-home point. Although it’s fun talking about seemingly important stats, like faceoff wins, shots on goal, etc., they may not relate to things that matter, like regular-season points and, perhaps, playoff wins. In any event, for now, let’s keep rolling along, riffing on analytics and enjoying Staples et al’s spiced-up debates and most-excellent insights. I know I will.