Longtime watchers of the National Hockey League know: there’s nothing so right that the NHL can’t find a way to do it wrong. So the League adding advanced stats to their website has gone pretty much par for the course.
On February 20, the NHL added metrics based on shot attempts for and against to their official site. The ideas behind most of these analytics have been around for years among hockey fans, but the NHL’s been somewhat remiss in acknowledging that. In an interview, Chris Foster, the NHL’s Director of Digital Development, acknowledged that fans and bloggers had “brought a whole new analysis, a whole new understanding to the game, and we're tremendously influenced by what they were doing,” but also stated wrongly that the NHL was the only place to find information on zone starts and primary versus secondary assists, both of them longstanding points of focus in the analytics community. An article by Ryan Lambert at Yahoo’s Puck Daddy site ran through some of the complaints that emerged about the site’s interface—for example, you can’t filter the NHL’s stats for game situations, or for minutes played (meaning rarely-used players with small sample sizes of ice time show up in strange places). And some have found discrepancies in the league’s numbers.
Still, it’s a legitimate step forward for the NHL. In hockey, as in other sports, observers have been trying for years to find new and better ways to measure players’ and teams’ effectiveness. Just as in baseball and basketball before, the new analytical measures for hockey were met with a mixture of defensiveness and derision from many people already in the sport, which then slowly gave way to a realization at least among some that there’s actually useful information to be found in correctly-analysed stats.
You can see how far behind other sports hockey is by the way a minor basketball controversy played out at about the same time that the NHL was finally adding advanced stats to its website. On February 10, Charles Barkley spoke out against new statistics that were changing his sport: “I’ve always believed analytics was crap,” he said, arguing that in baseball and basketball, teams built with advanced stats hadn’t won a championship. “All these guys who run these organizations who talk about analytics, they have one thing in common—they’re a bunch of guys who have never played the game, and they never got the girls in high school, and they just want to get in the game.”
Barkley’s comments prompted a thoughtful piece on Grantland by Bryan Curtis. Curtis felt that Barkley’s comments weren’t a part of traditional arguments over the use of analytics. After all, Curtis said, GMs and the media all know that advanced stats have value. He argued that Barkley’s comments have more to do with the way people talk about the game, and who gets to tell the story of a basketball game: the players, or the analysts. Do the players know what really happened on the court, or do the media have a better understanding with their statistical breakdowns?
For a hockey fan, Curtis’ piece is a sharp reality check. Because hockey’s nowhere near the state that Curtis describes in basketball. Again: according to Curtis, NBA front offices and the majority of the media recognize the importance of analytics. Hockey’s just not there yet — specifically with respect to the media.
Most NHL teams recognize that there’s something to the new stats. When ESPN recently ran a feature listing every team in every major league’s attitude to analytics, only one NHL franchise was listed as “nonbelievers” (the Colorado Avalanche) and three more as skeptics. Most teams have adopted analytics to some extent. Last summer, a number of teams hired prominent voices in the fan analytics community to crunch numbers for them. But when those teams—Edmonton, Toronto, and New Jersey—had poor results, some voices in the hockey media and among fans claimed that this was proof that analytics was hokum.
They were and are wrong. The reason teams have invested in analytics is because they work. But many veteran media people resist them for the same reason Curtis identified underlying resistance among NBA players: it challenges their ability to define the narrative of a game.
It’s certainly not the complexity of the numbers. As a humble former English Lit major, the new hockey stats don’t seem too terrifying to me. Most of them are based around the concept that to win you need goals, to get goals you need to take shots on net, and to get a shot you need to make a shot attempt (which might be blocked or go wide). Goals are relatively rare events, so it’s difficult to tell much statistically from them, but by the time you get to shot attempts you have something that happens relatively frequently. Shot attempts, called “Corsi,” track closely to possession time. Take Corsi, subtract blocked shots, and you have another stat called “Fenwick” which correlates strongly with scoring chances. Good analysts can compare Corsi and Fenwick ratings not just for teams but for individual players, taking into account the number of shifts which a player starts in the offensive zone of the rink, the quality of the player’s teammates, the quality of opposing players on the ice, and game situations (even-strength play versus power play versus shorthanded).
Then there’s PDO, which is based around adding shooting percentage to save percentage. Both shooting and save percentages have been tracked for ages, and are both just what they sound like—the percentage of shots that become goals, and the percentage of shots a goaltender stops. If you add the two things together league-wide, you have exactly 100 percent; every shot is either stopped or becomes a goal. If you do that for individual teams or players, you can find out who’s been lucky. What players are being victimized by never getting a save from their goaltender? What player that seems to have developed a scoring touch is actually just on a run of really good luck?
Other stats count scoring chances—shots from specific areas near the nets—and rates like a player’s goals and assists per 60 minutes of ice time. Conceptually none of this is difficult. Most of these and other advanced hockey stats are based simply on counting, not the relatively elaborate formulas behind baseball analytics. But even though these stats have consistently shown some predictive power (score-adjusted Fenwick rates during the last 20 games of the regular season will tell you who wins a playoff series almost 70 percent of the time), a number of veteran sportswriters, radio hosts, and TV analysts don’t like them. Because, I think, the stats challenge what they think they know, what they want to believe, and their ability to state confidently what’s happening on the ice. It’s a debate over how the swirl of action on the hockey rink is to be understood—and what values are most important in the sport.
A lot of younger and brighter voices in the hockey media world have embraced analytics, as have some respected veterans—James Mirtle of The Globe & Mail is an example of the first group, Bob McKenzie of TSN an example of the second. These are analysts interested in genuinely understanding what’s happening on the ice, and in determining which players are genuinely useful and which only look like they’re helping. Other media voices aren’t interested in having their prejudices challenged. Aren’t interested in considering whether, say, toughness and grit are as useful as they want to believe.
Of course those characteristics are useful. But they’re only two of many useful characteristics in hockey, along with speed, skill, and hockey sense. Sometimes tough players can be shown to have statistical use. Sometimes not. The problem is that there’s a general narrative about hockey, especially prized in Canada, which says that toughness is the way to win and if you’re lacking in skill you can make up for it in determination; if you just want victory enough you’ll get it. It’s about as realistic as a belief that the purest of heart will always conquer, but it’s a story in which a lot of people in the media seem to be heavily invested. Now things have reached the point where the desire to tell and re-tell this story is interfering with the accurate description of the game of hockey.
What’s frustrating is that these “new” advanced stats aren’t actually new. One coach, Roger Neilson, was counting scoring chances back in the late 1960s. Watching a DVD of the 1976 Canada Cup, a best-on-best international tournament, I was surprised to hear the commentators occasionally mention “shots at the net” along with “shots on net”—in other words, Corsi. And yet the contemporary hockey media’s slow to come around.
Two longstanding narratives, then. And a bigger issue: who controls the way the game is understood? Who defines the way we think about hockey? The media or the bloggers? Or the players? Of course, in hockey as in other sports, there’s an overlap between commentators and former athletes. Some are open to analytics; some, like Barkley, reject them. Do you want to tell a story that comforts you, or do you want to understand whether that story’s true?
Even to phrase the question that way is to establish a certain way of looking at the sport. The anti-analytics argument states that hockey’s too fast and complicated to reduce to numbers, with too many interrelated things happening on the ice at any one time. There’s some truth to that given the limitations of contemporary analytics, but it misses the broader point—the new stats are measuring results. Regardless of what specifically is being done, is it ending up producing shots, goals, and wins? How you get those things is less important than how many you get. Isn’t it?
It is, if you’re interested in wins. It isn’t, if you’re interested in telling a story. And a lot of media stories are just that, stories being told. In a story, the result is only one element of the whole. It can be ironic or predictable, it can help set the tone of the piece, but the way you get to the ending defines the story you tell as much as the ending itself does. The theme you have decided on determines what you as a storyteller do with the ending you get.
Analytics upend that. They focus on the result, and working back from there to try to understand what happened—as opposed to trusting one’s eyes to pick out the important symbolic moments from an entire game. Some people, usually employed by a team’s front office, are able to do that consistently; it’s not clear that most members of the media have that ability.
You could argue that there’s a different set of values implied by either approach. Many in the traditional hockey media, again especially in Canada, present the sport deeply seriously as an expression of national pride and identity. On the other hand, analytics would seem to be irreverent by definition; they’re essentially about looking at games critically.
Idiosyncratic whimsy’s even built into the very names of the stats. “PDO” sounds like an abbreviation for something, but it’s actually the handle of the message board poster who came up with the stat. “Fenwick” is somebody’s name. “Corsi” was picked as a name for a stat because a guy had a striking moustache. A blogger named Tim Barnes, who wrote about hockey stats under the name Vic Ferrari, heard then–Buffalo Sabres General Manager Darcy Regier talking about the idea of counting missed and blocked shots as well as shots on goal. Barnes liked the idea but thought that “Regier” lacked in euphoniousness. He looked through pictures of Sabres front office personnel on the team’s website, and eventually picked goaltending coach Jim Corsi as the man who’d give his name to the new stat. In an interview with hockey sage Bob McKenzie, Barnes was forthright about what went into his choice: “I just liked his moustache.” Ironically, Corsi actually was the member of the Sabres who’d originally come up with the idea of measuring shot attempts in the first place, as a way of tracking his goalies’ workload.
It all sounds like weird techie in-jokes, but then hockey’s full of weird in-jokes of its own, and doesn’t lack for eccentric humor (Google “Taro Tsujimoto” for another Buffalo Sabres–related example). Perhaps it’s unsurprising that analytics have slowly become accepted by the hockey community at large. The NHL, in its infinite wisdom, simply renamed Corsi and Fenwick when it put the stats on its website. Globe writer James Mirtle observed that the conference on advanced stats in major sports “now attracts athletes themselves, part of a new generation raised in the information age that views analytics as simply another way to gain an edge.” Mirtle cited a hockey example, Minnesota Wild star Zach Parise reading a paper on zone entries delivered at last year’s Sloan Conference. The athletes are taking on board not just information, but a new way of looking at the sport that the information implies. Because it helps them win.
Analytics aren’t magic, especially in hockey. Most statheads I’ve read have said, at one time or another, that these are early days in hockey analytics. Chance still seems to play a large role in the outcome of individual games, and even in the multiple games of a single playoff series. On the other hand, the NHL is promising an explosion of new data as soon as next season, based off sportvision cameras and tracking chips that will give detailed information on the location of the puck and players 30 times per second. Ken Campbell in The Hockey News reacted by proclaiming “the death of Corsi,” stating that the new information will make contemporary analytics obsolete.
It probably won’t. Information is one thing. Analyzing that information, correlating it with other information, and generally figuring out what it means and what to do with it—that’s what’s important. And how you use the understanding that results to think about and write about the game. Logically, people already hip-deep in analysing hockey numbers will have an advantage when a bunch of new numbers arrive.
What will happen then? What will change in what we think we know about hockey? Who will determine the new image of the game? I don’t know. No one knows. But I don’t see how it’ll be possible for even the hockey media to keep ignoring analytics. Just because a story’s satisfying doesn’t mean it can stand up against a true one.
—Follow Matthew Surridge on Twitter: @Fell_Gard