Brian Burke, the president of hockey operations for the Calgary Flames, has a reputation as an old-school executive who doesn't care for so-called advanced statistics. When newly hired Flames GM Brad Treliving was asked recently if the Flames would incorporate possession metrics in player evaluation, he responded with a smile, "If I say yes, Burkie might punch me."
Burke, after all, is the same guy who two years ago appeared at the MIT Sloan Sports Analytics Conference and said that "Everybody is looking for these Moneyball breakthroughs... I have yet to see anything that has value in terms of an alternative way of evaluating players."
He's also the same guy who, in a radio interview just last week, said that "analytics to me is can you take raw data and give me an edge on drafting or trading for a player? No one has come up with one yet."
Burke, of course, is not the only NHL executive wary of using advanced statistics. Maple Leafs GM Dave Nonis, for instance, has shared a similar view: "The biggest thing we use is going to watch a player play," Nonis said at a conference last fall. "I haven't seen anything that's going to stop that from being the primary source of our decisions."
Statements like those, though, demonstrate a fundamental misunderstanding of what stats are and how they should be used, and it's a dangerous mindset in a sport where advanced stats are still in their relative infancy.
You can almost forgive guys like Burke and Nonis for not being blown away by the existing, publicly available statistics that get lumped under the heading of "advanced." There are possession stats like Fenwick and Corsi, which essentially look at how often a team is shooting the puck compared to its opponent. (It's been shown to be a reliable indicator of success.) There's PDO, which uses shooting percentage and save percentage to flag when luck (good or otherwise) is a factor. And usage charts can give an overview of how a team deploys its players, like whether they start their shifts in the offensive or defensive zone, or whether they're usually matched up against the opponent's top players. There are others, too, but compared to the mind-boggling amount of data available in, say, baseball, hockey still has a long way to go.
But even the stats we have now are worth paying attention to.
Consider Nonis' Maple Leafs. Advanced stats pegged the Leafs to be pretty bad this year, thanks to poor possession numbers the previous season. Worse still, their high PDO suggested they'd gotten pretty lucky the year before, when they qualified for the playoffs. Even before this season began, Grantland's Sean McIndoe identified them as a sort of test case: Could a team succeed even when all the fancy-stat metrics say they shouldn't? Turns out they could not: Despite a good start to the season, the Leafs regressed, just as the number-crunchers believed they would, and did not qualify for the playoffs.
And though a postseason series offers a small sample size, existing stats can help explain how a team like Colorado (which won its division despite a young, inexperienced roster) could be upset in the first round. The Avs ranked 25th in the NHL in Corsi, and consistenty lost the possession game in the first round to Minnesota, who eliminated them in seven.
There couldn't be a worse time for teams to shy away from using statistical analysis, either. Because that reluctance, coupled with a lack of urgency to find the next useful application of information, could lead to an even larger gap between the savvy teams and the dinosaurs ones as new, more complex data becomes available.
And make no mistake, such data is on the way. One of the reasons hockey has lagged in the realm of advanced stats is that it's difficult to quantify every movement on a hockey rink. Much of what happens on a baseball field, for instance, can be broken down into a series of individual events, and technology has made it possible to also track the movement of the ball as well as the players, to better study things like fielding. Similar technologies are available in other sports, as well: Both the NBA and the UEFA Champions League use a sophisticated system of cameras called SportVu to track and quantify everything that happens during a game. But buried at the end of this Fast Company article from earlier this year is this tidbit: SportVu is coming to hockey next. And other advancements in analyzing data are on the way, too.
In 10 years, we may well look back at the stats we now consider advanced and laugh. (If I may bring up baseball again, it's funny to think that when Moneyball was released a little over a decade ago, assigning value to on-base percentage was a pretty revolutionary concept. Now it's so commonplace that it's posted on stadium scoreboards.)
This shouldn't need to be said, but just in case: Statistics are merely a tool, and don't need to serve as a team's sole method of evaluation. The smart teams understand this, and they also understand that any edge could be huge. And the more data available, the more those smart teams will be able to use analysis to their benefit.
Consider the Chicago Blackhawks, winners of two of the last four Stanley Cup titles, and a team two victories away from a return to the Western Conference Finals. They've completely embraced statistical analysis, and are already tracking some of the things a hockey SportVu system could provide, like where shots were taken from and which player carries the puck into the offensive zone on a given play (not to mention the manner in which he does so).
"Stats are what they are," Blackhawks GM Stan Bowman told the Chicago Sun-Times last week. "There's no disputing who scored the goal, or who was on the ice for the goal. That's fact. What you do with that is sort of the real value. And I think there's an art to it. The analytics themselves are very objective. But then you have to do something with them and draw conclusions."
He continued, with a quote that must have made stat-savvy fans across North America smile: "What we do is different. I think it's better, but I guess it's a matter of opinion. It's also a competitive advantage. That stuff's readily available, but what we have is more proprietary. Which is why I'm really trying not to talk about it. I think what we do gives us an advantage over other teams. They might say I'm wrong, but we're pretty confident that what we have works."
Paying attention to statistics -- and actively trying to gain an edge by using them -- may not be the only reason the Blackhawks have been so successful in recent years, but surely it's helped. Bowman, whose team is using proprietary data while some GMs struggle to use advanced stats at all, gets it.
Executives like Burke and Nonis seem to be waiting for a killer stat or a perfect system before diving in, and they're explicitly saying as much. Burke has said that if there's an analytics system that he thinks works, he'll buy it right now. (For what it's worth, Treliving, Burke's new hire in Calgary, says that the Flames actually will use statistical analysis in some capacity, though he didn't go into specifics.) And Nonis said last year that the Maple Leafs don't even spend the money they've budgeted for analytics because there's nothing worth spending it on. In reality, though, it's not that easy. As Bowman says, using numbers successfully is an art.
This isn't strictly a game of finding the most useful stats, just like Moneyball wasn't a book about why on-base percentage is good and RBIs are evil. It's all about finding market inefficiencies -- identifying players that might be overlooked and hunting out bargains. The same number-crunching can help on the other end of the market, too, allowing teams to figure out which players are worth the risk of a big contract. It's also constantly evolving: Today's cutting edge analysis is tomorrow's widely used metric. By then, the smart teams have already moved onto the next thing, and thing beyond that. We've seen this play out in baseball, and if it's true that hockey is now where baseball was 10 or 15 years ago, we're going to see it in the NHL even more soon. Teams would be wise not to get left behind.