Hockey analysts still racing to catch up to other sports’ stat mavens – Chicago Tribune
Jennifer Lute Costella, a graduate of Valparaiso University School of Law, put a career as a defense counsel on hold four years ago to stay at home with her young sons. When both reached school age, she didn’t return to lawyering and had time to expand an interest in hockey that began as a young adult.
“Now,” she said, “it has become a full-time obsession.”
To the tune of 60 hours a week, Costella found herself fixated not on being an impassioned fan of a team or the game but on learning how the sport was being analyzed in coldly numerical terms online. She had discovered dozens of people questioning the conventional hockey wisdom passed down over the decades when plus-minus rating was the cutting-edge stat.
With no background in statistics or data analysis, Costella, who now lives in Harper’s Ferry, W.Va., threw herself headlong into the brave, new, opinionated world of advanced hockey analytics.
Costella, 39, started on Twitter. After developing an idea that couldn’t be expressed in 140-character segments, she published a blog post titled “The Brandon Pirri Progression” on a personal website in November 2013. The reaction was positive.
Two years later, she still blogs at jenlc13.wordpress.com, with several new entries on the Stanley Cup playoffs. She said NHL teams have expressed interest in her work.
Costella is one of the rare women immersed in a field that a person present at its creation calls the “third generation of hockey analysis dudes.”
“When I started in 2003, I couldn’t identify anyone doing this online,” said Alan Ryder, of Toronto. “Now there is a large, new generation doing things in public, with others doing it in private for NHL teams.”
Ryder, who has a degree in statistics, became interested in applying them to hockey through Bill James’ groundbreaking baseball analyses, which have informed the way some teams (most notably Billy Beane’s Oakland Athletics) build their rosters.
“I found out that hockey was very poorly described statistically,” said Ryder, who cataloged data himself for an initial foray on shot quality, defining shot types and positing the likelihood of a particular shot becoming a goal.
Ryder, who said the Blackhawks approached him during the 2009-10 season with questions about his work, removed himself from the roster of major players in hockey analytics soon after that because he had a time-consuming job as CEO of Aurigen reinsurance company.
Whether these hockey analyses, public or private, become an important tool in assembling a team or determining its style of play remains to be seen. At the moment, shortcomings in data collection remain a significant obstacle.
But there are new analytical ideas already available on websites such as war-on-ice.com, puckalytics.com, hockeyanalysis.com, behindthenet.ca and hockeyanalytics.com.
And the mindset that hockey is too fluid a game to quantify usefully is disappearing, as evidenced by the Maple Leafs having hired a young hockey analytics guru, Kyle Dubas, as assistant general manager in 2014. Other teams have followed suit.
The Hawks? A team spokesman said they don’t talk publicly about analytics.
“I think information is good, and I think you’re naive not to take in everything around you,” said Lightning coach Jon Cooper. “If statistics is part of that, I want to have it at our disposal.
“How much we use and how much it changes your thinking … I can’t sit here and say, ‘Well, I’m 100 percent analytics says yes, so that’s our direction.’ The one thing analytics doesn’t take into effect is the human mind, the human spirit and the human soul.”
Starting small
For the record, plus-minus shows whether a player is on the ice when his team scores an even-strength or short-handed goal (a plus) or the opponent scores in those situations (a minus).
That leaves all circumstances related to the scoring of such a goal unrecorded. As such things were identified and quantified, the value of plus-minus to measure a player’s ability diminished.
According to Ryder, analysts began to have other statistics easily available about a decade ago, when an algorithm was developed to catalog the information in the play-by-play accounts available on the NHL’s website. They record events such as shots, hits, blocks, giveaways, takeaways, faceoffs and who was on the ice when such events occurred — a substantial increase in data over the basic shots on goal, saves and plus-minus.
With that information, goalie coach Jim Corsi developed the eponymous metric, Corsi, that has become a gold standard in the current state of hockey analysis.
It uses all shots in 5-on-5 situations — on goal, blocked and missed — to act as a proxy for puck possession, the idea being that the team with the higher Corsi number must have the puck more. And the theory is that puck possession in close games is a reliable measure of success.
Corsi touched off an advanced-stats boom, leading to the labeling of some three dozen other metrics to define areas of individual performance to be examined. Their merits are widely, heatedly and, of course, analytically debated.
“But we have outgrown the data being accumulated,” Costella said.
The problem is in verifiability, volume, and variety.
The play-by-play accounts are compiled by a home team’s stats crew, whom Kings coach Darryl Sutter once called “the retired gentlemen.” In a fast-moving game — Ryder says an average possession is between 10 and 15 seconds — they fall short of noting all the events. Costella resorts to tracking turnovers on her own.
At this year’s All-Star Game, the NHL experimented with chips in the puck and in the players’ jerseys that allow for finding where both are on the ice and the movement to get there at all times, which would provide reams of new data for analysis. NHL spokesman Frank Brown said the league expects to keep testing the system next season “as we work toward full implementation.”
The human element
Even with a record of that movement, much more will be needed to present a full statistical picture of what happens in a hockey game, and it will take eyes rather than computers to categorize it.
The problem is hockey is a comparatively chaotic game. Other than shootouts, hockey lacks the “discreet states” of offense and defense in “state-based” games like baseball, where they are completely separated, or football, where they are not only separated except for turnovers but also are subdivided into first-and-10, second-and-8 and so on.
“The closest analogy is basketball,” Ryder said, “but it has a finite number of offensive and defensive strategies and repeatable events. Basketball is way ahead of hockey on these sorts of stats right now.”
About Us
Free Shipping
If your order is $35 or more, you may qualify for free shipping. With free shipping, your order will be delivered 5-8 business days.
