Data, Pucks and Money: Analytics Applied to the National Hockey League

by   |   March 22, 2013 12:37 pm   |   0 Comments

Sidney Crosby 2009 photo by Michael Miller via Wikipedia 224x253 Data, Pucks and Money: Analytics Applied to the National Hockey League

Pittsburgh Penguins star Sidney Crosby. Photo by Michael Miller via Wikipedia.

TORONTO – A lecture on using data to analyze performance in hockey, the national sport, naturally drew one of the larger and more animated crowds at this year’s Predictive Analytics World Conference here.

While there weren’t any major penalties for fighting, presenters did describe an ongoing battle to convince the National Hockey League that implementing data analytics can build higher performing hockey teams and generate more revenue.

The economic stakes that are influencing decision-making in other professional sports apply here. With the average NHL team worth $282 million, and league ticket revenue of $1.2 billion in 2011, presenter Dan MacKinnon, director of player personnel for the Pittsburgh Penguins, said the “gut feeling” that once guided draft picks and trades is now “a little dangerous as we move into multimillion-dollar contracts and long-term deals.”

MacKinnon shared with audience members how the Pittsburgh Penguins teamed up with The Sports Analytics Institute in Park City, Utah, to create sophisticated computer models that measure just how much a player contributes to a winning team. These models, dubbed as the Player Evaluation System, consist of four metrics: Predicted Goals Scored, a player’s offensive and defensive contributions; Predicted Wins, a player’s offensive and defensive contributions in certain situations; Contributions to Winning, the specific number of games that each player contributed to winning; and Player Lifetime Value, which predicts the number of games that each player will contribute in the future.

Related Stories

Orlando Magic applies analytics to retain season ticket holders.

Read more»

Five elements of a data scientist’s job.

Read more»

Rethink your org chart for big data analytics teams.

Read more»

Use cases for analytics.

Read more»

In the case of Predicted Goals Scored, for example, Kevin Mongeon, co-founder of The Sports Analytics Institute and a speaker at the conference, explained, “Every time a shot is taken, we calculate the probability of it going into the net.” Analysts accomplish this by tracking the location of the player who shoots the puck, and assigning a value to each of the various areas of the offensive zone and the probability of a player scoring from each of these areas. Other determining factors include shot quality and what players are on the ice when shots are taken.

Coaches Can Use Data to Change Lines on the Fly
Armed with this information, Mongeon said a team can determine when to put a particular player on the ice to bring about a desired outcome. For example, he said, some players may prove to have a higher Contribution to Winning score when playing the top line as opposed to playing the third line. By being able to “predict what a player’s production will be in more difficult situations,” Mongeon said teams can increase their chances of winning.

But the NHL is still years’ away from the work described in the Michael Lewis book “Moneyball” that inspired the Hollywood movie. Part of the reason is that compared to the discrete actions (hitting, pitching) and set positions (infield, outfield, pitcher, catcher) of players on the baseball diamond, the fluid flow of play in hockey makes data collection more challenging, Mongeon said.

“The data is generated in a more complicated way in hockey,” Mongeon said. “The fundamental methods used in baseball don’t really apply in hockey.”

For instance, whereas teams and players play offense and defense separately in baseball, teams and players simultaneously play offense and defense, making it more difficult for number crunchers to assign values to individual players.

Familiar Business Challenges
Mike Boyle, another co-founder of The Sports Analytics Institute, said that NHL teams “are experiencing many of the same challenges as today’s businesses” when it comes to deploying an analytics system, such as poor adoption.

What’s more, Boyle pointed out that many NHL teams simply don’t know how to go about hiring the right people to launch and oversee an analytics system. “Hockey teams don’t really like to look to the data analytics industry for analytical talent. They like to hire a sports guy to be their director of analytics even though he doesn’t have a great analytics background.”

Complicating matters further is that many team owners are so inundated with calls from data analytics solution providers and consultants that, according to Boyle, “They just don’t know where to start. Everybody is pitching them something.”

But if there’s a dangling carrot that promises to drive adoption among NHL teams, it’s analytics’ broader business implications. Imagine, for example, being able to apply a tiered-ticket pricing system based on the anticipated performance of high-value players? Or enabling major arenas to use analytics to predict high-scoring games in order to better manage beer inventory levels.

For now, though, “We’re not at that level that,” said MacKinnon, the Pittsburgh Penguins executive. “Right now, the tribal knowledge with the NHL is simply to win.”

A Toronto hockey fan would add: Just as long as it’s not against the Maple Leafs.

Cindy Waxer, a contributing editor who covers workforce analytics and other topics for Data Informed, is a Toronto-based freelance journalist and a contributor to publications including The Economist and MIT Technology Review. She can be reached at cwaxer@sympatico.ca or via Twitter @Cwaxer.

Tags: , ,

Post a Comment

Your email is never published nor shared. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>