As a graduate student at Carnegie Mellon, Samuel Ventura helped the Tartans club hockey team win a 2015 College Hockey East championship recently. His biggest contribution to the sport, however, might take place off the ice.
As a graduate student studying statistics at CMU, Ventura co-founded the hockey analytics web site War-on-Ice.com in August 2014. His goal was to develop enhanced statistics that could help teams analyze player performance beyond conventional methods, such as goals and assists.
Ventura, a newly hired visiting assistant professor at CMU, belongs to a community of online hockey fanatics leading a movement to integrate enhanced statistics into player evaluation in the NHL. Similar to sabermetrics in baseball, enhanced statistics have slowly attracted mainstream attention, and, within the past year, a flurry of analytics-based hires has swept the NHL.
“Even if hockey is 90 percent random, smart teams would do whatever they can to gain an advantage in that remaining 10 percent,” Ventura said. “Baseball is random, too. That’s why we call baseball statistics ‘averages.’ ”
Penguins general manager Jim Rutherford made at least one such hire, bringing in Jason Karmanos as vice president of hockey operations. The Penguins, however, declined interview requests for this article.
As early as 2006, hockey bloggers started presenting new statistics as a way to evaluate team and individual player performance. It started with metrics that tracked offensive output by measuring how many shots a team generates with a particular player on the ice, a way to justify puck possession.
The metrics supported the idea that teams that take more shots not only generate a higher number of scoring chances, they limit scoring chances for the opposing team by holding onto the puck.
NHL teams eventually saw value in the new statistics and started to hire analytics consultants, including many of the bloggers who started the movement. James Mirtle, a hockey writer for the Globe and Mail newspaper based in Toronto, said today about 22 teams use analytics. Five years ago, it was 10 or fewer. Several of those teams hired analytics consultants before this season.
The NHL recognized the growing interest in enhanced statistics and partnered with German-based data company SAP to give anyone access to accurate enhanced statistics in as close to real time as possible. The league rolled out the enhanced statistics page Feb. 20.
“We think we have the greatest sport on the planet, and one of the reasons why is because it’s so fast,” said Curtis Foster, director of Digital Business for the NHL. “If we give [fans and analysts] the tools to slow down the game, we can really appreciate what these athletes are doing and the amazing stuff they can do over time at this speed.”
In many ways, the movement continues to evolve. For instance, one of the original enhanced statistics, the Corsi relative score, measures a player’s shot attempts — including shots on goal, missed shots and blocked shots — compared to shots generated by the team when that player is not on the ice.
Ventura, recently named an assistant coach for CMU’s club hockey team, said that metric fails to account for the quality of a player’s teammates among other variables, but he has developed a model to put those numbers in context.
For example, the Corsi relative score — which the NHL calls “shot attempts relative” on its Web site — shows the Penguins generate offense at the highest rate with left winger Chris Kunitz, not star centers Sidney Crosby or Evgeni Malkin, on the ice. Kunitz leads the Penguins with a 5.5 Corsi relative score and a 10.2 Corsi relative score per 60 minutes, which means that if Kunitz is on the ice for all 60 minutes of a game, the Penguins would generate about 10 more shots than if he were on the bench.
That might be one reason why Kunitz, despite a below-average season, is still a central part of the Penguins offense.
“Just because Crosby’s not leading the league [in Corsi relative score] doesn’t mean he’s not the best player in the world,” Ventura said. “At the same time, just because a player doesn’t produce a lot of offense doesn’t mean he doesn’t have value.”
But enhanced statistics go only so far, and they have their fair share of critics. Traditional methods of player evaluation often supersede newer models.
For example, Penguins center Brandon Sutter has a minus-4.6 Corsi relative score, one of the lowest on the team. But he is tied for second in the NHL with four short-handed goals this season. Better even-strength players exist, according to the enhanced statistics, but Sutter remains on the roster as a penalty-killing specialist.
The enhanced statistics only add a tool to evaluate players more objectively but fail to completely replace traditional methods of evaluation. When used properly, enhanced statistics serve a purpose beyond simple player analyzation.
“If a hockey analyst can use quantitative evidence to save a team from making one bad free-agent signing,” Ventura said, “they’ve already paid for themselves several times over.”
First Published April 11, 2015, 12:00am at https://www.post-gazette.com/sports/penguins/2015/04/11/Analytics-gaining-foothold-in-NHL-too/stories/201504110015
Jacob Betzner 