May 09, 2017
The Hockey Writers
For the third consecutive year, some of the best analytical minds in the sport gathered in Ottawa for the annual hockey analytics conference on May 6. Although advanced stats have been slow to catch on with the majority of hockey fans, Saturday’s conference was further proof that there exists a small but passionate community dedicated to the advancement of analytics in hockey.
One of the few presentations that didn’t look at individuals but more so the collective was provided by SMT “(the yellow line company,” known best for superimposing the first down line on NFL broadcasts). They used their 3D puck-tracking technology during last year’s World Cup to track shot types and precise shot locations and got a plethora of data.
Using this tech, they were able to measure the speed of various shot types (slap shots vs. wrist shots), though it was difficult for the technology to differentiate between wrist shots and snap shots. They also found that shot angle had a significant impact on shooting percentage, showing that lifting the puck increased shooting percentage by 1.5 to two times that of a shot along the ice.
Some of the other features SMT presented included the ability to track the quantity and quality of screens, as well as the quality and danger of shots and the ability the evaluate the type of shots being blocked. Each of these has exciting applications for the NHL, especially the latter which could add some context to a stat many in the analytics community are wary of.
While blocked shots are often used as a measure of defensive ability, Matt Cane of WinnersView proposed an alternative method of evaluating defensive ability by looking at passing data. Cane and his team found seven common pass types after analyzing the data, with cross-ice passes (what they call Royal Road) being the most effective in leading to goals.
Although they weren’t able to make too many generalizations, they found that players and teams who performed well in possession-based metrics tended to perform worse in expected goals off of passes, while the opposite held true of poor possession players. One example he used was that of Cody Ceci, who has a poor Corsi For percent but also a low expected goals rating.