Two of the most interesting sessions – at least as judged by twitter interest – at the 2014 Joint Statistical Meetings in Boston were the player tracking session (abstracts here) and the hockey analytics panel (here). Here’s a brief summary, with some relevant tweets.
A) First, some links
Here’s the .mp3 of our hockey analytics talk – we apologize about the sound quality!
Here are my slides on hockey’s point system
Here are Luke Bornn‘s slides from the player tracking session.
Here are Sam Ventura’s slides on quantifying defensive ability in hockey.
Lastly, click here for Andrew’s post, which links to talks from Michael Schuckers and Kevin Mongeon.
B) Eye in the Sky: The Player Tracking Revolution in Sports Analytics
Hearing the story of how Goldsberry’s NBA player tracking data came to be was really interesting. Dan Cervone is a graduate student in the statistics department at Harvard, and joined Goldsberry in a group interested in the NBA’s player tracking data. While he didn’t get a chance to speak at JSM, I have a lot of respect for Dan, and it was nice of Goldsberry to provide a walk-through of the student contributions to one idea for analyzing NBA player tracking data.
Given Goldsberry’s familiarity with player tracking data, I thought this was a really interesting observation.
The above image represents Prozone’s NHL player tracking software. Neat stuff. One unsolved issue that was discussed at JSM regarding tracking data in hockey is the uncertainty of which team possesses the puck. Unlike basketball, baseball, or football, possession in hockey is often unclear, and is perhaps best described in terms of a probability.
Intriguing comment from Kirk.
Related follow-up question: Is an academic career in sports research a career killer? (I hope not)
C) Statistics on Ice: Advances in Methods for the Analysis of Ice Hockey
Really interesting forthcoming work from Kevin, although this tweet wasn’t all that clear.
Kevin’s research has found that teams with several players from the same European country were more successful than those with players from several different European countries. However, when looking at individual shifts, it was not relevant for the players from the same European country to be playing on the same lines or on the ice at the same time. Thus, if there is a benefit to a diversified roster, it would be to due to a more welcoming locker room, and not because of a more comfortable style of play.
Because counting data in hockey is usually measured by a set of officials far above the ice, there is the chance for varying rates by arena. Ken Pomeroy has found similar results in college basketball. In any case, this was fun work from Schuckers, who channeled Edward Tufte with his graphics. I look forward to this work being formalized.
There are two aspects of player tracking that most statisticians aren’t ready for. First is the size – standard laptops are nowhere near big enough – and second is the spatial aspect. It will be interesting to see if, how, and when a familiarity with handling this type of data pays off for a franchise. My guess is that we are still a few years away.
Reblogged this on Stats in the Wild.