Friday, March 22, 2013

Quick Thoughts on Moneypuck/James Neal/the Penguins/Dan MacKinnon/SAI


Yesterday, I came across a very interesting article by the excellent James Mirtle of the Globe and Mail, talking about the Penguins use of analytics and its significance in their decision-making process in trading for James Neal (and signing Tomas Vokoun).

Also interesting was that they referenced the Sports Analytics Institute — which appears to be some sort of boutique hockey analytics consulting firm, established in 2008, ran by a trio of mathematically-inclined professors. They appear to be pretty "confidential" in the work they do judging from their website, as it is intellectual property they're selling after all.

A notable part of the piece was discussing SAI's use of shot quality data.
SAI’s analysis relies on breaking the offensive zone into sections based on the probability of scoring from those areas and weighting for other factors such as what type of shots players are taking. 
The use of shot-quality data is still hotly debated within analytics circles and some of the concern is over the accuracy of the league-tallied location information, which can vary from building to building. Much of the other advanced statistics work being done, both for teams and independently, is more focused on puck-possession metrics that use shot attempts for and against to measure the amount of time teams and players spend in the offensive zone. 
Boyle and Mongeon, however, have found converts for their “predicted goals scored” system with the Penguins and one other undisclosed NHL team they said is also among the best in the league.And a lot of their success making inroads in the typically old-school NHL came as a result of the Neal trade.
As Mirtle says, many prominent bloggers are in pretty strong agreement of the redundancy of shot quality data over the long-term. I've read a bunch of posts saying why, but the chorus is almost too much in agreement for my liking, and a little too counter-intuitive. That's also not to say what SAI is doing is perfectly valid either; that other NHL teams used them isn't really an endorsement of their methodology — we all know the NHL is a copycat league. It's possible that the success of the Neal deal wasn't  direct result of their findings, maybe it just sort of happened to turn successful, just like countless other NHL deals that went down far before moneyball ever entered cultural lexicon. I don't have access to that data, so I can't really say this with any justification, and NHL teams probably got people way smarter than me to review the findings; I'm just spitballing.

Another thing I had remembered was briefly researching analytics personnel of NHL teams a while ago and I had actually bookmarked an interview from the 2011 combine (embedded below) with Penguins' Director of Player Personnel Dan MacKinnon, who was referenced in the article. He seemed like an interesting guy (his bio on the Penguins site) that came from an interesting path (followed Shero from Nashville) with some interesting things to say on their approach. I had also recalled he was a Penguins representative at Sloan in 2012 and found what he had to say about analytics in the Globe and Mail piece very interesting. Particularly in saying, “I don’t think we’ve made an impact decision since then without consulting the analytics.”

Not much more I want to say other than that, just that I find stuff like this interesting.

No comments:

Post a Comment