The MIT Sloan Sports Analytics Conference is one of the premier gatherings of sports analytics in the country. The 2011 version was no different. Mark Cuban, Bill Simmons, Daryl Morey, Mike Leach, Brian Burke and Eric Mangini were some of the most notable names. And Malcolm Gladwell was there, too. This year for the first time there was a Hockey Analytics panel.
The panel was moderated by Katie Burke from Athletes' Performance. The rest of the panel included Stan Bowman (GM of the Blackhawks), Dan Fishman (Asst GM of the Capitals), Dan MacKinnon (Director of Player Personnel, Penguins), Jim Price (President of RinkNet), and Jeff Solomon (VP of Kings).
My overall impression was disappointment and, later, maybe a small amount of joy. My paper on a new method for goalie analytics was one of the finalists in the paper competition and so I was a speaker at the conference. The paper DIGR: A Defense Independent Rating of NHL Goaltenders using Spatially Smoothed Save Percentage Maps attempts to adjust a goalies save percentage by adjusting for the distribution of shots that they faced. That is, the DIGR ratings is the predicted save percentage if each goalie faced the league average distribution of shots accounting for shot location, strength of the opponent (even strength, power play, shorthanded) and type of shots (slap, snap, etc.). In the paper, I analyzed every shot from the 2009-10 regular season to develop the ranking.
My initial disappointment was in the lack of awareness of advanced statistical methods being done in hockey among the panelists. (Caveat: I realize there might be a competitive advantage to be gained by playing dumb.) MacKinnon has some awareness clearly but it is unclear to me if the others do. Maybe that is an opportunity for those of us who do this kind of work. I know that there were representatives from the Canucks, Lightning and Sabres in the audience for the panel.
Scott Cullen at TSN has a more detailed summary of the panel
I would also be remiss if I didn't mention that my friend Brian MacDonald also presented his work on adjusted +/- at the conference.
For those who've made it this far, the top 10 DIGR ratings for the 2009-10 season are given below:
1. Miller (0.9285)
2. Conklin (0.9280)
3. Halak (0.9269)
4. Hiller (0.9243)
5. Lundqvist (0.9237)
6. Nabokov (0.9227)
7. Bryzgalov (0.9226)
8. Rask (0.9218)
9. Niemi (0.9215)
10. Vokoun (0.9191)
Goalies who saw their DIGR drastically different from their raw save percentage include: Mike Smith (TBL) , Martin Biron (NYI) and Dwayne Roloson (NYI) who saw their DIGR be substantially larger than their save percentage and Jimmy Howard (DET), Tuukka Rask (BOS) and Tim Thomas (BOS) who saw their DIGR be substantially smaller than their save percentage. These kind of differences suggest that the shot distributions that these goalies faced were different from the league average and that they would do better or worse respectively if they faced the league average distribution of shots.
I'll post the 2010-11 DIGR ratings once the regular season is complete.
If you can find your way to Boston next year, it is a blast for anyone interested in analytics and sports.