A week ago (on August 6th), I appeared on a panel on the Joint Statistical Meetings in Boston along with Brian Macdonald, Andrew Thomas, Sam Ventura and Kevin Mongeon. The panel was a good one on some of the advanced statistical methods that are being used in hockey. Fluto Shinzawa of the Boston Globe attended the panel and wrote up a piece on the panel that can be found at this link. A rough audio of this can be found at Andrew Thomas’ website: www.acthomas.ca/?p=62. Andrew was the organizer of the session.
I spoke for about 5 minutes on a model for Rink Effects that Brian Macdonald and I developed. Rink effects are something that has been known about in Hockey Analytics for some time. The issue is that there has not been a way to correct for the differences between the recording of RTSS at various rinks. This issue has been cited as one of the hurdles to adoption of hockey analytics. Here is an article with a quote from Chuck Fletcher of the Wild on data inconsistencies. Sportsnet.ca article Fletcher’s wrong about the recording of events in Detroit relative to Minnesota though Minnesota does undercount shots relative to the rest of the league. The bigger picture is that we need a way to adjust event counts. Thus, the goal of or paper is to develop tools to allow analysis to adjust RTSS events (Shots, Hits, Misses, Blocks) so that they are consistent from rink to rink. We’re not aiming for absolute truth (i.e. that what is counted as a hit is genuinely at hit) but rather for relative truth meaning that events are counted, on average, the same from rink to rink.
My slides from a short presentation at the JSM Panel are found below. We, Brian Macdonald and I, are working on finishing the final paper before I present the full set of results at the Royal Statistical Society’s International Conference in Sheffield, UK at the beginning of September.
(This article was edited 8/28/14 to add the link to the Sportsnet.ca article.)