Putting this here so I have half a chance of finding it later: I agree with Puck Prospectus that this article [PDF] is the most interesting analytical novelty of 2010 in hockey. I also don't know how much practical use it is, since you apparently need three seasons of data to make the numbers come out looking realistic. But I've been experimenting with a similar approach to isolating player contributions myself, and it wouldn't surprise me if three years just happens to be how long it takes...
Comments (2)
The paper on Henrik and Daniel:
"... the model has a difficult time separating the contributions of the twin brothers."
:)
Posted by Ryan Cousineau | January 28, 2011 6:14 PM
Posted on January 28, 2011 18:14
It's a serious problem! Pretty much the first one you run into: if two guys play together almost exclusively, their small-sample apart-results will be volatile and perhaps unrealistically extreme. There's no free lunch when it comes to isolating individual contributions.
Posted by Colby Cosh | January 28, 2011 7:09 PM
Posted on January 28, 2011 19:09