Saturday, January 31, 2015

Strength of Schedule & Adjusted Statistics: An Unexpected Update

I've removed my previous posts on Strength of Schedule and adjusted statistics from the blog because I now believe that the method I was proposing does not always converge -- in some cases it appears that the adjusted statistics will end up with one team at 1.0 and all the others at 0.0 -- not a particularly useful approach!

That will teach me to post before I've implemented -- although it also goes to show that the process of experimentation is bound to be fraught with problems.  I usually avoid describing all the pitfalls, but the process of discovery is rarely as clean as the final product.


After some more work, I think I understand the conditions under which the approach converges, and I will go back, fix the original postings and put them up again.

No comments:

Post a Comment