Last week, curious as to whether Colgate's win over Cornell was the
biggest upset of the week, I started comparing the KRACH and CHODR
predicted outcomes of the game. This led to a script to cobble together
both ranking systems and output them for the weekly schedule. The results
can be seen at <http://gneiss.geo.cornell.edu/odds/>.
The two models tend to agree pretty well on the outcome of the game. Last
week's numbers, which aren't on the site because the KRACH is updated
continuously and I had to keep rerunning to get the bugs out, had a few
mismatches, but they were all on the order of +1/2.
In the comments in the source code, I included my dilemma over how exactly
to control the rounding with the chodr numbers. I would welcome a
convincing argument on that.
Pat Carr
Cornell '96
P.S. Colgate was in fact the biggest upset last week; Cornell was a 17:1
favorite.