Craig A. McGowan asks:
>Since there seems to be very few inter-conference games this year, I wonder,
>would it be possible to measure the "connectivity" of the TCHCR weekly?
Here is something I could do. In a fully connected schedule graph, everyone
has played everyone else. For 45 teams, that would be 44+43+42+...+1 = 990.
There will be no where near 990 connections, especially since 1 team can
only play ~34 games. But I can count the connections, as I am calculating
TCHCR, and give you a percentage.
>For example, what would the effect of Union beating Lake Superior State be
>on St. Lawerence at this point in time? In a well-connected schedule, this
>might not make much difference, but right now, I think it would.
Yes, the more different connections there are, the less effect any single
connection has.
>What I am looking for is some statistical measurement which would tell us
>how much faith to put into the ranking.
I can give you a percentage of total connections, but I can't translate that
number to tell you how much faith to have. If I say 34%, then is that enough?
Is only 5% enough? I have no magic cut-off; I can only say that the more
connections, the better the rating.
I was already working on a "connection count" (or in hockey terms, the number
of different opponents) for each team for a future TCHCR set of notes, so the
total connections will be a breeze to calculate. Look for it at a computer
near you.
Keith
|