Ross Bracco writes:
 
> Why does the Index count a team's winning percentage for less than that of
> it's opponents' opponents?  It would seem to make you a better team based
> more on the company you keep than on your performance against them.
 
You note correctly that the formula is heavily biased toward strength of
schedule.  I think this has to do with our oft-discussed problem of lack
of inter-league play.  The formula attempts to measure a team's performance
relative to its schedule, based only on wins, losses, and ties.  Teams that
schedule strong non-conference opponents will do better than those who don't.
Note that strength of schedule depends on whom you play, but not how you do
against a particular team.  A win against a first place team counts for no
more than a win against a last place team.
 
> also, have you tried a multiplicative model, wherein the final factor is the
> product of different powers of the three base terms?  This would make
> exaggerations like extraordinary winning percentage or terrible opponents'
> winning percentage much larger factors in a team's rank.
 
This is an interesting idea; I might look into it in the off-season, or you
might want to try it yourself.  I did not devise the formula; I am using what
I have found in the literature.  The program I use wasn't terribly difficult
to write.
 
-- Erik