>In fact, if you compare the *standard deviations* of winning
>percentage, opponents winning percentage and >opponents-opponents-win%, it
appears
>that winning percentage has more impact on RPI than OWP
>and OOWP combined. Yes, it seems that the 25/50/25 system
>actually OVER emphasizes record.
 
Exactly. The key is that the NCAA uses RPI *at the end of the season*
when OWP and OOWP are very close for most teams. This is when interconference
play results become important, because they're
the major contributor to  variability in OWP and OOWP.
 
In the meantime, teams who have powerhouses on their schedule
early on look better than they should, and teams who start out
with less successful opponents are penalized.
 
Chris Lerch
EDS
RIT '84 & '91