Beating a weak team can make your RPI lower than if you didn't play
them at all.

This is something which we've known about since the mid-1990s, back
when the NCAA was first using the current formula for RPI (25% win%,
50% oppwin%, 25% oppoppwin%), was documented at the time, and also has
been seen this season, but tonight's results are a clear illustration
of the problem:

Maine loses to BU and their RPI drops from .5949 to .5911
Cornell beats Princeton and their RPI drops from .5947 to .5895

Not only does Cornell get penalized for playing Princeton, even though
they beat them, they take a bigger hit despite winning than Maine does
for losing.

(The other results also affected strength of schedule, of course.
I'll post a follow-up soon on how to isolate the effects of these two
games.)

On the other hand, Cornell's KRACH RRWP goes from .8310 to .8313 as a
result of tonight's games (beating Princeton was overwhelmingly likely
already, so it doesn't help them much in the rankings) while Maine
drops from .8335 to .8221.  I maintain this is what should happen in
this scenario: the win over a much weaker team leaves Cornell's rating
basically unchanged, while the loss to a good team drops Maine
somewhat.
                                          John Whelan, Cornell '91
                                                 [log in to unmask]
                                     http://www.amurgsval.org/joe/

Enjoy the latest Hockey Geek tools at slack.net/hockey