Why the RPI is useless
The Ghost of Edward M. Kennedy
James Gibson
wrote:
> http://rivals.yahoo.com/ncaa/basketball/polls?poll=5
Any computer poll without preseason values is going to be useless this
early.
The Ghost of Edward M. Kennedy
> http://rivals.yahoo.com/ncaa/basketball/polls?poll=5
<
<Any computer poll without preseason values is going to be useless this
<early.
True, but the RPI is particularly so. How could Sagarin or Massey
put 2-1 Coastal Carolina at #7? Even MOV-less ELO chess wouldn't
do that.
--Tedward
Collins
wrote:
> http://rivals.yahoo.com/ncaa/basketball/polls?poll=5
The Pomeroy ratings are off in the weeds as well this early.
It would be interesting to see in retrospect how quickly they
converged.
I guess you could compare them to Sagarin along the way to get some
idea
of how much settling out they still had to do.
The Ghost of Edward M. Kennedy
>> http://rivals.yahoo.com/ncaa/basketball/polls?poll=5
> <
> <Any computer poll without preseason values is going to be useless this
> <early.
>
> True, but the RPI is particularly so. How could Sagarin or Massey
> put 2-1 Coastal Carolina at #7? Even MOV-less ELO chess wouldn't
> do that.
Sagarin has Duke's SOS at 147; RPI SOS is #1.
[Looks at schedule]
Eh, 147 looks about right. 4 patsies, ASU doesn't suck, and UCONN.
#1? Don't be ridiculous.
-Tedward
Milt Epstein
C'mon, we all know the RPI is useless (unless you're the selection
committee). But it's really useless to point out that it makes no
sense this time of year. Or to complain how, with your team having
an upcoming game against a total patsy, their RPI will go down even
if they blow them out.
I long ago decided the only way to think about the RPI without
driving yourself crazy is to consider it as a template with some of
the values pre-filled in based on a team's schedule, with the rest
getting filled in as the season progresses, thus it only makes sense,
such as it does, at the end of the season. It's not something that
gets built up incrementally, like Sagarin and most (all?) of the
other rating systems.
--
Milt Epstein
meps...@illinois.edu
The Ghost of Edward M. Kennedy
>>> http://rivals.yahoo.com/ncaa/basketball/polls?poll=5
>><
>><Any computer poll without preseason values is going to be useless
>><this early.
>>
>>True, but the RPI is particularly so. How could Sagarin or Massey
>>put 2-1 Coastal Carolina at #7? Even MOV-less ELO chess wouldn't
>>do that.
>
> C'mon, we all know the RPI is useless (unless you're the selection
> committee). But it's really useless to point out that it makes no
> sense this time of year. Or to complain how, with your team having
> an upcoming game against a total patsy, their RPI will go down even
> if they blow them out.
You can game the system by playing patsies that will eat up
their own shitty conference.
--Tedward
hughes
Unless you lose to them. Stop rubbing it in, Ted.