Rick O'Donnell writes:
>(1)  One thing RPICH (nor any other computer ranking) does not consider is
>"recency" of a win.  When the NCAA Hoop tournament field is announced, they
>usually mention a team that slid out because of recent poor play, despite a
>better season-long performance compared to other teams that made it.  In a
>human poll, recent events weigh a little more heavily. If recent level of play
>is indeed important (and I'm not saying it is, I'm not sure), why not weight
>the more recent games with a slight factor to account for this?
 
This is already a factor of its own.  Here's what happens in a
nutshell:
 
1) RPI is calculated and the .01 threshold is employed to figure out
which teams' ratings are close enough that other factors should also
be looked at.  For example, from the latest RPICH:
 
 4  Minnesota           16   10- 5- 1   .6563   .5748   .5217   .5819
 5  Denver              16   11- 5- 0   .6875   .5435   .5210   .5739
 
.5819 - .5739 is less than .01, so Minnesota and Denver are close
enough that other factors will be looked at.  (BTW, Brown's RPI of
.5709 would be close enough to also earn it a comparison with each of
these two teams.)
 
2) One of these other factors is record in the teams' last 20 games.
Others are record vs common opponents, head to head record, and
record vs teams under consideration.  You get one point for each of
the criteria in which you have an advantage over the other team.  Head
to head carries a point value equal to the net difference in the number
of head to head games.  For example, in a Minnesota-Denver comparison,
Minnesota is already down 2-0 because DU has a 2-0-0 record against
the Gophers.
 
I have named four factors here.  The RPI is the fifth.  (This is one
of the chief arguments I have with the process...if RPI is deemed
close enough that other factors must be brought in, then RPI should
be thrown out at that point in such a head to head comparison.  It
gets worse below...)
 
3) If the two teams are tied after step two - let's say Minnesota got
one point for RPI, CommOpp, and TeamsUnderConsideration for a total of
3, while DU got 2 for HeadToHead and one for Last20, also for a total
of 3 -
 
Then the RPI is the deciding factor, no matter how small the
difference is.  Minnesota would receive the nod.
 
Where we tend to hear that recent performance hurt a team, in hoop,
hockey, or whatever, is when the team loses by one point to another
team and Last20 is one category it did not win.  Basically, the
committee is saying that recent performance is important, but not as
important as some other things.
 
But to be honest, I have not followed the basketball selection process
closely enough to know in what ways it may differ from hockey.  I am
under the impression that it is very similar, but that's just an
educated guess.  We have been able to apply the process for hockey to
figure out ahead of time what the seeds will be...I wonder if people
who follow hoop are able to do the same.
---                                                                   ---
Mike Machnik                                            [log in to unmask]
Cabletron Systems, Inc.                                    *HMM* 11/13/93