> ---Lee Urton <[log in to unmask]> wrote:
> >
> > The Pairwise Ranking is the ranking system used by the NCAA Selection
> > Committee to determine participants for the NCAA Division I men's hockey
> > tournament.
 
Jason Patton responded
 
> Hmmmm... is this technically correct?  I remember sitting in Lake Placid
> last year after Princeton won the tournament, and I essentially used RPI to
> figure out who was going to make the tournament and who was not (just like I
> have done for the past 4 years).  And I know I was able to figure out who
> was going to be in the tournament before PWR was published on a widespread
> basis.
 
The NCAA has told a number of people over the past couple of years
that the tournament is seeded according to the results of the pairwise
comparisons between individual teams.  This is not quite the same as
comparing the PWR, which is the total number of comparisons won, but
it's a lot closer to than either method is to comparing the RPI.
 
There is a fairly well laid-out method by which the NCAA seeds the
national tournament, which is explained in the following places:
 
http://www.slack.net/~whelan/cgi-bin/tbrw.cgi?pairwise
http://www.uscollegehockey.com/news/1998/03/19_selectionfaq.html
http://www.uscollegehockey.com/tournament/032097.html
 
With only a twelve-team field, it's not inconceivable that two methods
(RPI and pairwise comparisons) would yield the same results by
coincidence.  That's why trying to guess the method based on past
results is dangerous.
 
> Because if you ask me, they use RPI a lot more than PWR (we only get to PWR
> when teams are very close in RPI and even then, it is the team vs. team
> comparisons and not how many PWR points a team has).
 
I disagree.  The RPI, aside from being an ingredient in pairwise
comparisons themselves, is only used as a tie-breaking method when
individual comparisons are inconclusive (e.g. A beats B, B beats C, C
beats A).  The NCAA hasn't been completely unequivocal on what the
definition of the bubble is, but Joe Marsh's explanation strongly
implies they decide that on the basis of the pairwise comparisons and
not just the RPI.
 
Lee's statement is not precisely accurate because the committee uses
pairwise comparisons in a more complicated way that just comparing
total PWR (which why I'd like to see USCHO list all of the comparison
results in some compact table).  But PWR certainly gives a better feel
for the seeding process (as of 1996) than straight RPI.
 
                                          John Whelan, Cornell '91
                                                  [log in to unmask]
                                     http://www.amurgsval.org/joe/
 
"We've got a whole mess of penalties"  -- Cornell PA Announcer
   Arthur Mintz, after the Cornell-Union brawl 1998 December 4
 
HOCKEY-L is for discussion of college ice hockey;  send information to
[log in to unmask], The College Hockey Information List.