A looks at the current pairwise comparisons at
http://www.slack.net/~whelan/cgi-bin/tbrw.cgi?rankings$pwc reveals a
difference between the idea of using the Pairwise Rankings (total
number of comparisons won) to choose the tournament field, versus
looking at the individual pairwise comparisons, which is what the NCAA
has told us they've done.  Right now, the 12th through 16th teams are
Cornell (19), Maine (19), Wisconsin (18), Clarkson (18) and
MSU-Mankato (18).  Within this "bubble", the comparisons are:

Cornell def. Maine, Wisconsin
Maine def. Clarkson, Mankato
Wisconsin def. Maine, Clarkson, Mankato
Clarkson def. Cornell, Mankato
Mankato def. Cornell

So Wisconsin, having won the most intra-bubble comparisons, would get
the #12 spot, not Cornell (or Maine).  Wisconsin's lower PWR comes due
to losing comparisons with Mercyhurst and Canisius, who are not in the
running for an at-large bid because they lose comparisons to everyone
else in the top 20 or so, and also likely because of the committee's
conference strength escape clause.

OTOH, there is a way of doing things so that Cornell would get the
12th spot: Mankato wins only one comparison on the bubble, so drop
them (since we're looking for the top team); in the resulting
four-team bubble, Cornell and Wisconsin win two comparisons, Maine and
Clarkson one each; Cornell wins the comparison with Wisconsin, so
they're on top.  The NCAA has never elaborated on just how the
committee handles a "non-transitive bubble" in which you can't order
the teams in such a way that the "better" team always wins all PWCs
with any "worse" teams.

Of course, the tournament field will also include the champions of the
MAAC and ECAC tournaments, and since no member of either conference is
in the top 11, who is #12 would not matter if the final comparisons
looked like the current ones.  (What would actually matter is who is
in the top 10, if there are no upsets in the other three conference
tourneys.)
                                          John Whelan, Cornell '91
                                                 [log in to unmask]
                                     http://www.amurgsval.org/joe/

Consider the alternative: http://slack.net/~whelan/cgi-bin/tbrw.cgi?kpairwise