On Sun, 18 Mar 2007, Dr. Bob Hamilton wrote:

> Seems they are still missing something.  Let's take an extreme case where a
> major player is injured and cannot play in the tournaments.  And to make it
> even more interesting, what if the injury happens in the last game of the
> season where the team loses the tournament championship and yet qualifies on
> the current PWR.  Or what if they win and a post-game brouhaha leads to
> significant DQ's, it seems they would still be in the regionals.  These are,
> of course, extreme examples but it does get at the importance of looking at
> recent performance.

That's exactly the sort of nonsense that's considered in the seeding
of the basketball tournament.  Tournament seeding should reflect the
position each team has earned through the results of their games, not
some expected level of performance in the tournament.  Are the
standings of an individual league adjusted at the end of the season
based on player injuries etc?

The NCAA never did that for hockey, although they did once have a
criterion that paid more attention to recent games.  But still only
game results.

> Seems this is different from strength of schedule.

It's completely different.  The point is that what the NCAA used to do
was just look at each team's winning percentage in their last 16 or 20
games.  But two teams may have played very different schedules in
those games, and going 13-2-1 in Atlantic Hockey is not necessarily a
better performance than going 11-4-1 in the WCHA.  (This is also true
in the "vs TUC" criterion, as was just pointed out, but at least
there's some selection of team strengths.)

Now, we did once propose a system to adjust for this in the various
criteria:

http://slack.net/~whelan/tbrw/tbrw.cgi?kpairwise

 					John Whelan
 					Cornell '91