Dave Smith writes:
>2.  If I understand it correctly, subsequent performance is used to correct
>your schedule stength over time.  If you play a team that is 4-0 at the time
>of the meeting and then they go 4-25, your schedule stength should reflect at
>the end of the year that you played a 4-25 team instead of an undefeated one.
>I don't believe that enough teams on one school's schedule could disintegrate
>(for reasons other than team talent) enough for it to significantly affect
>their ranking in the poll.
 
BTW, I know this is just semantics, but remember TCHCR is not a "poll", it
is a rating.  Calling it a poll, as I've heard many people outside this list
do, makes it sound as if the method used is similar to the other well-known
*real* polls out there.  You know and I know what it is, but for those who
aren't yet familiar with it, it helps not to confuse them, so I suggest
people try not to refer to it as a poll.
 
Dave's point is a good one, but I would add that it is not just that the team
in question turned out to be 4-25, it is *who* they went 4-25 against.  If
their 4 wins were against good competition and their 25 losses were against
poor teams (an extreme), they may rate better than a team that did the
opposite.**  This is another reason why it is good to consider all results as
the season goes on rather than just results up till the date the game was
played.  TCHCR becomes more "accurate" as the season goes on because more
games are included, and we have a better idea of how good a team really is
as the season goes on - and again, this makes sense, because at the end of
the year we have a better idea of who the better teams are than we did at the
start.
 
** - This depends on quite a few factors, such as margin of victory, etc. so
that it is still possible for the team that beat poor teams and lost to all
good teams to be ranked higher.  Again, everything must be considered.
 
If you play a team at the start of the year, you really don't know how good
they truly are.  Some teams projected to be good won't be, and some Cinderellas
will surprise us.  If the team you beat in your first game goes on to chalk
up a 25-9 season, then clearly beating them is worth something - to the degree
of how they performed against other teams.
 
But, the question with regard to beating a team that may have struggled early
and then put it all together to become a contender is very important to
consider.  This is a shortcoming of TCHCR, if you will.  It is a reason why
TCHCR in its current incarnation will never be used exclusively to decide
who will go to the national tourney.  That's how it should be.  TCHCR is not
meant to be a be-all, end-all determination of rankings.  I see it as something
that in many ways is better than some rankings we have (polls), yet its main
strength is in sparking discussion that hopefully will lead to even better
ways of evaluating teams.  I said something to Keith a while ago about it
not being God - I forget the quote, but he liked it enough to use it, so I
am sure he agrees with me.  But I still believe it is better than any other
rating I have seen - and there are some weird ones out there.
 
I strongly suggest that anyone further interested in the basics behind TCHCR
search out the paper Keith mentioned as a reference in his posting.  It may
be difficult to find (took me a few months), but you should come away with
a very good understanding of how TCHCR works.  And remember that there are
still several variables that are key to TCHCR's functioning that are
arguable.  You may not agree with awarding 10 points to the winner, for
example, or any of the other values such as road win/ot win.  However, I do
know that these weren't entirely arbitrary and that Keith settled upon them
after quite a bit of research and testing.
 
I also suggest that specific questions on TCHCR's functioning be directed to
The Man (Keith) but allow him some time to put together a response.  I'd be
happy to try to answer any questions too, I think I have a good understanding
of it.  But discussion like this of the concepts is also good here since so
many people are interested in it.  This in particular was a very good thread,
IMO.
 
 
- mike