Re: Winning versus scoring

• From: ehillema
• Date: Fri, 16 Feb 2001 21:18:01 -0800
• Message-ID: <000T8S2C9.001NFJE7J.12170551-quizbowl_at_...>

I am trying not to weigh in on the ongoing
discussion any more than necessary now, as I feel I've
already explained the basics of *what* NAQT does (and has
done for years) in the SCT performance rankings that
determine our invitations, and suggested *why* we chose to
stress the statistics we do. Whether we *ought* to do
something differently is another matter, and one for the
members of NAQT collectively to debate and decide; as
always we listen to and seriously consider all
suggestions that something could be done better or with
greater fairness.

But I had to make a response
concerning one factual point, again about what we do now:


Ahmed writes:
<<NAQT should consider *adding*
a term representing win-loss record to their
formula. [...] I don't think it would be too hard to
implement such a change.">>

Let me point out
that won-loss record is already a factor in our
rankings. Perhaps it should be a greater one; perhaps it
should take on added importance in comparison between
teams from the same tournament, to make less common the
apparent anomaly of our ranking teams from the same
tournament in an order different from that of the
tournament's final standings. But it misrepresents our ranking
criteria to suggest that taking won-loss record into
account at all would be a new departure.

As posted
in an earlier message about our process both this
year and last: "2. The remaining Division I teams from
the SCTs are ranked according to their SCT
performance by a statistical method based on tossup points
scored per tossup heard, adjusted for strength of
schedule; bonus conversion; and a small adjustment based on
won-loss record."

Ahmed makes his suggestion so
that "if two teams are otherwise very close, the team
with the higher win-loss record should still prevail
with a higher ranking." This is in fact already the
case--though we obviously can argue over what statistical
differences remain "very close," and whether the boost we do
give a team by virtue of a superior winning percentage
is so large as it ought to be. (Single game
differences in won-loss record will tend to have have only a
very tiny effect on our ratings; greater differences
begin to have a somewhat greater effect, in a
progression that is geometric, not arithmetic.)

Eric
H., NAQT ICT invitations coordinator