[blml] Swiss Teams formats

[richard willey]

>On 4/2/07, Herman De Wael <hermandw at skynet.be> wrote:
> Interesting first results Richard,
>
> how did you determine the results of the matches? I assume that the
> further the teams are apart, the less chances you give the weaker team
> of scoring a lot?

Girben's the one who wrote the code (I didn't bother to look it over),
however, here is a summary of the basic methodology.  We are
(essentially) using the same methods that baseball and the chess
communities have adopted.

We assume that the performance of a given team can be modelled as a
normally distributed random variable with a known mean.  (The mean
corresponds to the skill level of the team).  When team X and team Y
compete against one another, randomly selected a "result" from their
respective PDFs.  You can then compare the two results to determine
which of the two teams won the board.  Repeat for N boards, and you
know who won the round.

Obviously, we're making a lot of simplifying assumptions.  For
example, one could make a claim that board results are auto-correlated
with one another.  Equally significant, the bridge scoring tables
aren't smooth (game bonuses kink the curses).  However, you need to
start somewhere.

Long term, once we manage to nail down the methodology we're going to
attempt to make the model more realistic.  (For example, we have big
data sets that we can use to properly parameterize the model)



-- 
Aristotle was not Belgian, the principle of Buddhism is not "every man
for himself," and the London Underground is not a political movement!
Those are all mistakes, Otto. I looked them up