[robocup-small] objection to the group drawing

[Christopher Head]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Hi there,
I’m sorry to hear you’re not happy about this situation and that you’ve
been assigned to tough fields.

First, please understand that the field assignment is definitely
random, with certain constraints. Those constraints are (1) the top
four teams are distributed across the fields so they do not play each
other during round robin, (2) the next four teams are distributed across
the fields so they do not play each other during round robin, (3) weak
teams may be distributed across the fields so they do not play each
other during round robin, and (4) teams from the same country are
distributed across the fields as evenly as possible.

The field randomizer is available for anyone to download, build, and
run from <https://github.com/Hawk777/ssl-grouping>. Additionally,
Michael Bleier and I have both audited the code in grouping.cpp, and we
both agree that it is working properly: the random number generator is
decent, and the algorithm used to construct random permutations has
uniform probability distribution across all possible permutations (it
is quite easy to accidentally construct a permutation generator that is
not uniform by misusing array indices, but this code does not fall into
that trap). Please feel free to look at the code yourself; if you find
a bug that would have affected the results, we will be happy to look
into it and decide where to go from there.

As indicated by my June 10 e-mail, the two input integers needed to run
the generator were provided as the minutes into the games of the first
two goals scored in the FIFA Confederation Cup. According to the match
overview at
<http://www.fifa.com/confederationscup/matches/round=255199/match=300200118/index.html?intcmp=livecoverage_box_confederationscup_hp_en>,
this game had three goals, the first two scored at 3 and 48 minutes;
thus, the generator was run as “./grouping 3 48”.

Pending certain other issues being sorted out which we still need to
discuss further, I am not currently going to release the exact team
list file that we used as teams.txt for the grouping program; however, I
promise it *will* be released shortly, along with an explanation of how
it was produced, *if* we decide that the field groupings will stand as
currently published.

Most importantly, though, I would like to apologize to all teams for
NOT making sure that this random drawing process was done in a fully
transparent manner that would have prevented any possible appearance of
inappropriate activity. While I am personally convinced the drawing was
fair, I should have made sure better procedures were followed, namely
that the input team list should have been committed in public before
the seed numbers were generated, the same way I *did* commit to the
choice of seed numbers in my June 10 e-mail.

Thanks for your patience in dealing with this,
Christopher Head
SSL TC chair

On Thu, 20 Jun 2013 14:22:35 +0300
Feyza Varol <feyzavarol at gmail.com> wrote:

> We have been attending competitions for 5 years. We wonder how come
> we are almost always in the same group with Skuba (4 times in 5
> years) and Odens (3 times in 5 years). We do not complain being in
> the same group with these teams particularly. But we think that it is
> not random enough to be in the same group with the same teams. We
> guess that there is something more effective (e.g., the alphabetical
> order of teams) than randomness in the group drawing algorithm. We
> will appreciate if you clarify how the algorithm "exactly" works.
> 
> 
> On behalf of BRocks team
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.20 (GNU/Linux)

iF4EAREIAAYFAlHDd1sACgkQnfE3lq0v9Iz3EwD+IA/8+cHcbcME0IIhU4JhcT0Y
Ew4IQ97KcPSiXTiCYSMBAIHyHYeQGFS8Pu+L6YjYoJvkMyxpHlB5eAbgr2ZJ/aSp
=N0EY
-----END PGP SIGNATURE-----