[robocup-small] New Field Drawings

Christopher Head chead at chead.ca
Sat Jun 22 04:13:32 EDT 2013


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

Hi everyone,
Due to a number of problems, we are retracting the field assignments we
previously published:

(1) Immortals was included in the list of teams, even though we had
been informed ahead of time that they would not be attending the
competition.

(2) The choice of which teams to seed was made improperly, not
according to the published rules for seeding.

(3) The process was not fully transparent—although the method for
choosing random seed numbers was published ahead of time, the list of
input teams for the generator and the source code to the generator
itself were not published.

(4) The way we chose seed numbers was not ideal. FIFA games are 90
minutes long, so using the number of minutes of goals allowed the seed
numbers to be only between 1 and 90; using two numbers in that range
provides only about 13 bits of entropy. The actual entropy would even
be slightly lower (about 12 bits), because the first number will very
often be smaller than the second. This is not enough entropy to ensure
that all valid field assignments could potentially be generated.

To address these issues, we will be rerunning the field assignment
process tomorrow, Sunday, 2013-06-23.

The source code to the generator program is published at
<https://github.com/Hawk777/ssl-grouping>. This program takes as input
the list of teams, with each team assigned to a bin (1 for top four
seeded, 2 for next four seeded, and 3 for all the others, in this case)
and also a country. The program generates a random assignment of teams
to fields, such that the bin 1 teams are spread across fields and the
bin 2 teams are also spread across fields. Also, if the result has two
teams from the same country in the same field, it will start over and
generate a new assignment.

The input file we will be using is attached to this e-mail. How that
file was generated is as follows. First, the top four teams from last
year (Skuba, ZJUNlict, KIKS, and Parsian) are listed from highest to
lowest rank in bin 1, because the generator does not reorder these top
teams—this ensures that Skuba will be on field A, ZJUNlict on field B,
KIKS on field C, and Parsian on field D; we have extended this
privilege to Skuba because, if there is any difference between the
fields, then field A is likely the more desirable field (e.g. due to
bleachers for spectators). Second, bin 2 has MRL and RoboDragons listed
in alphabetical order, followed by two dummy entries to bring the bin
up to four teams; normally this bin would finish the top eight teams,
but that is not possible because Immortals will not be attending the
competition and RoboFEI did not submit an ETDP. Finally, all the
remaining teams are in bin 3, again in alphabetical order, with one
dummy team to bring the count up to a multiple of four. You can see all
the dummy teams in the input file are listed as the same country; this
ensures that the fields will only have one team more or less than each
other, because the program will not put two dummies on the same field.

To select the two numbers the generator uses as seeds, we will be using
a service called random.org. As described above (problem #4), most ways
of choosing two numbers from things like soccer games or lotteries do
not provide enough bits of entropy to perform a fully random assignment
of teams. Random.org is a service that provides random data for many
purposes, both free and commercial, based on radios tuned to pick up
atmospheric noise. One of the services available on random.org is the
ability to generate random integers based on a date; the site generates
a list of numbers on that date, and from that time forward, anyone can
go to the site and get the same numbers by entering the same parameters
and date. To find this service, go to random.org, under Numbers choose
Integers, click Switch to Advanced Mode, and choose the option to use a
pregenerated randomization from a particular date. We will set the
number of requested numbers to 2, the minimum value to 1, the maximum
value to 1000000000 (1 billion), and the date to 2013-06-23. As an
example, if we had used 2013-06-22 (today) instead, the two numbers
would have been 415456377 and 424795746.

We will publish the field grouping as soon as this is done; however, if
you wish, as soon as midnight UTC 2013-06-23 has passed, you should be
able to generate the grouping yourselves by using random.org and
downloading and building the generator.

Thanks, and sorry for the confusion!
Chris
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.20 (GNU/Linux)

iF4EAREIAAYFAlHFXLYACgkQnfE3lq0v9IzaKwD+PHC0+uv/dkNou20OpRFLm6Rm
dG8kqjLJqXPro/ySKs8BAICc3FP8Kj3CkdRG61j5HBOwGpA5VAvRXU9tZ1RYH2c6
=qLkz
-----END PGP SIGNATURE-----
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: teams.txt
URL: <https://mailman.cc.gatech.edu/pipermail/robocup-small/attachments/20130622/f468b041/attachment-0001.txt>


More information about the robocup-small mailing list