WMCCC / Jakarta information
Robert Hyatt
I got several queries about my "pairing" comment, particularly asking
about "what is an accelerated pairing."
I studied the first round, and this is indeed what was done, although
I can't imagine why it was done, and, in fact, it was actually an
incorrect decision to do so I believe.
First, what is accelerated pairings? The basic principle of a
tournament is to "choose #1." If you have fewer than Log2(players)
rounds, you are at risk of not doing so. For example, with 32
players and 4 rounds, assuming the initial seeding order is perfect,
you can have two players finish with perfect 4-0 scores without
playing each other. The "accelerated pairing" scheme lets you
get a #1 while only playing Log2(players)-1 rounds. In this
example, with 32 players, you would still get a #1 (no tie with
any luck) with only 4 rounds, and here's how it works.
Normal pairing for round one would be to seed the players based on
rating, then (in the case of the WMCCC with 28 players) you take the
top 1/2 of the group, slide it down and match players, so that #1 plays
#15, #2 plays #16, ..., #14 plays #28. If your seeding order is
"perfect", and the players perform up to their ratings, #1-#14 will
all win. You take players with like scores, still in seed order,
and do this again, so that #1 plays #8, ... #7 plays #14. You
continue this for 4 total rounds, but at the end, #1 likely won't
meet #2.
In accelerated pairings, you simply assume that the first round
was played, and that the top half of the seeds won. So round one
is paired as though #1-#14 have scores of 1, and #15-#28 have
scores of 0... which means that if the top seeds all win, in
round 4 #1 will meet #2. This is because in round 1, #1 plays #8
rather than #15, and if #1 and #2 keep winning, they will run into
each other in the final round. That's the point. If you have enough
rounds to avoid the tie in the first place (rounds >= Log2(players)
then this is not needed, and, in fact, actually disturbs the final
result. In round 1, the top 1/4 of the field plays the second 1/4
and the second 1/4 should lose. The bottom half does the same.
Now you get intermixing of the two groups of 1's, and the top 1/4
should again win, as should the second 1/4 of the seeds, and so
forth. Doesn't accomplish anything, other than to give the 3rd tier
(of 4 tiers) of players a quick trip to the top group, and the
2nd tier gets a quick dump to the bottom.
No idea why this scheme was used, because with 11 rounds, there's
not much worry of not having enough rounds. In fact, there's really
way too many games. In this case, pairing like this tends to lead
to lots of problems later in the tournament with colors being
wrong, A having already played B once, etc... This happens with or
without accelerated pairings, but more frequently with the latter.
That's my thumbnail sketch of why the first/second round games were
paired as they were, although why this method of pairing was chosen
is anybody's guess. Of course, lots of decisions made surrounding
this event can be fuel for debate...
Kenneth Sloan
In article <53j50n$j...@juniper.cis.uab.edu>,
Robert Hyatt <hy...@crafty.cis.uab.edu> wrote:
>I got several queries about my "pairing" comment, particularly asking
>about "what is an accelerated pairing."
I agree with the sentiment that accelerated pairings should NOT be used
in an event with 28 players and 11 rounds. As Bob says - in this case
you have more than enough rounds to select the #1 player.
I suspect that the organizer/TD has in mind the idea of having as many
"top" programs as possible participate in a round robin. We often hear
complaints about how the Swiss system involves too much "luck". Using
accelerated pairings (perhaps even continuing the "acceleration" through
round 3) will mean that it is highly likely that the player finishing #1
will have played ALL of the top competition. So, we get a Swiss for the
field, and a RR for the very top players. Everyone near the top will
have a very rough schedule, and everyone in contention will get their
chance to play the eventual winner. During the "accelerated" part of
the event, the lower ranked players are essentially playing a
"qualifying" event - only the "best of the worst" will play against the
top players (so no top player will get a "lucky" pairing).
Yes, there will be problems with colors - but with 11 rounds it should
be possible to get very close to even colors (eventually), and a color
split of 7-5 is not all that much of a disadvantage.
The bottom line is - don't think of this as a typical "Swiss". It's
really closer to a RR for the top players.
--
Kenneth Sloan sl...@cis.uab.edu
Computer and Information Sciences (205) 934-2213
University of Alabama at Birmingham FAX (205) 934-5473
Birmingham, AL 35294-1170 http://www.cis.uab.edu/info/faculty/sloan/
Komputer Korner
Kenneth Sloan wrote:
If you meant 7-4, I disagree with you. If you meant 6-5, that is the
best you can get with an odd no. of rounds. The pairing rules should
give more priority at the end so that having an even or close to even
no. of colours is the highest priority.
--
Komputer Korner
Tim Mirabile
sl...@crestline.cis.uab.edu (Kenneth Sloan) wrote:
>Yes, there will be problems with colors - but with 11 rounds it should
>be possible to get very close to even colors (eventually), and a color
>split of 7-5 is not all that much of a disadvantage.
Yes, I'd love to get 7 blacks and 5 whites in an 11 round event.
:^)
+----------------------------------------------------------------------+
| Tim Mirabile <t...@mail.htp.com> http://www.angelfire.com/pg9/timm/ |
| TimM on FICS - telnet://fics.onenet.net:5000/ PGP Key ID: B7CE30D1 |
+----------------------------------------------------------------------+