Report: 3rd Korea Prime Minister Cup - Tournament Conditions
Robert Jasiek
Information started a bit late. Flight tickets (which - besides
accommodation and meals - were free, thanks to the sponsors!) were
issued much too late. This is a major problem not only because of visa
troubles for some countries: Some players had to delay purchase of
connection flight or train tickets until they got doubled prices while
other players were lucky to be paid also their connection flights but
had to start travelling without knowing whether they would indeed have
their major flight e-ticket when checking in.
DATE
The event date was 2008-11-07 to -12, the tournament was -09 to -11.
Finding a good date amidst many other important tournaments may be
difficult. So it was good that the KPMC did not coincide with the
WMSG's date. However, the interested KPMC players had no chance to
listen to most of or even speak at the 5th ICOB; this is very
unfortunate. The AGF's AGM and teaching meeting were also too close to
the KPMC.
VENUE
The congress center may have been a suitable place. The first
tournament day's playing hall (ball room) was packed with three
tournaments, noisy players, and a demonstration board though. Although
such is inappropriate, it is not quite as bad as it may sound because
all the noise was go-related and therefore not too distracting. The
room for rounds 3 to 8 was of a reasonable size to host the 68 players
and the organizers. Both rooms did not have daylight.
FURNITURE
The chairs were ok but maybe slightly too low. The tables were too low
to stretch one's legs easily but not quite as low as the WMSG tables.
Each board had its own table with enough free space around it. Scribes
had extra tables.
PLAYING MATERIAL
The boards were thick. On the too low tables, the board surface was at
a suitable height though. Quite some boards had some small but darkish
dirt on the surface. Why use what looks like quality boards if their
appearance cannot be maintained clear? The stones seemed to be of a
rather low plastics quality. Is it really that difficult to organize
better material (not Ing material, of course) for a very important
international tournament?
CLOCKS
The grey Korean clocks were used. That Ing clocks are even worse is no
excuse. Why design playing clocks where pressing the button is tough
work and requires luck? Why annoy all players by beep confirmation of
every move and by spoken time? It is horrific. The worst aspect is
this: While the not so critical basic time is shown in a big font, the
essential byoyomi data use a tiny font. This is all the worse the
lower the thinking time and byoyomi are.
THINKING TIME
The basic time was 30 minutes, the byoymi 3*30s. For an almost world
championship like tournament, this is ridiculously small and would be
appropriate for a rapid side tournament. Even with the 1h basic time
of the WMSG, many players made huge blunders. In the KPMC, it was not
different. With up to 3 games per day, 100 min. basic time plus soft
Canadian Byoyomi would be possible easily. Without mandatory
sightseeing and 4 tournament days (2 games per day), 2h basic time
become possible.
RULES OF PLAY
While the referees were announced before every round, the rules of
play were not announced in time. Only persistent asking let the author
know before round 1 that Korean Rules were used. For everybody, an
announcement was made only before round 3: "The rules we use are close
to international rules." A couple of rounds later, it was also
announced that alternate dame filling is mandatory.
Korean rules close to international rules? What are "international
rules"...? Presumably "Verbal Japanese Rules" were being meant and
described by the deceiving euphemism (one might also say: outrageous
lie). Close? How close is it, to mention just the most apparent, that
Korean Rules compared to the Applicable Traditional Japanese Rules
have these exceptional extras? Unclear removals of dead stones from
sekis, unclear perfect pass rule, unclear play out during the
alternation rules, ambiguous clearly more ko threats rule, unclear
long cycles playout rule, and unclear triple ko with one eye
exception. - A statement closer to the truth would have been: "Korean
Rules are used. They without all their exceptional extras are close to
Verbal Japanese Rules. However, alternate filling of two-sided dame
and teire is mandatory."
RESULT LIST
The walllist and result list lists the players' initial numbers as the
opponent numbers. This makes it very hard to read. Every Swiss and
McMahon list should use the current / final player numbers for the
opponent numbers! Good pairing programs with that feature are
available.
WINNING CRITERIA
The winning criteria for the final results ordering were Wins - SOS -
SOSOS. Wins is obvious; SOSOS is lottery. - The interesting aspect is
the first tiebreaker SOS. Before round 1, it was announced: "You
cannot influence your SOS - only your opponents can!" Although this
comes pretty close to one aspect of the nature of SOS, it hides
another very dominating aspect: the effect of the pairing method. That
was so extremely great that even the Wins were affected by up to 2 or
3 points per player! I.e., Wins as the first winning criterion was not
obvious at all! For the winning criteria, already a discussion of SOS
is somewhat esoteric - discussing Wins only is sufficiently
embarrassing. Just a quick look at the ordering of ranks should be
convincing enough. - SOS plays a greater role for place 1
determination though.
PAIRING SYSTEM
The initial player numbering depended on a country's previous year's
KPMC final places; new countries were somehow inserted. Round 1 used
cross pairing. This is a reasonable choice, and elsewhere other
organizations have concluded this, too, for early rounds Swiss
pairing.
The most unfortunately, adjacency pairing (#1-#2...#y-#z) was used per
score group in rounds 2 to 8. The effect is that some early winners
are paired against strong opponents for the rest of the tournament and
make not so many wins while some early losers are paired against weak
opponents for the rest of the tournament and make rather many wins.
Hence quite some weaker players will end up with a higher final place
than quite some stronger players. Such can be seen in the result table
easily. - Many players have criticized the adjacency pairing very
strongly. The author has not heard any player who might have praised
this pairing method. - It is irresponsible to use such a highly
doubtful pairing method. Although go has not found its best pairing
method or combination of pairing methods in different rounds yet, we
may be grateful that the 3rd KPMC serves as a good example for how bad
adjacency pairing actually is. Instead fold pairing, cross pairing, or
random pairing might be considered for most (last) rounds of a Swiss /
McMahon.
7 WINS PLAYERS
Opponents
# Name XX SOS SOSOS
1 ChienLiChen TW 47 312 19+ 6+ 15+ 3+ 4+ 2- 11+ 7+
2 LeeSangHun KR 43 311 18+ 34+ 17+ 11+ 28+ 1+ 3- 4+
3 ZhaoWei CN 42 323 26+ 32+ 30+ 1- 5+ 12+ 2+ 11+
Opponents-SOS
# Name XX SOS SOSOS
1 ChienLiChen TW 47 312 5+ 6+ 5+ 7+ 6+ 7- 5+ 6+
2 LeeSangHun KR 43 311 5+ 4+ 5+ 5+ 4+ 7+ 7- 6+
3 ZhaoWei CN 42 323 4+ 4+ 4+ 7- 6+ 5+ 7+ 5+
Opponents-SOS Summary
# Name XX SOS SOSOS 7 6 5 4
1 ChienLiChen TW 47 312 2 3 3 0
2 LeeSangHun KR 43 311 2 1 3 2
3 ZhaoWei CN 42 323 2 1 2 3
Player #1 won the tournament because he did not get any SOS=4
opponent. (Instead he got 2 SOS=6 opponents more than #2 and #3.)
Why did #1 not get any SOS=4 opponent? This is explained by the early
rounds SOS and the adjacency pairing that rewarded early high SOS:
Early Rounds SOS Before the Round Starts
# Name XX SOS SOSOS Round
1 2 3 4
1 ChienLiChen TW 47 312 0 0 2 6
2 LeeSangHun KR 43 311 0 0 2 5
3 ZhaoWei CN 42 323 0 0 2 5
The players have similar values here. The interesting thing is though
that during the first 3 rounds SOS has only little impact on the
pairing because the top players have beaten them. Therefore the
relative final SOS input from the early rounds opponents is almost
independent from the top players' early achievement and from the
pairing method, which has its relevant impact almost only during the
later rounds.
Again: Why did #1 not get any SOS=4 opponent? Since it cannot be
explained by the top players' own performance, it must be explained by
the opponents' own performce. So let us study the SOS=4 opponents (* =
moment of top player pair):
Opponent #34
Opponents:
34 RasmussenJ DK 32 263 68+ 2-* 40+ 8- 16- 46+ 35- 49+
Wins after R Rounds:
34 RasmussenJ DK 32 263 1 1 2 2 2 3 3 4
Opponents-SOS:
34 RasmussenJ DK 32 263 0 7 4 6 5 3 4 3
#34 played in most rounds as expected except round 7 that he might as
well have won. For #2, #34 was simply a bad pairing luck.
Opponent #28
Opponents:
28 FanJun CA 37 284 62+ 47+ 45+ 13+ 2-* 4- 9- 18-
Wins after R Rounds:
28 FanJun CA 37 284 1 2 3 4 4 4 4 4
Winning the first 4 rounds, then meeting #2, and then also losing the
rest is not plainly surprising, except for the last round. Rather here
adjacency pairing hits hard.
Opponent #26
Opponents:
26 SoldanLeszek PL 40 273 3-* 25+ 23+ 30- 33+ 27+ 8- 13-
Wins after R Rounds:
26 SoldanLeszek PL 40 273 0 1 2 2 3 4 4 4
Opponents-SOS:
26 SoldanLeszek PL 40 273 7 5 5 4 4 4 6 5
For a 4 wins player, #26 has incredibly strong SOS. #3 would have had
better pairing luck in round 1 getting a finally weak SOS opponent
with more wins.
Opponent #32
Opponents:
32 MoralesPablo ES 34 253 58+ 3-* 35- 37+ 21- 50+ 48+ 19-
Wins after R Rounds:
32 MoralesPablo ES 34 253 1 1 1 2 2 3 4 4
Opponents-SOS:
32 MoralesPablo ES 34 253 3 7 4 4 5 3 3 5
Everything looks unspectacular for #32. #3 would have had better
pairing luck in round 2 getting some other unspectacular player with
finally more wins.
Opponent #30
Opponents:
30 JasiekRobert DE 35 296 60+ 33+ 3-* 26+ 12- 15- 17- 44+
Wins after R Rounds:
30 JasiekRobert DE 35 296 1 2 2 3 3 3 3 4
Opponents-SOS:
30 JasiekRobert DE 35 296 2 4 7 4 5 5 5 3
Opponents' Place in Their Final Score Group
30 JasiekRobert DE 35 296 #2 #8 #3 #1 #2 #5 #7 #1
#30 had a good start on wins in the first two rounds, then got a SOS=7
opponent, then tended to be (about) #1 in his score group during later
rounds. This can also be seen by the opponents' places in rounds 4, 5,
and 8. Adjacency pairing hit #30 hard. After round 8, his score group
place suddenly dropped to #5. So his final SOS hides what happened
during the pairings in rounds 4 to 8.
Conclusion for the SOS=4 Opponents
Some of the SOS=4 opponents can be explained by adjacency pairing,
others are unspectacular and got their final number of wins for no
particularly specific reason.
This explains why #3 has got this place. It does not explain yet why
#1 made a better place than #2 by avoiding 4 points opponents.
The apparent difference between #1 and #2 is round 2. (In the rounds
4+, SOS=4 opponents are pretty unlikely for top players anyway. So
only the anomalies of rounds 1 to 3 are interesting here.) So what
about
Opponent #6
Why could he get 6 wins?
6 HsiangThomasY US 38 308 46+ 1-* 36+ 17+ 11- 29+ 13+ 14+
Wins after R Rounds:
6 HsiangThomasY US 38 308 1 1 2 3 3 4 5 6
Opponents-SOS:
6 HsiangThomasY US 38 308 3 7 4 5 5 4 5 5
Opponents-SOS Summary: 7 6 5 4 3
1 0 4 2 1
Because he is stronger than 5 wins players and was lucky not to get
any 6 wins player! Why did he have that luck? Maybe because he
postponed 3 of his wins until the last 3 rounds. Having played #1, #3,
#4 of the final Wins=5 group confirms that he is stronger than all 5
wins players.
Final Conclusion
So #1 could hardly have had any better pairing luck for round 2: A
player weaker than himself but among those one of the strongest and
one with not too strong opponent's opponents.
So although the SOS difference between #1 and #2 is significantly
greater than the SOS difference between #2 and #3, the former is
explained mostly by early rounds pairing luck while the latter is
explained to quite some extent by adjacency pairing. SOS is too
unpredictable overall to explain the exact sizes of either difference.
For the 3 top players, the impact of adjacency pairing is smaller than
for the rest of the player field but can still be noticed. Therefore
one should not even argue that adjacency pairing might have been
chosen for nothing but the purpose of ordering the top, wins-tied
players among each other. Adjacency pairing has only one - very
doubtful - advantage: It launches some weaker players to higher final
places than they deserve and thereby gives them some fake motivation
for improving in future.
CONCLUSION
The tournament conditions of the KPMC can be improved greatly:
- better preparation
- better dates of other, minor events
- slightly better venue, furniture, playing material
- better clocks
- longer thinking time
- better announcement of and better rules of play
- better concept of result lists
- more meaningful winning criteria and a reconsideration of using
tiebreakers for the final results ordering (not to be confused with a)
pairings, or b) initial placement) at all
- much better pairing method in round 2 to 8
While the sightseeing, the (wo)man power, and the presence of high
ranking politicians are impressive (sorry: the culture was just too
loud though), the same cannot be said about the tournament conditions.
It is strange that the core of the event is the highly neglected part
so far.
regg...@gmail.com
Great report! Only one small correction:
"...new countries were somehow inserted..." New countries were put at
the end of the previous year's
KPMC final places list.
Robert Jasiek
>Great report!
Thank you!
Basically there are two kinds of opinions on this report and my WMSG
report. Either like your opinion or some have a contrary opinion: "The
report(s) are heavily one-sided, unrespectful, and endanger the future
of Asian sponsorship of amateur world tournaments." I can't understand
such a view (which implies or is accompanied with recommendations of
self-censorship of one's freedom of speech) but one thing also I agree
to: More reports by other writers from also different perspectives
might give a broader impression than my emphasis on technicalities and
tournament organization.
Meanwhile politicians actually read such reports. While we are free to
praise or criticize tournaments, they are free to do so with our
reports. One thing really bothers me though: Rumours that criticism of
our reports substitutes reflection of whether the quality of
tournaments should be improved indeed. An influential politician even
suggested I should not attend next year's WAGC so that I will not
write another critical report. At the same, such critics of our
reports do not even want to discuss their contents. IMO, better
quality of the tournaments would also make the sponsors happier - much
more so than hiding existing shortcomings by questioning the freedom
of speech.
After the KPMC, queries have been given to the players for their
comments and suggestions. During the 5th ICOB, expression of free and
alternative opinion was encouraged by the moderators. So what are some
Western politicians afraid of? Have they not realized yet that sincere
factual discussion and freedom of speech is customary also in East
Asia?
***
For one statement I should apologize though. I wrote: "Presumably
'Verbal Japanese Rules' were being meant and described by the
deceiving euphemism (one might also say: outrageous lie)." Here
"outrageous lie" is an overreaction by me. I can't know if "The rules
we use are close to international rules." were said intentionally or
carelessly and might have wanted to say something like: "Application
of the rules we use is close to application of Verbal Japanese Rules."
Robert Jasiek
diagram, which to some extent allows a visual analysis of a
tournament's pairing quality. Here is an example:
http://www.dgob.de/yabbse/index.php?action=dlattach;topic=3301.0;attach=2103;image
The row r shows the player with the final place r and his results
against opponents. Column c is his opponent with the final place c.
Green means "has won". Red means "has lost".
The example is for a 7 rounds tournament (WMSG Individual Open
preliminary group 4). The dark blue region extends 7 (= number of
rounds) fields from the black diagonal. Light blue is an (arbitrary)
extension of that core region. So a player winning all his games would
fill the core region right of the diagonal; a player losing all would
fill it left of the diagonal. In an ideally paired tournament, all
results would fit onto the dark blue strip.
The pairing quality can be measured by, e.g., the sum of distances of
all result fields outside the dark blue strip to it.
The pairing quality depends on several factors. Among them are density
of the player field, pairing algorithm, and number of rounds.
http://www.dgob.de/yabbse/index.php?action=dlattach;topic=3301.0;attach=2107;image
shows a typical German R=5 rounds McMahon weekend tournament, paired
presumably by the program "MacMahon": St. Augustin, Germany. Final
table:
http://www.dgob.de/tourn/tourn.cgi?f=08destau.txt&mode=cml
The quality of the combination of tournament system and pairing
algorithm (which relies on globally optimized, SOS-SOSOS-dependent
fold pairing) should be obvious: Although the player field is not
particularly dense, almost all results fall into a 2R strip.
http://www.dgob.de/yabbse/index.php?action=dlattach;topic=3301.0;attach=2112;image
shows the diagrams for each round of the Korea Prime Minister Cup
2008. Light green / red are the current round's new game results; dark
green / red are previous rounds' results. It is very apparent how
relatively few results fall into blue even after 8 rounds. Only from
round 7 on, the new results start being focussed in (at least) the 2R
strip. Despite the great range of player ranks in the Swiss
tournament, this is a very convincing confirmation about the very low
quality of the used adjacency pairing method: It is very weak at
creating a high percentage of pairs with finally close places in the
result list.
henric
> On Tue, 18 Nov 2008 05:54:01 -0800 (PST), reggi...@gmail.com wrote:
> >Great report!
>
> Thank you!
>
> Basically there are two kinds of opinions on this report and my WMSG
> report. Either like your opinion or some have a contrary opinion: "The
> report(s) are heavily one-sided, unrespectful, and endanger the future
> of Asian sponsorship of amateur world tournaments."
The sponsored international amateur events are extremely valuable in
stimulating go world wide, provided that they are handled in a
responsible way by the participating countries (qualification through
competitions or in other ways that stimulate go activities). In Sweden
for example, the new big events (KPMC and WMSG) have meant in
particular that the younger players who are developing quickly could
qualify faster for representing their country in such events.
Also, though I didn't participate myself I'm sure that the first hand
impression by the participants in WMSG and KPMC must have been that
it's a great experience and very positive.
That's a lot more important than e.g. some little problem with a clock
or suchlike.
In that perspective your reports, which focus a lot on negative
aspects, are very one sided and might have negative effects overall.
On the other hand it's of course good to mention problems and things
that might be improved. Everybody for sure won't agree with your
priorities in tournaments, so your reviews are personal, not
necessarily representative for the majority of participants. But they
have at least been thorough and detailed and of course interesting as
such. The best solution for the future is hardly that you change your
Jasiek review style or stop writing, rather there seems to be a need
for other opinions than yours on the table, for balance.
I think your points about the pairing are interesting, although it is
probably not possible to design a perfect algorithm anyway, so it's
not trivial to judge on which aspects of the pairing are most
important. It's also good that you mention the thinking times, which
are unusually short, it would be nice to see other opinions too on
that. Perhaps not everybody was aware that the games in the KPMC are
that fast. Maybe there are advantages too with a big tournament with
shorter thinking time.
best regards,
Henric
Robert Jasiek
<henricb...@gmail.com> wrote:
>The sponsored international amateur events are extremely valuable in
>stimulating go world wide,
Absolutely, although I would add that also pro events (for later game
study) or non-sponsored events can also be extremely valuable for that
purpose. However, a not sponsored amateur world tournament would see
much fewer represented countries and a smaller variety of players. A
lot of players of also other sports have said this during WMSG or
KPMC; air fares are simply too high for many.
>I think your points about the pairing are interesting, although it is
>probably not possible to design a perfect algorithm anyway,
Not yet. Pairing theory deserves a lot of further study, incl. a
precise formulation and weighting of aims. Quite something is already
understood though, and why should one ignore it?
The pairing algorithm used at KPMC has given the weaker countries'
players about the maximal number of "teaching" games within the
tournament. However, I think that it is not necessary to have this as
an aim for the tournament. If one wants it formal, one can let the
high scoring players teach the low scoring players in formal friendly
games during the afternoon. But is this even necessary? A lot (eh,
quite some at least) of strong players play many friendly even or
handicap games against everybody who likes during the afternoon or the
night.
Therefore the tournament itself can concentrate on what it is
announced to be: A competition for the first place. Then the weaker
players get maybe only 1 or 2 instead of 3 or 4 teaching games inside
the tournament. I think this is better because every tournament's
first objective should be the best possible determination of the top
places.
>the thinking times, which
>are unusually short, it would be nice to see other opinions too on
>that.
1 or 2 low dan Latin Americans told me that they don't know how to use
thinking time and could play 1 minute basic time games as well as 30
minutes games. Otherwise the opinions tended towards longer thinking
times but opinions differed about the exact length.