From: Jochum Drechsler <drjochum <at> web <dot> de>
Date: 20 sep 2004
Subject: Reply to Erics remarks about: Shogi SOS-based pairing at the EC/WOSC

This is a multi-part message in MIME format.
--------------010609080403010102070503
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: quoted-printable

Your free subscription is supported by today's sponsor:
-------------------------------------------------------------------
Do you live vicariously through Reality TV? Let us know and win!
http://click.topica.com/=
caacDv1a2i6YsbnuqMaa/PermissionData
-------------------------------------------------------------------

Dear Eric,

I am sorry but in most parts I can not follow your argumentation. Let me=
=20
explain.

>> The system is not strange. Most tournaments in Germany used this=20
>> pairing system in the past. All major Go-tournaments in Europe also=20
>> used this method (and the programme).
>> Also: This pairing system has nothing to do with MacMahon. The=20
>> programme supports the MacMahon system and this methode can also be=20
>> used in the MacMahon system, but it is not restricted to it!
>>
> Later in this mail, you give the name of this program "Gerlachs=20
> MacMahon program" that indicates that it was first designed for=20
> MacMahon system tournaments. And major Go European tournaments are=20
> mostly organized with MacMahon points for each grade. I can understand=
=20
> that this system can work correctly with MacMahon points (+3 for 4=20
> Dan, +2 for 3 Dan, ... -1 for 1 kyu and so on) with a good number of=20
> players for each grade.

The programme was indeed designed for Go tournaments using a MacMahon=20
system. Nevertheless the SOS-based pairing method is completely=20
independent of the MacMahon system!!
And by the way a tournament where all players start with same number of=20
points in round one (Swiss system) is just a special case of the=20
MacMahon system. All players start with the same MacMahon score!

> But the WOSC has not the same characterics at all With an inside=20
> knock-out tournament, no MacMahon point and a relatively small number=20
> of players at each grade, this system shows its limits.
>
Before I answer to this I would like to point out that there are (at=20
least) two aims of good tournamts:
A) Find the best player and produce a good (meaningful) ranking at the top.=

B) Make the tournament exciting for all players. That means mainly to=20
produce exciting and lose games.
You can not necessarily fulfil both points at the same time. I think=20
that A has clear priority over B. So for the top places you follow A=20
without restriction.
I agree with you that the WOSC is (was) indeed different to the MacMahon=
=20
system as all players started with the same number of points from the start=
.

>> 1. No two players will be paired twice.
>> 2. The global number of players paired up or down has to be minimized.
>> 3. The choice of players that are paired up or down.
>>
> Nothing about a rule preventing a player from being "punished" several=
=20
> times in the tournament as in the swiss system.

It is not clear to me what you mean by punishment. If you mean in regard=
=20
to interesting games or not, see below.
If you mean as a result of unavoidable disatvantageous pairings, you are=
=20
right that the rules do not mention that directely. However, the rules=20
are more like axioms that have to be followed by the pairing methode.=20
The pairing method itself guarantees that no player is treated worse=20
than any other player.

>> Pairing within a point (or MacMahon) group
>> The players within one group (with the same number of points or=20
>> macmahon points) are paired like this: the first player (according to=
=20
>> his SOS) against the last; the second against the second last. The=20
>> players that had a strong tournament so far (shown in their high sos)=
=20
>> will get a relatively weaker opponent.
>>
> Here is the main bad side effect : when you take part in the knock-out=
=20
> tournament and when you lose the semi final and the final for the=20
> third place of the ESC, you have "only" 3 points out of five. You are=20
> somewhere in the middle of the tournament but you have a high SOS=20
> number of points (probably the highest SOS points of the 3 out of 5=20
> players).
>
> With the above rule, not only you just lost the semi final and the 3rd=
=20
> place final of the ESC but you will be punished again with a "weak"=20
> pairing. This happened for Gert twice in rounds 6 and 7 :-(( And this=20
> ruined his final place in the tournament table after his good start in=
=20
> the tournament.

I think your argument does not work.
Are you saying that it this is fair?: If you have a player with a high=20
SOS (the highest of all players with the same number of points) should=20
get a strong opponent again, that he is rather likely to loose against?=20
And that a weaker player should get again a weak player, he is rather=20
likely to win? That have the result that that weaker player has one more=
=20
point than that stronger player, but the weaker player had much, much=20
less SOS than the stronger player. That would be strange!
A player that already had very strong opponents (but few match points)=20
should get a weaker player so that he gets the chance catch up in points=
=20
on the other players that have more points but had weaker opponents so far!=
!
This is what it is all about if you level the SOS.
You do not say anything about the unfairnesses of the big diversions of=20
SOS in the final ranking that would otherwise result.

> Sorry, but I don't agree with this point. As by "relatively weak=20
> opponent", your program chooses a player somewhere in the weakest=20
> players of the group. The main consequence of this is that after two=20
> hard fought games in round 4 and 5, Gert played against two much=20
> weaker players in rounds 6 and 7. The difference of strength between=20
> his opponents is really too wide, this system prevents the "punished"=20
> player from playing interesting games. How can you still be interested=
=20
> in the tournament with such a pairing ? You simply lose your=20
> concentration for the next games. I am afraid that seeing this, the=20
> next loser of the 3rd place final, will expect bad pairings in the=20
> next rounds. His interest in the tournament can simply vanish ;-(

You are talking about another aspect here: Excting games.
Before we have looked at the aim A now we are talking about aim B. This=20
is indeed an important point. Look at the first two games of Mr. Sera.=20
These pairings had a difference of more than 2000 Elo points. A problem=20
of the swiss pairing as well as this SOS-based pairng is the elevater=20
effect: that you get a very strong opponent and lose; than you get a=20
very weak opponent and win. Up and down.
This problem can only be avoided if you take other means. One=20
possibility would be to use MacMahon groups (say four groups for=20
example). You could also have completely seperate groups (a dan and a=20
kyu group for example).
I completely agree with you that the games do not get so exciting if you=
=20
have all players in one group from the start. However this nothing to do=
=20
with the pairing method used to find the top ranked players.
FESA dediced this year to do it differently next year but not exactly how.

For all players that have a chance to win the tournament (a predefinded=20
top group) all other considerations (exciting games or not; playing=20
against players from the same club) are of course not relevant.

So far, Jochum.

part 2 will follow.

Your free subscription is supported by today's sponsor:
-------------------------------------------------------------------
Give your opinion on Abortion and Win!
http://click.topica.com/=
caacERNa2i6YsbnuqMaf/PermissionData
-------------------------------------------------------------------

--^----------------------------------------------------------------
This email was sent to: =
shogi-l <at> shogi <dot> net

EASY UNSUBSCRIBE click here: http://topica.com/u/?a2i6Ys.=
bnuqMa.=
c2hvZ2kt
Or send an email to: shogi-unsubscribe <at> topica <dot> com

For Topica's complete suite of email marketing solutions visit:
http://www.topica.com/?p=3DTEXFOOTER
--^----------------------------------------------------------------




--------------010609080403010102070503
Content-Type: text/html; charset=us-ascii
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
  <meta http-equiv=3D"Content-Type" content=3D"text/html;charset=3DISO-8859=
-1">
  <title></title>
</head>
<body text=3D"#000000" bgcolor=3D"#ffffff">=
<PRE>Your free subscription is supported by today's sponsor:
-------------------------------------------------------------------
Do you live vicariously through Reality TV? Let us know and win!
<A HREF=3D"http://click.topica.com/=
caacDv1a2i6YsbnuqMab/PermissionData">http://click.topica.com/=
caacDv1a2i6YsbnuqMab/PermissionData</A>
-------------------------------------------------------------------
</PRE>

<p class=3D"MsoNormal"><span lang=3D"EN-GB" style=3D"">Dear Eric,<o:p></o:p=
></span></p>
<p class=3D"MsoNormal"><span lang=3D"EN-GB" style=3D"">I am sorry
but in most parts I can not follow your argumentation. Let me explain.<br
 style=3D"">
<!--[if !supportLineBreakNewLine]--><br style=3D"">
<!--[endif]--><o:p></o:p></span></p>
<blockquote type=3D"cite">
  <blockquote type=3D"cite">
    <p class=3D"MsoNormal"><span lang=3D"EN-GB" style=3D"">The system
is not strange. Most tournaments in Germany used this pairing system in
the
past. All major Go-tournaments in Europe also used this method (and the
programme).
    <br>
Also: This pairing system has nothing to do with MacMahon. The
programme
supports the MacMahon system and this methode can also be used in the
MacMahon
system, but it is not restricted to it! </span></p>
  </blockquote>
  <span lang=3D"EN-GB"
 style=3D"font-size: 12pt; font-family: &quot;Times New Roman&quot;;">Later=
 in this
mail, you
give the name of this program "Gerlachs MacMahon program" that
indicates that it was first designed for MacMahon system tournaments.
And major
Go European tournaments are mostly organized with MacMahon points for
each
grade. I can understand that this system can work correctly with
MacMahon
points (+3 for 4 Dan, +2 for 3 Dan, ... -1 for 1 kyu and so on) with a
good
number of players for each grade. <br style=3D"">
<!--[if !supportLineBreakNewLine]--></span></blockquote>
<p class=3D"MsoNormal"><span lang=3D"EN-GB" style=3D"">The
programme was indeed designed for Go tournaments using a MacMahon
system.
Nevertheless the SOS-based pairing method is completely independent of
the
MacMahon system!! <br>
And by the way a tournament where all players start with same
number of points in round one (Swiss system) is just a special case of
the MacMahon system.
All players start with the same MacMahon score!<o:p></o:p></span></p>
<blockquote type=3D"cite"><span lang=3D"EN-GB"
 style=3D"font-size: 12pt; font-family: &quot;Times New Roman&quot;;">But t=
he WOSC
has not the
same characterics at all </span><span
 style=3D"font-size: 12pt; font-family: &quot;Times New Roman&quot;;"><!--[=
if gte vml 1]><v:shapetype id=3D"_x0000_t75"
 coordsize=3D"21600,21600" o:spt=3D"75" o:preferrelative=3D"t" path=3D"m@4@=
5l@4@11@9@11@9@5xe"
 filled=3D"f" stroked=3D"f">
 <v:stroke joinstyle=3D"miter"/>
 <v:formulas>
  <v:f eqn=3D"if lineDrawn pixelLineWidth 0"/>
  <v:f eqn=3D"sum @0 1 0"/>
  <v:f eqn=3D"sum 0 0 @1"/>
  <v:f eqn=3D"prod @2 1 2"/>
  <v:f eqn=3D"prod @3 21600 pixelWidth"/>
  <v:f eqn=3D"prod @3 21600 pixelHeight"/>
  <v:f eqn=3D"sum @0 0 1"/>
  <v:f eqn=3D"prod @6 1 2"/>
  <v:f eqn=3D"prod @7 21600 pixelWidth"/>
  <v:f eqn=3D"sum @8 21600 0"/>
  <v:f eqn=3D"prod @7 21600 pixelHeight"/>
  <v:f eqn=3D"sum @10 21600 0"/>
 </v:formulas>
 <v:path o:extrusionok=3D"f" gradientshapeok=3D"t" o:connecttype=3D"rect"/>=

 <o:lock v:ext=3D"edit" aspectratio=3D"t"/>
</v:shapetype><v:shape id=3D"_x0000_i1025" type=3D"#_x0000_t75" alt=3D":-("=
 style=3D'width:14.4pt;
 height:14.4pt'/><![endif]--><!--[if !vml]--><img
 width=3D"19" height=3D"19"
 src=3D"file:///C:/WINDOWS/TEMP/msoclip1/02/clip_image001.gif" alt=3D":-("
 v:shapes=3D"_x0000_i1025"><!--[endif]--></span><span lang=3D"EN-GB"
 style=3D"font-size: 12pt; font-family: &quot;Times New Roman&quot;;">With =
an
inside knock-out tournament, no MacMahon point and a relatively
small number of players at each grade, this system shows its limits. <br
 style=3D"">
<!--[if !supportLineBreakNewLine]--><br style=3D"">
<!--[endif]--></span></blockquote>
<span lang=3D"EN-GB" style=3D"">Before I
answer to this I would like to point out that there are (at least) two
aims of
good tournamts:<o:p></o:p></span>
<br>
<span lang=3D"EN-GB" style=3D"">A) Find the
best player and produce a good (meaningful) ranking at the top.<o:p></o:p><=
/span>
<br>
<span lang=3D"EN-GB" style=3D"">B) Make the
tournament exciting for all players. That means mainly to produce
exciting and
lose games. <o:p></o:p></span>
<br>
<span lang=3D"EN-GB" style=3D"">You can not
necessarily fulfil both points at the same time. I think that A has
clear
priority over B. So for the top places you follow A without
restriction. <o:p></o:p></span>
<br>
<span lang=3D"EN-GB" style=3D"">I agree
with you that the WOSC is (was) indeed different to the MacMahon system
as all
players started with the same number of points from the start.<o:p></o:p></=
span><br>
<blockquote type=3D"cite">
  <blockquote type=3D"cite">
    <p class=3D"MsoNormal"><span lang=3D"EN-GB" style=3D"">1. No two
players will be paired twice. <br>
2. The global number of players paired up or down has to be minimized. <br>=

3. The choice of players that are paired up or down. <o:p></o:p></span></p>=

  </blockquote>
  <span lang=3D"EN-GB"
 style=3D"font-size: 12pt; font-family: &quot;Times New Roman&quot;;">Nothi=
ng about
a rule
preventing a player from being "punished" several times in the
tournament as in the swiss system. <br>
<!--[endif]--></span></blockquote>
It is not clear to me what you mean by punishment. If you mean in
regard to interesting games or not, see below.<br>
If you mean as a result of unavoidable disatvantageous pairings, you
are right that the rules do not mention that directely. However, the
rules are more like axioms that have to be followed by the pairing
methode. The pairing method itself guarantees that no player is treated
worse than any other player.<span lang=3D"EN-GB"
 style=3D"font-size: 12pt; font-family: &quot;Times New Roman&quot;;"><br s=
tyle=3D"">
<!--[endif]--></span>
<blockquote type=3D"cite">
  <blockquote type=3D"cite">
    <p class=3D"MsoNormal"><span lang=3D"EN-GB" style=3D"">Pairing
within a point (or MacMahon) group <br>
The players within one group (with the same number of points or
macmahon
points) are paired like this: the first player (according to his SOS)
against
the last; the second against the second last. The players that had a
strong
tournament so far (shown in their high sos) will get a relatively
weaker
opponent. <o:p></o:p></span></p>
  </blockquote>
  <span lang=3D"EN-GB"
 style=3D"font-size: 12pt; font-family: &quot;Times New Roman&quot;;">Here =
is the
main bad
side effect : when you take part in the knock-out tournament and when
you lose
the semi final and the final for the third place of the ESC, you have
"only" 3 points out of five. You are somewhere in the middle of the
tournament but you have a high SOS number of points (probably the
highest SOS
points of the 3 out of 5 players). <br>
  <br>
With the above rule, not only you just lost the semi final and the 3rd
place
final of the ESC but you will be punished again with a "weak"
pairing. This happened for Gert twice in rounds 6 and 7 :-(( And this
ruined
his final place in the tournament table after his good start in the
tournament.
  <br style=3D"">
<!--[if !supportLineBreakNewLine]--></span></blockquote>
I think your argument does not work.<br>
Are you saying that it this is fair?: If you have a player with a high
SOS (the highest of all players with the same number of points) should
get a strong opponent again, that he is rather likely to loose against?
And that a weaker player should get again a weak player, he is rather
likely to win? That have the result that that weaker player has one
more point than that stronger player, but the weaker player had much,
much less SOS than the stronger player. That would be strange!<br>
A player that already had very strong opponents (but few match points)
should get a weaker player so that he gets the chance catch up in
points on the other players that have more points but had weaker
opponents so far!!<br>
This is what it is all about if you level the SOS.<br>
You do not say anything about the unfairnesses of the big diversions of
SOS in the final ranking that would otherwise result.<br>
<br>
<blockquote type=3D"cite"><span lang=3D"EN-GB"
 style=3D"font-size: 12pt; font-family: &quot;Times New Roman&quot;;">Sorry=
, but I
don't agree
with this point. As by "relatively weak opponent", your program
chooses a player somewhere in the weakest players of the group. The
main
consequence of this is that after two hard fought games in round 4 and
5, Gert
played against two much weaker players in rounds 6 and 7. The
difference of
strength between his opponents is really too wide, this system prevents
the
"punished" player from playing interesting games. How can you still
be interested in the tournament with such a pairing ? You simply lose
your
concentration for the next games. I am afraid that seeing this, the
next loser
of the 3rd place final, will expect bad pairings in the next rounds.
His
interest in the tournament can simply vanish ;-( <br>
<!--[endif]--></span></blockquote>
You are talking about another aspect here: Excting games.<br>
Before we have looked at the aim A now we are talking about aim B. This
is indeed an important point. Look at the first two games of Mr. Sera.
These pairings had a difference of more than 2000 Elo points. A problem
of the swiss pairing as well as this SOS-based pairng is the elevater
effect: that you get a very strong opponent and lose; than you get a
very weak opponent and win. Up and down.<br>
This problem can only be avoided if you take other means. One
possibility would be to use MacMahon groups (say four groups for
example). You could also have completely seperate groups (a dan and a
kyu group for example).<br>
I completely agree with you that the games do not get so exciting if
you have all players in one group from the start. However this nothing
to do with the pairing method used to find the top ranked players.<br>
FESA dediced this year to do it differently next year but not exactly
how.<br>
<br>
For all players that have a chance to win the tournament (a predefinded
top group) all other considerations (exciting games or not; playing
against players from the same club) are of course not relevant.<br>
<br>
So far, Jochum. <br>
<br>
part 2 will follow.<br>
<br>
<br>
<blockquote type=3D"cite">
  <p class=3D"MsoNormal"><span lang=3D"EN-GB" style=3D""><o:p></o:p></span>=
</p>
  <span lang=3D"EN-GB"
 style=3D"font-size: 12pt; font-family: &quot;Times New Roman&quot;;"><!--[=
endif]--></span></blockquote>
<PRE>Your free subscription is supported by today's sponsor:
-------------------------------------------------------------------
Give your opinion on Abortion and Win!
<A HREF=3D"http://click.topica.com/=
caacERNa2i6YsbnuqMag/PermissionData">http://click.topica.com/=
caacERNa2i6YsbnuqMag/PermissionData</A>
-------------------------------------------------------------------
</PRE>
<PRE>--^^-------------------------------------------------=
--------------
This email was sent to: =
shogi-l <at> shogi <dot> net

EASY UNSUBSCRIBE click here: <A HREF=3D"http://topica.com/u/?a2i6Ys.=
bnuqMa.=
c2hvZ2kt">http://topica.com/u/?a2i6Ys.=
bnuqMa.=
c2hvZ2kt</A>
Or send an email to: shogi-unsubscribe <at> topica <dot> com

For Topica's complete suite of email marketing solutions visit:
<A HREF=3D"http://www.topica.com/?p=3DTEXFOOTER">http://www.topica.com/?p=
=3DTEXFOOTER</A>
--^^---------------------------------------------------------------</PRE>
</body>
</html>

--------------010609080403010102070503--