On Mar 16, 6:31 am, "Chess One" wrote:
Chessbase has put up a good article which these newsgroups also researched
this past January.

Here is the URL

http://www.chessbase.com/newsdetail.asp?newsid=4513

Hi, this is Mark Galeck, whose proposal was criticised in this
article by Mr Tiong. I take it on the chin, my proposal isn't perfect
and has it's flaws too. My proposal is not original, it is used in
some Go tournaments, but perhaps it is more psychologically suited to
Go than to Chess.

Anyhow, we had a correspondence with Mr Tiong about this, and I want
to notice one thing, about Mr Tiong's scoring system, which can be
described as: For win/draw/loss, B gets 1.1, 0.6, 0 : W gets 1, 0.4,
-0.1 ...

it turns out, that this system is equivalent to a simpler system, that
only changes the scores for draws compared to the traditional system.
So the adjustment to win and loss scores, is unnecessary. By
"equivalent", we mean that with both systems, the tournament tables at
the end of a tournament, would have the same order of players.

Here is the proof:

We start from Mr Tiong's system and step by step make equivalent
scoring systems. Let us assume first that every player plays the same
number of games with white (odd number of total players). So we can
modify Mr Tiong's system, increasing all white values by 0.1, and
obtain an equivalent system - for W/D/L black scores 1.1, 0.6, 0, and
white scores 1.1, 0.5, 0. (This system is equivalent to the previous
one because we just added to each player, the value of 0.1 x number of
white games).

Now, a digression - If a tournament has an even number of players,
then half of them will have played one more game with white (than the
others). If we modify Mr Tiong's system as above, we would have to
subtract 0.1 point for every one of those players to get to a system
that is equivalent to Mr Tiong's. Regardless of their results, half
of all players would all just get penalized. I think we agree that
this is not a change for the better. So, in case of an even number of
players, the system with 1.1, 0.6, 0 for black, and 1.1, 0.5, 0 for
white, is even slightly better than Mr Tiong's (but really it does not
make much of a difference, and is not important - this is just an
inherent flaw with tournaments with even number of players).

Now we divide all values by 1.1 and again obtain an equivalent
system: black: 1, 6/11, 0, white: 1, 5/11, 0. As you can see we
can get rid of most of the modifications of Mr Tiong's proposal, and
just use the suitable modification for the draw results, black 1/2 +
some constant c, white: 1/2 - c, and obtain an equivalent system (or,
in the case of even number of players, a slightly better system). .

End-of-proof

So I think both me and Mr Tiong, and reasonable people would have to
agree, the only question is: how to determine "c". This is exactly
the question I posed in my first proposal quoted on Chessbase. I
proposed an auction system. I agree this is a bit complicated and
unwieldy.

On the other hand, remember, what the goal is: to get rid of
"grandmaster draws" (short unfought draws). I submit to you that if
we just go for the kill (Topalov style), and set c = 1/2, that is for
the draw, black gets 1, white gets 0, there would be ***NO***
grandmaster draws. Period. We would immediately solve the problem.

Of course, that would have it's own flaw. Namely, black would always
play for draw, no incentive to play for the win. OK, then let's fudge
a little, I don't know, I leave that to the discussion of experts,
let's assign a large enough value for c. Perhaps c = 0.2, that is 0.7
for black draw, 0.3 for white draw, would suffice? If we play
tournaments this way and it turns out that is not enough and there are
still some unfought draws, try c = 0.3. I think one can choose this
constant to be large enough to prevent most if not all of unfought
draws, and still make black worthwhile to play for a win, in most if
not all cases.

Mark Galeck

Hi Mark Galek,

Along with other people who wish to address this subject, I wonder if you
would like to address it to a broad public?

Please e-mail me if you would like to do so, and perhaps we can use
Chessville's pages to substantially air these timely issues, on which there
is a broad range of opinion world-wide.

Cordially, Phil Innes

"Mark_Galeck" wrote in message
...
On Mar 16, 6:31 am, "Chess One" wrote:
Chessbase has put up a good article which these newsgroups also
researched
this past January.

Here is the URL

http://www.chessbase.com/newsdetail.asp?newsid=4513

Hi, this is Mark Galeck, whose proposal was criticised in this
article by Mr Tiong. I take it on the chin, my proposal isn't perfect
and has it's flaws too. My proposal is not original, it is used in
some Go tournaments, but perhaps it is more psychologically suited to
Go than to Chess.

Anyhow, we had a correspondence with Mr Tiong about this, and I want
to notice one thing, about Mr Tiong's scoring system, which can be
described as: For win/draw/loss, B gets 1.1, 0.6, 0 : W gets 1, 0.4,
-0.1 ...

it turns out, that this system is equivalent to a simpler system, that
only changes the scores for draws compared to the traditional system.
So the adjustment to win and loss scores, is unnecessary. By
"equivalent", we mean that with both systems, the tournament tables at
the end of a tournament, would have the same order of players.

Here is the proof:

We start from Mr Tiong's system and step by step make equivalent
scoring systems. Let us assume first that every player plays the same
number of games with white (odd number of total players). So we can
modify Mr Tiong's system, increasing all white values by 0.1, and
obtain an equivalent system - for W/D/L black scores 1.1, 0.6, 0, and
white scores 1.1, 0.5, 0. (This system is equivalent to the previous
one because we just added to each player, the value of 0.1 x number of
white games).

Now, a digression - If a tournament has an even number of players,
then half of them will have played one more game with white (than the
others). If we modify Mr Tiong's system as above, we would have to
subtract 0.1 point for every one of those players to get to a system
that is equivalent to Mr Tiong's. Regardless of their results, half
of all players would all just get penalized. I think we agree that
this is not a change for the better. So, in case of an even number of
players, the system with 1.1, 0.6, 0 for black, and 1.1, 0.5, 0 for
white, is even slightly better than Mr Tiong's (but really it does not
make much of a difference, and is not important - this is just an
inherent flaw with tournaments with even number of players).

Now we divide all values by 1.1 and again obtain an equivalent
system: black: 1, 6/11, 0, white: 1, 5/11, 0. As you can see we
can get rid of most of the modifications of Mr Tiong's proposal, and
just use the suitable modification for the draw results, black 1/2 +
some constant c, white: 1/2 - c, and obtain an equivalent system (or,
in the case of even number of players, a slightly better system). .

End-of-proof

So I think both me and Mr Tiong, and reasonable people would have to
agree, the only question is: how to determine "c". This is exactly
the question I posed in my first proposal quoted on Chessbase. I
proposed an auction system. I agree this is a bit complicated and
unwieldy.

On the other hand, remember, what the goal is: to get rid of
"grandmaster draws" (short unfought draws). I submit to you that if
we just go for the kill (Topalov style), and set c = 1/2, that is for
the draw, black gets 1, white gets 0, there would be ***NO***
grandmaster draws. Period. We would immediately solve the problem.

Of course, that would have it's own flaw. Namely, black would always
play for draw, no incentive to play for the win. OK, then let's fudge
a little, I don't know, I leave that to the discussion of experts,
let's assign a large enough value for c. Perhaps c = 0.2, that is 0.7
for black draw, 0.3 for white draw, would suffice? If we play
tournaments this way and it turns out that is not enough and there are
still some unfought draws, try c = 0.3. I think one can choose this
constant to be large enough to prevent most if not all of unfought
draws, and still make black worthwhile to play for a win, in most if
not all cases.

Mark Galeck