Re: computer-go: Most simple Go rules

[Nick Wedd <nick@xxxxxxxxxxxxx>]

Bob Myers <rtm@xxxxxxxxxxxxxxxxx> writes

>2) What ruleset would you consider ideal for a computer go tournament?

However this message is also in reply to a private email from Robert
Jasiek.


I have to specify the rules for the CG event in the European Go
Congress, in Dublin on August 1st.  Here is what I propose:

  Chinese (area) scoring.
  Standard ko rule, i.e. 2-cycles forbidden.
  Komi 6.5.
  No suicide allowed.

Here are my reasons.

__ Chinese (area) scoring.
    In Ing 2000, in David Fotland's words, "Three crucial games were
decided by a program messing around in the opponent's territory ([names
listed]). The worst offender was [name given], which played on for 90
moves after the dame were filled. In all three cases, the program
messing around eventually won, so it seems that for computer vs computer
go, this is an important feature."
    This kind of "messing around" is not pretty, and should not be
rewarded.  The best way to avoid rewarding it is for programs to make
defensive moves inside their own territory.  They will be unwilling to
do this if territory scoring is used.

__ Standard ko rule, i.e. 2-cycles forbidden.
    Implementing superko is difficult.  Some of the programs at Dublin
will not be world leaders, they will be newcomers to CG.  I do not want
to make things difficult for them.

__ Komi 6.5.
    I think that this is the right number.  I see no reason to prefer a
different number.  I don't feel strongly about it.

__ No suicide allowed.
    Some programs do not allow suicide.  If suicide is allowed, and is
attempted against such a program, it refuses to accept the move, and
therefore loses.
    Ing rules allow suicide.  If a program had entered Ing98, and played
random bad moves, except that it made a suicide move in every game, it
would have won against opponents which failed to accept suicide, and
scored above average.
    I do not want to encourage programs to make meaningless suicide
moves in the hope of winning in this way.  The ideal solution is to
persuade programmers to support suicide.  In a world-class event this
may be appropriate.  But in Dublin, I prefer to be supportive of new
programs.


Bob - I would like to see some standardisation of CG rules, and would
like to co-operate with you in this.

Nick
-- 
Nick Wedd