Re: computer-go: Most simple Go rules

[Robert Jasiek <jasiek@xxxxxxxx>]

Bob Myers wrote:
> 1) In what respect do you consider the AGA rules "logical"?

"logical" or "well-defined" rules have the following feature:

WHAT IS DEFINED CAN ALWAYS BE APPLIED.
 
This is only the case if the rules are complete, i.e. the
objects mentioned in them have a treatment in any state 
that they might ever assume. E.g. regardless of the actual
"position" a "move" can always be executed. E.g. regardless 
of the "position" at the "game end" the "score" can always
be determined.

This is only the case if the rules are unequivocal, i.e.
every state that the rules objects might ever assume has
one and only one possible treatment. E.g. for any
particular "position" and "prior sequence of positions" the 
set of all "intersections" that the actual "move" may not be 
played on is a particular subset of all "intersections". E.g.
at the "game end" the "position" has one and only one possible
"score value".

The AGA 1991 rules can always be applied because they are
complete and unequivocal. (However, the wording is a little
tough, non-formal and it requires some effort to realize 
their unequivocality. Furthermore, the AGA rules include
tournament rules and options for them so that one has to 
agree on some options before a start of the game; the handicap
system and method of counting should be specified in advance.)

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

A ruleset for tournament go always consists of
A) core rules of play (the minimally necessary part of all
rules to describe what a legal game and who the winner is),
B) tournament rules (all rules besides the core rules of play).

(B) includes things like mane-go restriction, if any, time,
handicap, komi, sportsmanship, tournament system, method of
counting, method of handling the game end, if for some obscure
reason it shall be more complex than just ending the game due 
to passes, plagiarism, penalties, participation, seeding, 
arbitration, organization, sponsoring, hardware, etc.
For a start, I suggest only (A) in a separate thread. (B) can
be discussed later. My suggestion uses my own preferences for
ko rules and for removals of own stones, a wording derived 
from the International Rules http://home.snafu.de/jasiek/int.html,
the methodically simple game end due to nothing but two
successive passes, and the simplest scoring that determines the
winner as expected by any go rules used in the word in almost
all cases. The scoring is informally known as "Chinese" and is
that of Tromp-Taylor Rules. The most important point is the
game end handling (nothing but two successive passes). Once one
would start to include agreement or confirmation phases, one
gets into trouble; humans may be able to agree rather easily in
practice but programs do not have any easy language of agreement,
would have to resort to human programmers, and a complex
interaction of program agreement, player agreement, and 
supervision by the tournament organizers would be the very nasty
consequence.

--
robert jasiek
http://home.snafu.de/jasiek/rules.html