[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [computer-go] Super Ko on KGS ignores player to move

Interesting. So you are saying that sites such as
http://www.britgo.org/rules/compare.html have it wrong? Or that the
examples in Chapter 3, Section 20 at
http://www.cs.cmu.edu/~wjh/go/rules/Chinese.html have been removed
from the official rules? Could you point us to a more authoritative
and up-to-date source?

IMO chapter 1 one section 6 alone is described too informal to draw
any strong conclusions without taking the full text, including
examples and commentaries into account. (For example, it is not even
specified what exactly a position is.) If one does this it is possible
to design logical rules that are in much better agreement than what
has been implemented now.

Personally I quite like Makrai's idea to declare a draw when a
position (including prisoner difference) repeats (in a long cycle) for
the 3rd time. It effectively removes any false/unbalanced cycles and
appears to be in excellent agreement with the official rules. (For
more info see http://dnc.yogar.hu/go/rules.html )

On 8/14/05, William M. Shubert <wms@xxxxxxxxxxxxxxxxx> wrote:
> John has it right. The chinese exact rules specify positional superko,
> then they have generalizations that indicate that the TD may overrule
> superko. The KGS source code of "Rules.java" contains this function:
>   public final int getKoType() {
>     if (type == JAPANESE) {
>       return(KO_SIMPLE);
>     } else if (type == CHINESE) {
>       return(KO_SAMEBOARD);
>     } else {
>       return(KO_SAMESTATE);
>     }
>   }

Which is incorrect when a position includes triple ko, quadruple ko,
eternal life, round-robin ko, Cho-sei, Molasses Ko, Shimada's ko, etc.
Further, but that's more relevant for programmers than for human play,
it even undermines the application of Benson's unconditional life...

> Which is hopefully clear enough to a programmer to explain exactly what
> to expect ("type" is a member variable of a rules description). 

It's perfectly clear what you intend, and (for me) it is not too
difficult to cripple my program to comply to such rules. However,
until you come up with more authoritative and up-to-date sources about
the Chinese rules I simply don't believe this to be "the most accurate
interpretation possible".

computer-go mailing list