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

[Peter McKenzie <peter.mckenzie@xxxxxxxxx>]

I noticed this super ko rule a while back and decided to add a unit
test to my program to ensure it complies.  In designing my test I
tried to find the shortest possible game that had a superko.  Here is
what I came up with:

a2 b1 b2 c2 b3 d1 a3 a1 c1

Now the move b1 is illegal on KGS under Chinese rules (although legal
under all other KGS rule variants).

cheers,
Peter


On 8/13/05, Rémi Coulom <Remi.Coulom@xxxxxxxxxxxxxxxxx> wrote:
> Hi,
> 
> When testing my go program on KGS with Chinese rules, I noticed that
> super ko does not take the player to move into consideration. It seems
> that not all programs have the same behaviour in this respect. LeGoBot
> has a problem with this, as can be seen in this game:
> 
> http://kgs.kiseido.com/games/2005/7/22/LeGoBot-CrazyStone-2.sgf
> 
> I had to resign the game on behalf of CrazyStone to unlock LeGoBot. I
> noticed that GNU Go has the same problem when started with
> --chinese-rules. Maybe gnugo has an option that I missed ? In this game,
> it wanted to play J3 at the end, which my program refused:
> 
> (;FF[4]CA[ISO8859_1]GM[1]GN[cs0001-01-87]AP[TwoGtp:0.7.1]SZ[9]KM[6.5]RE[W+9.5]
> PB[GNU Go:3.6]PW[Crazy Stone - CMCSearchPlayer:0001.01]DT[2005-08-13]
> C[Black\: gnugo --mode gtp --chinese-rules
> White\: ./cs0001-01
> Result[Black\]\: W+9.5
> Result[White\]\: W+9.5
> Host\: zizi.grappa.univ-lille3.fr
> Date\: Saturday, August 13, 2005 1\:39\:16 AM CEST]
> ;B[ee];W[dg];B[eg];W[eh];B[df];W[cc];B[fg];W[cg];B[fh];W[ed]
> ;B[bg];W[cf];B[bf];W[de];B[ch];W[ef];B[dh];W[be];B[fe];W[bh]
> ;B[gc];W[ei];B[bi];W[ff];B[gf];W[ah];B[df];W[ef];B[fi];W[fd]
> ;B[gd];W[cd];B[eb];W[fc];B[fb];W[gb];B[ba];W[hc];B[hd];W[he]
> ;B[ge];W[db];B[ec];W[dd];B[da];W[ha];B[ib];W[ca];B[cb];W[dc]
> ;B[hb];W[ga];B[ea];W[bb];B[ab];W[ca];B[ag];W[af];B[cb];W[bc]
> ;B[ca];W[ac];B[ic];W[aa];B[ff];W[fa];B[ci];W[id];B[df];W[ie]
> ;B[ef];W[ia];B[hc];W[ai];B[di];W[eh];B[ei];W[ag];B[tt];W[hg]
> ;B[tt];W[hf];B[hh];W[ih];B[hi];W[ea];B[tt];W[gh];B[tt];W[gg]
> ;B[ig];W[if];B[tt];W[ii];B[ig];W[ae];B[tt];W[ih];B[tt];W[ii])
> 
> Also, in this game:
> http://kgs.kiseido.com/games/2005/8/1/SchokoBot-CrazyStone.sgf
> There was a seki at the end. This game was played with Japanese rules,
> so there may be a different rule for superko here. Instead of passing,
> if playing as White, my program would have played J1 and killed the
> black group (since D4 is super ko and cannot be played). If there is a
> superko rule, then I guess there is no seki anymore and one of the two
> groups has to die.
> 
> To summarize:
>  * Programmers, beware of KGS's behavior with respect to super ko in
> Chinese games.
>  * Is there an option in gnugo so that it plays with KGS rules ?
> 
> Rémi
> _______________________________________________
> computer-go mailing list
> computer-go@xxxxxxxxxxxxxxxxx
> http://www.computer-go.org/mailman/listinfo/computer-go/
>
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/