[gnugo-devel] A bug on Chinese scoring with seki eyes?

[Go Fast]

Not sure whether it was reported before or not, there seems to be a bug on scoring by GnuGo when there is a seki eye in the final position. In my understanding, a seki eye if completely surrounded by one color, should be solely assigned to the surrounding side in Chinese rules. However, GnuGo does not assigned it to the correct side.

Copied below is the sgf text of a sample game. In Chinese rules, the correct results is W+0.5. However, GnuGo is giving B+0.5. The position G1 should belong to the white side.

The command line argument I used is gnugo --mode gtp --chinese-rules


(;FF[4]SZ[9]KM[7.5]
;B[ee];W[gf];B[ff];W[bd];B[ge];W[he];B[gd];W[gg];B[hd];W[cc];B[eh];W[fh];B[fg];W[hf]
;B[hh];W[gh];B[ie];W[hi];B[ih];W[eg];B[dh];W[ch];B[cg];W[dg];B[bh];W[cf];B[ci];W[bg]
;B[ch];W[if];B[df];W[fi];B[id];W[eb];B[ce];W[ed];B[gb];W[be];B[bf];W[fd];B[fc];W[fe]
;B[ef];W[ec];B[hc];W[fb];B[ib];W[ig];B[ei];W[ga];B[ha];W[fa];B[ae];W[dd];B[gc];W[ac]
;B[de];W[cd];B[af];W[ca];B[ad];W[bb];B[dg];W[db];B[ag];W[aa];B[];W[])