KGS scoring discrepancy?
Mike Yankee
This surprised me (black) because I was quite sure I had won the
game.
When I downloaded and reviewed the sgf file, GnuGo and Many Faces
scored it as B +9.5. But CGoban still counted it as W +0.5.
It was a simple, uncomplicated board position with clear territorial
lines, no sekis, no unplayed teire, etc.
This is quite a large discrepancy and one that changed the outcome.
Fortunately it wasn't a rated game, but I wonder:
a. Why such a discrepancy?
b. Had it been a rated game on KGS, could the result be contested?
Can anyone help me understand this? If so, thanks in advance.
I'm posting the sgf below:
(;GM[1]FF[4]CA[UTF-8]AP[CGoban:3]ST[2]RU[Chinese]SZ[19]KM[7.50]TM[1500]
OT[5x30 byo-yomi]PW[fullgote]PB[mikeyango]BR[8k]DT[2009-10-21]PC[The
KGS Go Server at http://www.gokgs.com/]C[fullgote [-\]: GTP Engine for
fullgote (white): Fuego version 0.4]RE[W+0.50];B[pd]BL[1496.184];W[dd]
WL[1499.748];B[pq]BL[1494.26];W[cp]WL[1499.497];B[qk]BL[1493.341];W[qj]
WL[1483.983];B[pj]BL[1486.502];W[pk]WL[1469.751];B[qi]BL[1475.576];W
[rj]WL[1460.378];B[ri]BL[1472.206];W[oj]WL[1444.639];B[pi]BL
[1464.042];W[pl]WL[1437.053];B[sj]BL[1461.615];W[rk]WL[1417.302];B[sk]
BL[1458.604];W[rl]WL[1405.433];B[sl]BL[1457.13];W[oi]WL[1390.086](;B
[oh]BL[1455.857];W[nh]WL[1349.428];B[og]BL[1450.009];W[ng]WL
[1337.394];B[of]BL[1433.596];W[ph]WL[1323.582];B[qh]BL[1428.818];W[nf]
WL[1306.116];B[ne]BL[1415.22];W[me]WL[1291.757];B[nd]BL[1412.014];W[md]
WL[1280.341];B[nc]BL[1404.589];W[gq]WL[1262.892];B[rm]BL[1398.633];W
[mc]WL[1246.868];B[ql]BL[1379.856];W[ni]WL[1231.551];B[pm]BL
[1377.263];W[nl]WL[1215.003];B[oo]BL[1362.019];W[ck]WL[1201.044];B[kq]
BL[1351.742];W[hh]WL[1185.646];B[mn]BL[1313.684];W[nb]WL[1168.731];B
[ob]BL[1310.4];W[ch]WL[1152.318];B[cc]BL[1280.811];W[dc]WL[1138.482];B
[cd]BL[1277.992];W[cb]WL[1125.899];B[bb]BL[1231.83];W[ce]WL[1116.945];B
[db]BL[1226.926];W[eb]WL[1106.031];B[da]BL[1221.015];W[de]WL
[1090.049];B[bd]BL[1219.429];W[ea]WL[1077.17];B[ca]BL[1215.97];W[km]WL
[1065.325];B[ko]BL[1210.457];W[eq]WL[1049.993];B[iq]BL[1157.057];W[hj]
WL[1034.627];B[in]BL[1152.641];W[hm]WL[1022.438];B[jm]BL[1119.949];W
[jl]WL[1013.019];B[kn]BL[1112.038];W[im]WL[1001.234];B[hn]BL
[1104.488];W[gn]WL[992.505];B[lm]BL[1094.095];W[kl]WL[980.95];B[ll]BL
[1091.867];W[lk]WL[972.041];B[ml]BL[1089.984];W[mk]WL[960.586];B[nm]BL
[1088.471];W[be]WL[947.177];B[ae]BL[1085.463];W[mb]WL[934.529](;B[pb]BL
[1083.491];W[hb]WL[922.565];B[go]BL[1025.069];W[fo]WL[915.711];B[gp]BL
[1018.941];W[fp]WL[906.442];B[hq]BL[1011.422];W[cn]WL[894.945];B[jn]BL
[1007.507];W[ec]WL[883.861];B[gr]BL[1003.076];W[fn]WL[873.917];B[ag]BL
[981.137];W[af]WL[863.881];B[bf]BL[979.039];W[cf]WL[852.484];B[bh]BL
[976.028];W[bi]WL[844.859];B[ai]BL[964.411];W[bj]WL[834.822];B[aj]BL
[961.869];W[ak]WL[827.587];B[ah]BL[954.765];W[bg]WL[815.214];B[cg]BL
[949.521];W[dg]WL[809.601];B[af]BL[946.719];W[ad]WL[799.756];B[ac]BL
[941.574];W[ie]WL[788.338];B[ds]BL[874.668];W[bq]WL[776.552];B[cr]BL
[866.902];W[nk]WL[769.174];B[br]BL[859.841];W[cq]WL[761.159];B[er]BL
[852.014];W[kb]WL[750.007];B[fq]BL[847.853];W[ar]WL[739.479];B[dq]BL
[842.023];W[dp]WL[732.992];B[ep]BL[839.637];W[eo]WL[727.186];B[eq]BL
[838.828];W[bg]WL[720.549];B[ad]BL[831.994];W[cg]WL[716.555];B[na]BL
[827.49];W[ma]WL[710.842];B[oa]BL[826.298];W[bl]WL[701.868];B[ol]BL
[787.538];W[ok]WL[696.349];B[om]BL[786.83];W[gm]WL[689.06];B[bs]BL
[784.047];W[do]WL[678.787];B[as]BL[781.56];W[aq]WL[673.442];B[kh]BL
[780.114];W[jh]WL[667.451];B[kj]BL[773.617];W[kg]WL[660.775];B[kk]BL
[771.668];W[jk]WL[654.823];B[lh]BL[767.246];W[lg]WL[647.33];B[jj]BL
[737.241];W[hk]WL[637.459];B[ji]BL[729.847];W[mf]WL[630.826];B[ih]BL
[724.042];W[jg]WL[625.885];B[hi]BL[699.528];W[hg]WL[623.889];B[gi]BL
[697.234];W[ig]WL[616.319];B[ii]BL[695.043];W[fi]WL[610.849];B[gj]BL
[686.86];W[gk]WL[606.286];B[fj]BL[680.318];W[ei]WL[601.479];B[ej]BL
[676.205];W[dj]WL[597.125];B[di]BL[665.388];W[ci]WL[591.954];B[eh]BL
[637.573];W[fh]WL[586.067];B[dk]BL[629.621];W[fk]WL[580.221];B[cj]BL
[627.659];W[cl]WL[575.919];B[ek]BL[625.626];W[el]WL[571.113];B[dl]BL
[605.832];W[fl]WL[567.39];B[]BL[579.691];W[dj]WL[562.844];B[]BL
[579.691];W[dh]WL[558.344];B[]BL[579.691];W[eg]WL[553.881];B[]BL
[560.202];W[ic]WL[547.079];B[]BL[560.202];W[pr]WL[539.873];B[or]BL
[557.394];W[qr]WL[535.993];B[qq]BL[549.057];W[rq]WL[532.415];B[rp]BL
[542.932];W[sq]WL[529.361];B[sp]BL[537.978];W[rr]WL[525.771];B[rs]BL
[526.143];W[qs]WL[523.403];B[ss]BL[517.6];W[sr]WL[519.178];B[oq]BL
[512.625];W[qp]WL[514.555];B[qo]BL[502.695];W[ro]WL[507.016];B[pp]BL
[499.025];W[so]WL[503.974];B[qp]BL[496.435];W[os]WL[498.093];B[ns]BL
[479.716];W[ps]WL[496.224];B[nr]BL[476.057];W[qn]WL[493.208];B[]BL
[476.055];W[dm]WL[488.179];B[]BL[476.055];W[ho]WL[484.133];B[hp]BL
[472.557];W[io]WL[480.595];B[]BL[472.557];W[lj]WL[474.095];B[]BL
[472.557];W[em]WL[468.299];B[]BL[472.557];W[qm]WL[464.511];B[sm]BL
[463.424];W[rn]WL[460.797];B[]BL[463.424];W[oc]WL[456.469];B[pc]BL
[459.342];W[qd]WL[452.578];B[qc]BL[449.332];W[rd]WL[447.622];B[rc]BL
[446.564];W[sc]WL[444.7];B[sb]BL[441.719];W[sd]WL[441.795];B[rb]BL
[436.16];W[]WL[441.795];B[]BL[436.159]TW[fa][ga][ha][ia][ja][ka][la]
[fb][gb][ib][jb][lb][fc][gc][hc][jc][kc][lc][ed][fd][gd][hd][id][jd]
[kd][ld][ee][fe][ge][he][je][ke][le][df][ef][ff][gf][hf][if][jf][kf]
[lf][fg][gg][mg][eh][gh][ih][kh][lh][mh][di][gi][hi][ii][ji][ki][li]
[mi][cj][ej][fj][gj][ij][jj][kj][mj][nj][bk][dk][ek][ik][kk][al][dl]
[gl][hl][il][am][bm][cm][fm][an][bn][dn][en][ao][bo][co][ap][bp]TB[aa]
[ba][pa][qa][ra][sa][ab][cb][qb][bc][oc][sc][od][qd][rd][sd][oe][pe]
[qe][re][se][pf][qf][rf][sf][pg][qg][rg][sg][ph][rh][sh][si][qj][rj]
[rk][rl][mm][qm][ln][nn][on][pn][qn][rn][sn][ho][io][jo][lo][mo][no]
[po][ro][so][ip][jp][kp][lp][mp][np][op][gq][jq][lq][mq][nq][rq][sq]
[dr][fr][hr][ir][jr][kr][lr][mr][pr][qr][rr][sr][cs][es][fs][gs][hs]
[is][js][ks][ls][ms][os][ps][qs][rs][ss])(;B[na]BL[1062.593];W[oc]WL
[924.744]))(;B[nh]BL[1455.587])(;B[oh]BL[1451.172]))
eric_...@alum.mit.edu
> Last night KGS gave my opponent (white) the victory by half a point.
> This surprised me (black) because I was quite sure I had won the
> game.
>
> When I downloaded and reviewed the sgf file, GnuGo and Many Faces
> scored it as B +9.5. But CGoban still counted it as W +0.5.
KGS used Chinese scoring, because apparently that is what was agreed
on before the game started. Probably that's because most Go robots
understand only Chinese rules, as they are simpler to program than
Japanese rules. However, GnuGo defaults to Japanese scoring, and Many
Faces probably does as well. Unfortunately, it appears that neather
GnuGo nor Many Faces read the scoring method out of the SGF file.
Perhaps this is not too surprising, since SGF standard defines only a
few rulesets, and Chinese is not one of them.
Usually there is not much difference between these methods of
scoring. However, in this game there are a bunch of places where
Black passes, but white keeps playing (inside white territory). Under
Chinese rules, this has no effect on the score. Under Japanese rules,
each time White played, the White score was reduced by one. GnuGo, at
least, can be told to use chinese scoring:
> gnugo --score estimate test.sgf
Black wins by 9.5 points
> gnugo --chinese-rules --score estimate test.sgf
White wins by 0.5 points
--
Eric
Nick Wedd
<8f2082ee-584c-44e5...@l31g2000vbp.googlegroups.com>,
Mike Yankee <mikey...@aol.com> writes
>Last night KGS gave my opponent (white) the victory by half a point.
>This surprised me (black) because I was quite sure I had won the
>game.
>
>When I downloaded and reviewed the sgf file, GnuGo and Many Faces
>scored it as B +9.5. But CGoban still counted it as W +0.5.
When I count it, I make White half a point ahead (White has 177 points
on the board + 7.5 komi = 184.5, Black has 184 points on the board). KGS
agrees. My copy of Many Faces of Go (version 12.001) also agrees.
My guess is that your versions of GNU Go and of Many Faces of Go were
unaware that the game was set to Chinese scoring.
>
>It was a simple, uncomplicated board position with clear territorial
>lines, no sekis, no unplayed teire, etc.
>
>This is quite a large discrepancy and one that changed the outcome.
>Fortunately it wasn't a rated game, but I wonder:
>
>a. Why such a discrepancy?
Because of the difference between Japanese and Chinese rules. About ten
times in the course of the game, Black passed while White continued to
make defensive moves. These exchanges (of a pass for an unnecessary
defensive move) have no effect on the score if you are using Chinese
rules, but benefit Black if you are using Japanese rules.
The default on KGS is Japanese rules. However this game, with the
agreement or acquiescence of the players, was played with Chinese rules.
>b. Had it been a rated game on KGS, could the result be contested?
No. But as completed (i.e. all the dame filled) games are as far as I
know always counted correctly, this is not a problem.
Nick
--
Nick Wedd ni...@maproom.co.uk