Re: [gnugo-devel] t patterns audit

[max-d]

here GG demonstrates it makes fewer error in end game than Go++
(It is the single game i have run GG 3.4 level 12 vs Go++ lv 5)


move 192
at this moment gg estimates the score W+3.5 (+14 ,-7)
Go++ plays S2 ?
(gg says J1)
Then it seems that Go++ had no idea about what could happen...

229 (score is B+7)
at the end score is B+14
(Go++ says W +55 considering having killed black on the right edge )

Game 1

 --mode gtp --level 12 --boardsize 19 -M 8 --komi 5,5
Size 19
Black = gnugo_3.4 White = GO++
Level black = 12 Level white = 5
Time black = 476,8 Time white = 550,7
Score = = B+14.0 (upper bound: +14.0, lower: +14.0)
(
;B[pd]
;W[dd]
;B[pp]
;W[dp]
;B[pj]
;W[nc]
;B[lc]
;W[qc]
;B[qd]
;W[pc]
;B[od]
;W[nb]
;B[me]
;W[nq]
;B[lq]
;W[qp]
;B[qo]
;W[qq]
;B[po]
;W[pq]
;B[cn]
;W[fq]
;B[bp]
;W[cq]
;B[ck]
;W[li]
;B[fc]
;W[cf]
;B[db]
;W[cc]
;B[ic]
;W[ci]
;B[hq]
;W[bq]
;B[no]
;W[lk]
;B[nh]
;W[fe]
;B[ek]
;W[kn]
;B[nl]
;W[hj]
;B[jq]
;W[rd]
;B[re]
;W[rc]
;B[se]
;W[lb]
;B[kb]
;W[cb]
;B[en]
;W[mc]
;B[la]
;W[mb]
;B[ld]
;W[hm]
;B[ei]
;W[he]
;B[ro]
;W[rp]
;B[je]
;W[id]
;B[hc]
;W[bj]
;B[bk]
;W[jd]
;B[kf]
;W[ak]
;B[ji]
;W[gg]
;B[kh]
;W[jk]
;B[ho]
;W[ih]
;B[fo]
;W[bl]
;B[gm]
;W[gl]
;B[hn]
;W[im]
;B[fl]
;W[eb]
;B[fb]
;W[da]
;B[kd]
;W[mr]
;B[jc]
;W[kr]
;B[gk]
;W[hl]
;B[kq]
;W[lr]
;B[if]
;W[fj]
;B[ej]
;W[ir]
;B[gj]
;W[gi]
;B[hr]
;W[gr]
;B[fi]
;W[hk]
;B[fk]
;W[hs]
;B[sp]
;W[sq]
;B[so]
;W[cl]
;B[dl]
;W[dh]
;B[cj]
;W[bi]
;B[di]
;W[bo]
;B[co]
;W[cp]
;B[bn]
;W[ap]
;B[eh]
;W[dg]
;B[gh]
;W[hg]
;B[hi]
;W[ii]
;B[ll]
;W[kl]
;B[lm]
;W[km]
;B[op]
;W[lh]
;B[lg]
;W[nj]
;B[oj]
;W[ni]
;B[oi]
;W[nk]
;B[ok]
;W[mh]
;B[mg]
;W[mk]
;B[ml]
;W[jh]
;B[ki]
;W[kj]
;B[fg]
;W[jj]
;B[kg]
;W[ma]
;B[ef]
;W[ee]
;B[jr]
;W[js]
;B[ka]
;W[eg]
;B[ff]
;W[ge]
;B[oc]
;W[ob]
;B[oq]
;W[or]
;B[ie]
;W[mp]
;B[hd]
;W[do]
;B[dn]
;W[gn]
;B[gp]
;W[fm]
;B[gq]
;W[fp]
;B[fn]
;W[em]
;B[dm]
;W[mo]
;B[fr]
;W[gs]
;B[mn]
;W[eo]
;B[es]
;W[er]
;B[df]
;W[de]
;B[hh]
;W[gd]
;B[ln]
;W[gc]
;B[gb]
;W[nd]
;B[ne]
;W[sd]
;B[mq]
;W[np]
;B[ec]
;W[dc]
;B[iq]
;W[rr]
;B[is]
;W[ds]
;B[fs]
;W[gs]
;B[gr]
;W[dr]
;B[ks]
;W[oe]
;B[qf]
;W[jo]
;B[ls]
;W[ms]
;B[js]
;W[lo]
;B[jg]
;W[in]
;B[fa]
;W[ea]
;B[kp]
;W[ig]
;B[gf]
;W[hf]
;B[fh]
;W[nn]
;B[oo]
;W[ol]
;B[nm]
;W[om]
;B[on]
;W[qj]
;B[oh]
;W[ko]
;B[cg]
;W[ch]
;B[bg]
;W[bf]
;B[am]
;W[al]
;B[aj]
;W[ai]
;B[ah]
;W[dj]
;B[dk]
;W[af]
;B[bm]
;W[ao]
;B[cm]
;W[ip]
;B[io]
;W[jp]
;B[hp]
;W[pe]
;B[qe]
;W[nf]
;B[pf]
;W[qk]
;B[qi]
;W[qm]
;B[rj]
;W[rm]
;B[pk]
;W[pm]
;B[pl]
;W[sm]
;B[ql]
;W[sk]
;B[rl]
;W[sl]
;B[sj]
;W[rk]
;B[rh]
;W[si]
;B[ri]
;W[ed]
;B[hb]
;W[md]
;B[go]
;W[gm]
;B[fd]
;W[le]
;B[mf]
;W[el]
;B[lp]
;W[an]
;B[hs]
;W[tt]
;B[aj]
;W[ak]
;B[al]
;W[ag]
;B[aj]
;W[bh]
;B[ak]
;W[tt]
)

btw ,i also ran a 125 game level 10 match gg3.4 vs gg.3.3.16
i stopped after  122 games ...
result 61/61 !

As for the cosmic gnugo , i would like to run a test vs Go++ with the
version Arend got after his last message "patterns audit".
I have  not in mind to say you to make GG a go software just tuned against
Go++ but just verify Arend's feeling about the old patterns.
i already have a 89 game match against standard GG3.4 (GO++ 46 , gnugo_3.4
43)
Just need a windows binary to do that  since i think you can get more
precise  informations about "old patterns"  by an other 89 match GO++ vs
Arend GG.
md.