gnugo-devel
[Top][All Lists]
Advanced

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

[gnugo-devel] some tests , one bug


From: max-d
Subject: [gnugo-devel] some tests , one bug
Date: Sun, 20 Apr 2003 02:38:10 +0200

HI,
i have made some tests between GnuGo 3.3.16 and 3.2
(65 games level 10, size 19).

3.3.16 won 50 games , 3.2 won 15 games.


i selected a game lost by gg 3.3.16
(see the 2 sgf files at the end of this message)

it seems to me that things turn wrong at move 87

as i have often observed 3.3.16 has problems to stop invasions.
btw ,Gnugo 3.2 plays the "correct" 87: J11 but its top move-list strangely
says that 87:j14 (3.3.16's move)is better by almost 2 points.

(notice that 3.3.16 plays the good move at level 14).


3.3.16 scores the position 11.5 for white after move  86: K12
i have played black at this stage of the game (2nd sgf file)...
at the end of the 2nd sgf file ,
1/ it scores white +6.5 but it seems to me black position is far more
better now than after move 86.
(i don't know , but maybe this is why it plays  these situations wrongly)
2/ 3.3.16 and 3.2 are never able to punish me (even at level 15) after i
played 103:A10 ???..
(That seems a bug)
whireas gnugo 3.0 can do that very well !

hope this can help

(i had also made 11 games vs 3.0 .
 3.3.16 won all games so i stopped.
Also 10 games : 3.3.16 vs itself ,black using 40 Mb hashtable , white 8 mb
...
black won 8/2 but playing time was almost even  . i believed increasing the
tables size had just for effect to decrease the time needed .I am wrong here
? ) .

md.

sgf 1

black =3.3.16 , w=3.2
level 10

(
;B[pd]
;W[dp]
;B[pp]
;W[dd]
;B[fc]
;W[df]
;B[db]
;W[gc]
;B[gd]
;W[hc]
;B[dn]
;W[fp]
;B[cp]
;W[cq]
;B[co]
;W[cc]
;B[bq]
;W[fd]
;B[dq]
;W[nc]
;B[fe]
;W[ed]
;B[he]
;W[pn]
;B[np]
;W[pj]
;B[qo]
;W[kc]
;B[qf]
;W[pb]
;B[qc]
;W[lq]
;B[nq]
;W[iq]
;B[ch]
;W[eq]
;B[ck]
;W[dr]
;B[qb]
;W[br]
;B[fn]
;W[cg]
;B[dh]
;W[fb]
;B[dg]
;W[bg]
;B[ef]
;W[bh]
;B[hn]
;W[qn]
;B[ro]
;W[kg]
;B[ih]
;W[kk]
;B[ij]
;W[ik]
;B[hk]
;W[bp]
;B[nn]
;W[bo]
;B[rn]
;W[il]
;B[ob]
;W[hl]
;B[gk]
;W[nb]
;B[gl]
;W[oc]
;B[qm]
;W[nj]
;B[jn]
;W[jj]
;B[ji]
;W[lj]
;B[rj]
;W[pm]
;B[ql]
;W[pl]
;B[qh]
;W[ki]
;B[nf]
;W[qk]
;B[rk]
;W[ig]
;B[hg]
;W[jh]
;B[if]
;W[ii]
;B[hj]
;W[hh]
;B[gg]
;W[gi]
;B[cn]
;W[ee]
;B[fg]
;W[mr]
;B[nr]
;W[qa]
;B[bn]
;W[rb]
;B[rc]
;W[fj]
;B[ek]
;W[bi]
;B[ej]
;W[ns]
;B[os]
;W[ms]
;B[pr]
;W[nm]
;B[hd]
;W[ei]
;B[di]
;W[ao]
;B[jd]
;W[ic]
;B[an]
;W[aq]
;B[ho]
;W[hp]
;B[nh]
;W[ph]
;B[pg]
;W[qi]
;B[ri]
;W[qj]
;B[oh]
;W[bk]
;B[bl]
;W[cj]
;B[dj]
;W[al]
;B[cl]
;W[hm]
;B[am]
;W[ak]
;B[mn]
;W[gm]
;B[lm]
;W[gn]
;B[sb]
;W[go]
;B[ra]
;W[fm]
;B[em]
;W[pa]
;B[jo]
;W[mm]
;B[ip]
;W[jq]
;B[ll]
;W[jc]
;B[hi]
;W[gh]
;B[eh]
;W[fi]
;B[kp]
;W[jf]
;B[je]
;W[ke]
;B[kq]
;W[kr]
;B[jp]
;W[kd]
;B[im]
;W[jl]
;B[lp]
;W[mq]
;B[mp]
;W[ls]
;B[ci]
;W[bj]
;B[jm]
;W[fo]
;B[en]
;W[pi]
;B[id]
;W[fl]
;B[fk]
;W[me]
;B[mf]
;W[lf]
;B[mi]
;W[mk]
;B[lh]
;W[lg]
;B[mg]
;W[ml]
;B[ne]
;W[md]
;B[kh]
;W[li]
;B[mh]
;W[oi]
;B[el]
;W[eo]
;B[od]
;W[kl]
;B[km]
;W[on]
;B[oo]
;W[fh]
;B[hq]
;W[gp]
;B[pc]
;W[oa]
;B[mj]
;W[lk]
;B[nd]
;W[po]
;B[ni]
;W[ok]
;B[do]
;W[ih]
;B[gj]
;W[tt]
)

sgf 2
me =black after move 86

(
;B[pd]
;W[dp]
;B[pp]
;W[dd]
;B[fc]
;W[df]
;B[db]
;W[gc]
;B[gd]
;W[hc]
;B[dn]
;W[fp]
;B[cp]
;W[cq]
;B[co]
;W[cc]
;B[bq]
;W[fd]
;B[dq]
;W[nc]
;B[fe]
;W[ed]
;B[he]
;W[pn]
;B[np]
;W[pj]
;B[qo]
;W[kc]
;B[qf]
;W[pb]
;B[qc]
;W[lq]
;B[nq]
;W[iq]
;B[ch]
;W[eq]
;B[ck]
;W[dr]
;B[qb]
;W[br]
;B[fn]
;W[cg]
;B[dh]
;W[fb]
;B[dg]
;W[bg]
;B[ef]
;W[bh]
;B[hn]
;W[qn]
;B[ro]
;W[kg]
;B[ih]
;W[kk]
;B[ij]
;W[ik]
;B[hk]
;W[bp]
;B[nn]
;W[bo]
;B[rn]
;W[il]
;B[ob]
;W[hl]
;B[gk]
;W[nb]
;B[gl]
;W[oc]
;B[qm]
;W[nj]
;B[jn]
;W[jj]
;B[ji]
;W[lj]
;B[rj]
;W[pm]
;B[ql]
;W[pl]
;B[qh]
;W[ki]
;B[nf]
;W[qk]
;B[rk]
;W[ig]
;B[hg]
;W[jh]
;B[ii]
;W[mm]
;B[bi]
;W[ce]
;B[bn]
;W[mr]
;B[nr]
;W[qa]
;B[om]
;W[ol]
;B[jp]
;W[jq]
;B[kp]
;W[rb]
;B[rc]
;W[ai]
;B[aj]
;W[ip]
;B[hp]
;W[hq]
;B[gp]
;W[gq]
;B[io]
;W[ah]
)






reply via email to

[Prev in Thread] Current Thread [Next in Thread]