gnugo-devel
[Top][All Lists]
Advanced

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

[gnugo-devel] vital attack patterns having no effect


From: Paul Pogonyshev
Subject: [gnugo-devel] vital attack patterns having no effect
Date: Sun, 25 May 2003 23:28:17 -0400
User-agent: KMail/1.5.9

in the attached game, although a vital attack pattern matches,
gnu go still thinks the dragon is alive. at least two fixes are
needed: one in either owl.c or optics.c, one in dragon.c

first bug shows up when owl doesn't try vital pattern, because
compute_eye_pessimistic() returns eye value of 4444 and the
same pessimistic min. either we should have a policy to _always_
play out vital moves, or tweak guess_eye() in optics.c.

the second problem is in dragon.c. since dragon genus is 2222,
owl is never called to determine dragon status. apparently, we
cannot trust optics code in such important matter as dragon
life. however, i think it would be safe to reduce the heuristic
to something like "if we have _three_ _separate_ solid eyespaces,
each having minimum number of eyes of at least 1, the dragon is
alive". note that three eyespaces (should) guarantee against
vital attacks which join eyespaces together and thus reduce the
outcome.

Paul


see the very end of this game:

(;GM[1]FF[3]
RU[Japanese]SZ[19]HA[2]KM[0.5]
PW[White]
PB[Black]
GN[White (W) vs. Black (B)]
DT[2003-05-25]
SY[Cgoban 1.9.14]TM[-]AB[pd][dp];W[cd];B[pq];W[po];B[qo];W[qn];B[qp];
W[pn];B[nq];W[qj];B[ed];W[ee];B[fe];W[de];B[gd];W[cn];B[cl];W[cq];
B[dq];W[cp];B[do];W[bn];B[em];W[qf];B[hp];W[mo];B[lp];W[ci];B[nd];
W[ec];B[fc];W[eb];B[ek];W[ha];B[kn];W[rd];B[qc];W[rc];B[ll];W[id];
B[kc];W[mm];B[nk];W[ol];B[fi];W[if];B[ff];W[hg];B[gh];W[ii];B[hj];
W[ke];B[kk];W[dg];B[ki];W[ib];B[md];W[qb];B[pb];W[qh];B[ij];W[cr];
B[dr];W[rb];B[ro];W[rn];B[ih];W[kb];B[lb];W[jc];B[ka];W[jb];B[cs];
W[lc];B[mf];W[mc];B[mh];W[bs];B[ds];W[so];B[sp];W[sn];B[ma];W[pa];
B[pg];W[nb];B[ph];W[qg];B[oa];W[qa];B[ob];W[bk];B[bl];W[ar];B[bj];
W[bi];B[ak];W[ok];B[ai];W[nj];B[ag];W[be];B[cj];W[di];B[mk];W[mb];
B[na];W[pf];B[la];W[og];B[pi];W[oh];B[ld];W[qk];B[kd];W[op];B[jd];
W[je];B[mp];W[oq];B[or];W[rq];B[qr];W[rp];B[pp];W[np];B[lo];W[mn];
B[qd];W[rr];B[he];W[qs];B[pr];W[ps];B[os];W[rs];B[mr];W[gb];B[ie];
W[nf];B[fb];W[fa];B[jf];W[dd];B[dj];W[oe];B[hd];W[od];B[nc];W[gc];
B[fd];W[ic];B[dn];W[oc];B[pc];W[ng];B[co];W[bo];B[mg];W[nl];B[mj];
W[ni];B[hc];W[hb];B[mi];W[lm];B[km];W[ef];B[fg];W[eh];B[ei];W[qe];
B[bm];W[am];B[ja];W[ea];B[cf];W[cm];B[eg];W[df];B[cg];W[bh];B[ah];
W[dm];B[dl];W[an];B[al];W[ap];B[ne];W[pe];B[nh];W[oi];B[ae];W[ad];
B[af];W[bd];B[fh];W[dh];B[ml];W[ga];B[bf];W[ln];B[ia];W[bg];B[ch];
W[qq];B[cb];W[tt];B[dc];W[tt];B[tt]
)




reply via email to

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