Re: [gnugo-devel] further test revision

[Daniel Bump]

Do the sgf files work from Trevor's regressions? Seems maybe broke.

Arend wrote:

> arb.tst:
> 200: Agree with Gunnar. I suggest to remove this problem, as there are too
> many plausible moves (M11, H4).

I agree this problem is clearly wrong but can wait for Teun (arb)
to maintain this file and the regression/games/arb directory. That
reminds me, we now have two contributors with the initials ab.

On the other hand Nicklas doesn't seem to be maintaining his regressions
any more so I used your patch for these tests.

> 1803: A6 is a nice tesuji.

I think A8 and A6 should both be accepted. The rationale for
A6 is that W can't fight the ko but that might be a little subtle
for GNU Go currently. A6 is the move played.

1102: You didn't mention this one but O9 should be accepted.
The ko will disappear after atari connect.

1205: Seems not such a good test.

> Nicklas5:
> 1203: I think K18 is playable, in fact I tend to favor it over the
> given alternatives. After B K18, W N18, B L19 (not N19 as in the game),
> W O18, B J16 B has a good attack and can make life at J18 if necessary.

Yes, it seems playable if B is willing to give up L15 as well as
the stones in the upper right. Here's some variations:

(;GM[1]FF[3]
RU[Japanese]SZ[19]
PW[GNU Go 3.1.13 (level 10) load and print]
PB[Unknown]
HA[0]KM[0.5]
GN[GNU Go 3.1.13 load and print Random Seed 1006053845]
AW[kb][qb][fc][hc][lc][qc][kd][ld][md][pd][ie][je][me][ne][oe][pe][nf][qi][dj][nj][pj][pk][rk][ql][rl][fm][pm][en][pn][do][cp][pp][cq][qq]
AB[lb][ob][pb][jc][kc][mc][nc][pc][dd][jd][nd][od][fe][ke][kf][of][cg][lh][nh][ph][ki][qk][ol][pl][dm][qm][rm][gn][fp][dq][kq][nq][br][cr][pr]
PL[1];B[jb];W[mb];B[ka];W[nb];B[id];W[hd];B[he];W[hf];B[ge];W[jg]
(;B[kg];W[ib]
(;B[ig];W[if];B[hg];W[gf])
(;B[if];W[jf];B[ig];W[jh])
)
(;B[hb];W[gb];B[ib]
(;W[kg];B[eb])
(;W[dc])
)
)

Dan