(;GM[1]FF[3]
RU[Japanese]BR[]WR[]CP[]SZ[9]HA[0]KM[5.5]
PW[GNU Go]
PB[GoLife I]
GN[GNU Go(W) vs. GoLife I (B)]
DT[2000-11-02]
SY[Cgoban 1.9.2]TM[30:00(5x30:00)]
C[This is a game between GNU Go version 2.7.187 and GoLife I played on november
2nd 2000. The game was for a position on the computer go 9x9 ladder. The result
is that white wins by 49.5 - 5 (44.5) after a big fight with many mistakes.
]
;B[fd]BL[1800];W[gc]WL[1800]
C[This move is taken from a database of fuseki moves from games between dan
players or low kyu players.
]
;B[cg]BL[1800];W[fc]WL[1800];B[ed]BL[1800];W[db]WL[1800]LB[dc:A]
C[A was GNU Gos second favorite move.
];B[ge]BL[1800];W[cd]WL[1800];
B[hd]BL[1800]LB[fg:A]
C[GNU Go would have played at A to stop the invasion at the 3-3 point in the lower right.
I believe that A is bigger than the game move, but maybe not because white
will have a lot of sente moves to destroy blacks territory even after A.
]
;W[gg]WL[1800]C[This invasion is GNU Gos only chance to win.
];B[bc]
BL[1800]
C[This is a somewhat strange move. It can be viewed in many ways. A high level
interpretation is that B7 is a probe allowing a continuation at B5 or D7, but
I doubt that this is what GoLife was thinking. GNU Go would have played at A
hopefully killing G3.
]
LB[eg:A];W[hf]WL[1800]
C[Doubtful. I think that E3 is bigger. Both moves probably live, but E3 is bigger and
leaves fewer threats. Strangely enough, GNU Go thinks that B7 is dead although
it can clearly connect to C3 atB5.
]
;B[cf]BL[1800]
C[Now white really has to reinforce the white stones at the top. The best move is probably
D7 since it both connects with C6 and removes blacks big move at D7.
]
;W[bd]WL[1800]
C[This move links all the white stones together.  If black plays D7 now, white can answer at
C7 and black can do nothing to prevent the connection.
]
;B[dd]BL[1800]C[GoLife doesn't see this big threat, however.
];W[ac]
WL[1800]
C[This move is due to a bug in GNU Go.  Some of the points here like A7 (A), B8 (B), 
C7 (C) and C8 (D) are overvalued. The correct move is D7 (E) which connects everything.
]
LB[ac:A][bb:B][cc:C][cb:D][dc:E];B[ab]BL[1800]
C[But GoLife plays badly also and doesn't punish the mistakes of GNU Go.
]
;W[cc]WL[1800]
C[Now everything is saved, but with highly suboptimal shape.
];B[gb]
BL[1800]C[Futile. After white H7, black can neither kill nor cut.
];
W[hc]WL[1800];B[hb]BL[1800];W[ib]WL[1800];B[ia]BL[1800];W[fb]WL[1800]
C[This move is due to a bug. GNU Go thinks that F8 attacks G8 which, of course, it
doesn't. A much better move, J7, was valued second highest. The correct move, H9,
got a still lower score.
]
;B[bb]BL[1800]
C[But GoLife doesn't use its advantage. Instead it tries to save the B7 stone. 
Although GNU Go failed to see the connection last move, here it gave a value
of 50 points to J7. This is of course much too high since white can still live.
]
LB[ge:A];W[ha]WL[1800]
C[GNU Go saves the situation on the upper edge.  Note that the lower right corner can
connect to the upper side at H5 (A) if attacked at e.g. E3 or F2.
]
LB[he:A];B[ic]BL[1800]C[A bug?
];W[id]WL[1800];B[ga]BL[1800];W[ag]
WL[1800]
C[GNU Go gives a value of 8 to A3 and 7 to E3. I would say that E3 is bigger.
]
;B[ie]BL[1800];W[ad]WL[1800]
C[GNU Go thinks that A6 kills A8. But A8 is already dead. The really important move
here is H5 which keeps the connection. GNU Go sees that move but only values
it to 12 points against 16 for A6.
]
;B[eh]BL[1800]
C[This attack is bad since H5 still exists.  However GoLife is starting to get so much
behind that desperate means are needed. The only way for GoLife to win is to
kill the lower right, but this will have to be due to severe life and death
shortcomings in GNU Go.
]
;W[he]WL[1800]C[And there goes the last chance.
];B[gh]BL[1800];W[hh]
WL[1800];B[fh]BL[1800];W[gd]WL[1800];B[ce]BL[1800];W[fe]WL[1800];B[ee]
BL[1800];W[gf]WL[1800]
C[Of course, F4 is much better since it takes one more points, and leaves black with no
followup threat. It doesn't actually threaten to kill black (there is still B2), but there
is plenty of room for mistakes. G4 got a ridiculous value of 15 points from GNU Go,
and F4 the equally ridiculous 11 points.
]
;B[ff]BL[1800];W[bh]WL[1800]C[Good!
];B[dh]BL[1800]
C[And this move lives. It looks as if white could kill, but the A3 stones are not thightly
enough connected that it would work. However, if there was a stone at A4, black would
be dead (see variation).
]
(;W[fa]WL[1800]C[A wasted move. -1 point in gote.
];B[fg]BL[1800];
W[ge]WL[1800];B[df]BL[1800]
C[Unnecessary. Since black has one eye at the edge, black could play some sente endgame
moves, e.g. at H1.
]
;W[ch]WL[1800];B[bg]BL[1800];W[af]WL[1800];B[di]BL[1800];W[hi]WL[1800]
;B[fi]BL[1800]C[This move is a complete mystery to me.
];W[ec]WL[1800]
;B[bi]BL[1800];W[ah]WL[1800];B[ci]BL[1800];W[be]WL[1800];B[bf]BL[1800]
;W[dc]WL[1800];B[ai]BL[1800];W[ae]WL[1800];B[gi]BL[1800];W[tt]WL[1800]
C[This is were the game should have ended. But GoLife played 11 more moves at which
GNU Go passed at every one. At the 12th move GNU Go was forced to answer. So the
difference increased 11 points which could very well have proved fatal in a close game.
]
;B[cb]BL[1800];W[tt]WL[1800];B[ba]BL[1800];W[tt]WL[1800];B[da]BL[1800]
;W[tt]WL[1800];B[if]BL[1800];W[tt]WL[1800];B[hg]BL[1800];W[tt]WL[1800]
;B[ig]BL[1800];W[tt]WL[1800];B[eb]BL[1800];W[tt]WL[1800];B[ea]BL[1800]
;W[tt]WL[1800];B[ca]BL[1800];W[tt]WL[1800];B[ga]BL[1800];W[tt]WL[1800]
;B[gb]BL[1800];W[tt]WL[1800];B[ih]BL[1800];W[ii]WL[1800]
TW[aa][ba][ca][da][ea][ga][ia][ab][bb][cb][eb][gb][hb][bc][ic][hd][ie][if][hg][ig][ih]
TB[de][ef][dg][eg][ei]
C[The game is over.  Final score:
   White = 21 territory + 23 captures + 5.5 komi = 49.5
   Black = 5 territory + 0 captures = 5
White wins by 44.5.
])

(;W[fg];B[ef];W[di];B[ch];W[ci];B[ei];W[gi];B[bi];W[ci];B[bg]
(;W[ah];B[af])

(;W[ai];B[af];W[ae];B[ah]
C[A big ko, but white has no ko threats and black can play B4 whenever he wants.
])
)

)