Re: computer-go: new ladder member

[magnus.persson@xxxxxxxxxx]

Hi!

Viking has made a small step on the 9x9 computer go ladder as 
shown in the attached game. I also tried to add some comments 
and try to explain why (I think) Viking played as it did.

If the file is corrupt for some reason then the original on NNGS is:

viking-babel17-200101191445

Best wishes
Magnus

(;
GM[1]FF[4]VW[]AP[Many Faces of Go:10.0]
SZ[9]
HA[0]
ST[0]
PB[babel17]
PW[Viking]
DT[2001-01-19]
PC[NNGS Worldwide: nngs.cosmic.org 9696]
KM[5.5]
TM[720]
RE[W+44.5]
US[Brought to you by No Name Go Server]
EV[None]
GN[Viking-babel17(B) NNGS]
CP[
  This game was played on the No Name Go Server
  Permission to reproduce this game is given.]
BR[NR ]
WR[18k*]
;B[ee];W[fd];B[ec];W[ed];B[fe];W[dd];B[de];W[fc]
C[Black has more potential but white is a little stronger. Sometime
9x9 go matches against Viking are very peaceful - and in such
games Black might win easily.
]
;B[eb]
C[This move is heavy and Viking ignores these two stones to make
a base of secure territory with  a hane at G5 and threaten the
black half of the board.
];W[ge];B[cg]
C[The san- san is not a wasted move since it makes black secure
in any fighting but clearly G4 is the move here.
];W[gf];B[gg]
C[This is bad but Viking often has problems of handling this shape.
];W[fg];B[hg]
C[I would prefer Viking to simply play f4. But it makes "secure"
eyespace in the corner instead because Black H3 undermined the
white group.
];W[fb];B[ff];W[eg]
;B[eh]
C[This contact move makes the fight more hopeful for white.
];W[fh];B[dg];W[ce];B[df];W[cc]
C[This is a safe move and a move by black here could have been
painful. Capturing a stone would be good too. But Viking cannot
see that the white stone would live and plays elsewhere.
];B[dh];W[bf];B[bg];W[hf]
;B[cb]
C[White can now capture the lower right corner, but prefers to
attack the big black group first. It does not realize that the
black group will live big if the white stones are captured.
];W[ag];B[ah];W[af];B[bh];W[ig];B[ih]
C[Viking can now capture black with H2. But it also know that Black
is thratening to capture with J4. Viking do not realize that
H2 defends against J4. 

A problem with the current implementation is that even if Viking
can read several things correctly in a position  it cannot integrate
that information unless it makes a deep fullboard evaluation
until the position is quiet.
Which it don't. I just looks at the best black moves for one
ply and compares them with the best possible white moves on the
board and make a compromise.
];W[if];B[fi];W[gi]
;B[ef];W[ei];B[hi]
C[Here some kind of bug creeps up in the tactics. White fails to
see that a move at H2 will give black an Eye an immunity against
the white eyeless stones.
];W[di]
C[If black plays H2 now 
]
(;B[ci]
C[Viking does read snapbacks and now there is no choice.
];W[hh];B[gh];W[hh];B[fi];W[gh]
C[A small mistake. Again Viking defends against the greater threat
by playing the vital point of the opponent and not the correct
move.
];B[ei];W[bb]
C[Viking cannot read tactically that the upper left black group
is dead and the heuristc group evaluation thinks it just about
80% dead. So it attacks the remaining 20%.
];B[cf]
;W[bd];B[ae];W[be];B[ia];W[db]
C[D8 is worth .2 points to Viking. It could be set to stop 
playing moves like this earlier but then the endgame might not
be played to the end. This is a compromise assuming chinese rules.
];B[ib];W[hc];B[tt];W[tt];)
(;B[hh]
(;W[bi];B[ci];W[ch]
C[Viking do not know that kothreats are possible, and treat all
ko's as impossible to win. This is probaly what Viking read out
and why it believed this was a safe way to play.
])
(;W[ci];B[bi];W[cf]
C[This way it is an unusual seki.
])))