Re: [gnugo-devel] CGF2004

[Gunnar Farneback]

Dan wrote:
> I've added the games from CGF2004 to the cvs in the directory
> regression/games/cgf2004. You must use -d when you run cvs
> update to get the new directory.

Some random comments about the CGF games:

Aya
---
Move 28 (S16)
Looks like GNU Go is completely tricked here. Can white even capture
the S18 stones? To me it looks more natural to sacrifice two stones
with R15, P17, P18 and then connect at O16.

Move 34 (S14)
S14 is unreasonable since black can just play P18 to capture the
corner. (The ladder doesn't work.) I don't see that white has anything
better than to sacrifice the R18 stones as before.

Move 118 (K19)
This must be hallucination. White can't make anything with those
stones after black cuts at G19.

Move 162 (M14)
The N4 connection is urgent and the only move.


Caren's Whisper
---------------
Move 96 (A6)
No territory to gain here. This move only helps black.

Move 98 (E9)
Connecting at B7 is quite big now and locally bigger than E9. Another
alternative is to stop the monkey jump on the top. Furthermore the N16
stones can still get away so capturing those is probably the biggest
move on the board.

Move 130 (J17)
This move has no positive effect whatsoever. (It was thought to be an
owl attack on H18.)

Move 154 (A4)
D11 is clearly bigger, although the double threat to either defend D13
or cut off C9 is subtle for GNU Go. A hint that there's a problem can
be found by the connection reading which correctly sees that E13 and
C9 can be disconnected by black D11. Unfortunately it's not clear how
to make use of such information robustly.

Move 172 (F7)
The enclosed white stones can't make eyes, nor win a semeai against
the surrounding stones. F7 is hopeless.

Move 202 (L10)
M10 is locally bigger than L10. M9 is locally even better.


Gargoyle
--------
Move 213 (P4)
White is thrashing but P4 is not a natural restraining move. E.g. Q3
would make more sense.


GoriMutyu
---------
The opponent was very weak and most of its moves didn't make sense.
It's an open question whether anything would have survived if the game
had been played to completion.


Goro
----
From GNU Go's opening one would almost think that the cosmic option
was enabled. White is allowed to easily invade the moyo and to
eventually start surrounding a black group inside the former moyo. GNU
Go wins at move 125 because Goro had to forfeit after crashing. At
that time GNU Go was in rather bad trouble after having had its weak
group completely surrounded. Had it died the game would have looked fairly
close. 


Katsunari
---------
Move 74 (G8)
It's urgent to attack the invading stone. C9 looks natural.

Move 108 (D18)
D18 is artless since the black D17 string can be captured with the
H19+F18 tesuji combination.

Move 184 (Q1)
Q2 is locally better than Q1 and R1.


Martha
------
Poor Martha lost all her stones but still thrashed on until move 410
and lost by 370.5 points (377 on the board).

Move 127 (J2)
It must be better to capture the H3 string in the ladder with K3.


ShikoSakugo
-----------
Move 33 (S2)
This doesn't look like an appropriate double hane.

Move 93 (M14)
M14 lets the L13 string get away. L14 captures it.

Move 137 (B4)
L2 is bigger than B4.

/Gunnar