233 lines
6.2 KiB
Scilab
233 lines
6.2 KiB
Scilab
# CATEGORY=FUSEKI_STRATEGY
|
|
# SEVERITY=5
|
|
loadsgf games/arend/gnugo-gnugo1.sgf 25
|
|
1 reg_genmove black
|
|
#? [C4]*
|
|
|
|
# CATEGORY=KO
|
|
# SEVERITY=3
|
|
loadsgf games/arend/gnugo-gnugo1.sgf 112
|
|
2 reg_genmove white
|
|
#? [H15]
|
|
|
|
# CATEGORY=CONNECTION
|
|
# SEVERITY=10
|
|
loadsgf games/arend/gnugo-gnugo1.sgf 139
|
|
3 reg_genmove black
|
|
#? [P6|Q5]
|
|
|
|
# These kind of fights might be very difficult to do well
|
|
# with current GNU Go methods (How should it choose between the different
|
|
# 'correct' alternatives here? - R14 seems clearly best to me).
|
|
# But certainly it should not tenuki.
|
|
loadsgf games/arend/exper1.sgf 7
|
|
4 reg_genmove black
|
|
#? [R14|S15|R17]
|
|
|
|
#CATEGORY=STRATEGY
|
|
#DESCRIPTION=GNU Go should escape with its cutting stone.
|
|
loadsgf games/arend/exper1.sgf 25
|
|
5 reg_genmove black
|
|
#? [D7|E7|E6]*
|
|
|
|
#CATEGORY=STRATEGY
|
|
#DESCRIPTION=In this position, GNU Go plays L10 because of an enormous assumed strategic effect on the M16 dragon, which is quite safe.
|
|
loadsgf games/arend/gnugo-gnugo2.sgf 57
|
|
7 reg_genmove black
|
|
#? [C3]
|
|
|
|
#CATEGORY=BLUNDER
|
|
# S 16 can be captured. S17 is better, but another big move
|
|
# should be accepted as well
|
|
loadsgf games/arend/gnugo-gnugo2.sgf 60
|
|
8 reg_genmove white
|
|
#? [!S16]
|
|
|
|
#CATEGORY=TACTICAL_READING
|
|
# GNU Go rejects S17 because it is "strategically or tactically unsafe".
|
|
# Must be a misread.
|
|
loadsgf games/arend/gnugo-gnugo2.sgf 61
|
|
9 reg_genmove black
|
|
#? [S17]
|
|
|
|
#CATEGORY=STRATEGY
|
|
# J7 is extremely overvalued. This is caused by pattern CB11.
|
|
loadsgf games/arend/gnugo-gnugo2.sgf 73
|
|
10 reg_genmove black
|
|
#? [!J7]
|
|
|
|
#CATEGORY=STRATEGY
|
|
# K6 is overvalued by GNU Go 3.0.0.
|
|
loadsgf games/arend/gnugo-gnugo2.sgf 91
|
|
11 reg_genmove black
|
|
#? [!K6]
|
|
|
|
#CATEGORY=BLUNDER
|
|
loadsgf games/arend/gnugo-gnugo2.sgf 163
|
|
12 reg_genmove black
|
|
#? [C16|B16]
|
|
|
|
#CATEGORY=STRATEGY
|
|
# GNU Go does not see that P17 helps his group as well, and plays O13.
|
|
loadsgf games/arend/gnugo-gnugo4.sgf 27
|
|
14 reg_genmove black
|
|
#? [P17]*
|
|
|
|
#CATEGORY=FUSEKI_STRATEGY
|
|
# C6 is clearly the wrong side for approaching LL corner.
|
|
loadsgf games/arend/gnugo-gnugo4.sgf 31
|
|
15 gg_genmove black 1
|
|
#? [F3]*
|
|
|
|
|
|
# GNU Go loves this move! (H2 is the point to play here.) It is impossible
|
|
# to teach GNU Go not to play there with minor pattern tuning. Problem is
|
|
# huge follow_up value; there is no check at all in the move valuation
|
|
# whether a sente move is actually worth something if opponent answers
|
|
# (here it is aji-keshi).
|
|
# This looses only a few points, but comes up very often.
|
|
loadsgf games/arend/gnugo-gnugo4.sgf 45
|
|
16 reg_genmove black
|
|
#? [!G3]
|
|
|
|
#CATEGORY=TACTICAL_READING
|
|
# GNU Go moves at C9. I do not know what it is afraid of.
|
|
loadsgf games/arend/gnugo-gnugo4.sgf 153
|
|
17 reg_genmove black
|
|
#? [C7]
|
|
|
|
# Maybe this problem is suboptimal, because there is a ko involved, and
|
|
# GNU Go gets it wrong even without the ko.
|
|
loadsgf games/arend/gnugo-gnugo4.sgf 168
|
|
18 reg_genmove white
|
|
#? [P18|P19|R18|R19|S18|S17|S16|S14|S13|T14|T16]
|
|
|
|
#CATEGORY=PATTERN_TUNING
|
|
# GNU Go often plays this inefficient shape. Should be correctable with an
|
|
# additional pattern. Core of the problem is however that GNU Go does not
|
|
# understand the essence of the iken tobi -- it can be cut, but Black does
|
|
# not mind that, since a cut only strengthens Black.
|
|
loadsgf games/arend/gnugo-gnugo5.sgf 39
|
|
20 reg_genmove black
|
|
#? [!L15]
|
|
|
|
#CATEGORY=STRATEGY
|
|
# This inefficient move is produced by owl. But of course it should not
|
|
# be owl in charge of strengthening this big dragon.
|
|
loadsgf games/arend/gnugo-gnugo5.sgf 78
|
|
21 reg_genmove white
|
|
#? [!F9]
|
|
|
|
# Immediately take the double sente point!
|
|
loadsgf games/arend/gnugo-gnugo5.sgf 80
|
|
22 reg_genmove white
|
|
#? [D14]*
|
|
|
|
# H10 is quite overvalued here.
|
|
loadsgf games/arend/gnugo-gnugo5.sgf 105
|
|
23 reg_genmove black
|
|
#? [!H10]
|
|
|
|
# Low Shimari clearly better here
|
|
# See also arend2:130
|
|
loadsgf games/arend/gnugo-gnugo6.sgf 20
|
|
24 restricted_genmove white P16 P17
|
|
#? [P17]*
|
|
|
|
#CATEGORY=ENDGAME_TUNING
|
|
# C19 is either double sente, or worth at least 5 pts in reverse sente,
|
|
# which is 10 pts. T14 (chosen by GNU Go) is dame. F7 is nearly as large
|
|
# as C19.
|
|
loadsgf games/arend/gnugo-gnugo6.sgf 174
|
|
25 reg_genmove white
|
|
#? [C19|F7]
|
|
|
|
#CATEGORY=FUSEKI_STRATEGY
|
|
# D17 is wrong direction; in this position this is really bad.
|
|
loadsgf games/arend/gnugo-gnugo7.sgf 17
|
|
28 reg_genmove black
|
|
#? [C16]
|
|
|
|
#CATEGORY=FUSEKI_STRATEGY
|
|
# Black C15 is sente here, so W C14 is mandatory.
|
|
loadsgf games/arend/gnugo-gnugo7.sgf 24
|
|
29 reg_genmove white
|
|
#? [B14|C14]
|
|
|
|
#CATEGORY=BLUNDER
|
|
# GNU Go played F8.
|
|
loadsgf games/arend/gnugo-gnugo7.sgf 79
|
|
30 reg_genmove black
|
|
#? [C13]*
|
|
|
|
#CATEGORY=OWL_DEFEND (?)
|
|
# B14 can (and should) be saved here
|
|
loadsgf games/arend/gnugo-gnugo7.sgf 81
|
|
31 reg_genmove black
|
|
#? [B13|C13]
|
|
|
|
#CATEGORY=CONNECTION
|
|
#SEVERITY=9
|
|
loadsgf games/arend/gnugo-gnugo7.sgf 137
|
|
32 reg_genmove black
|
|
#? [F9|E8|G8]
|
|
|
|
loadsgf games/arend/gnugo-gnugo7.sgf 140
|
|
33 reg_genmove white
|
|
#? [F9]*
|
|
|
|
#CATEGORY=ENDGAME_TUNING
|
|
loadsgf games/arend/gnugo-gnugo7.sgf 173
|
|
34 reg_genmove black
|
|
#? [!K8]
|
|
|
|
# This is a serious mistake. H17 is not huge, its HUGE.
|
|
loadsgf games/arend/gnugo-gnugo8.sgf 32
|
|
35 reg_genmove white
|
|
#? [H17|J19]
|
|
|
|
#CATEGORY=ENDGAME_TUNING
|
|
# Its just one point difference, but it happens a couple of times every
|
|
# game.
|
|
loadsgf games/arend/constructed1.sgf
|
|
36 reg_genmove black
|
|
#? [C5]
|
|
|
|
#CATEGORY=ENDGAME_TUNING
|
|
loadsgf games/arend/constructed2.sgf
|
|
37 reg_genmove black
|
|
#? [A5]
|
|
|
|
#CATEGORY=STRATEGY
|
|
# I think this problem summarizes well two common misjudgements
|
|
# of the influence function:
|
|
# 1. Breaking into an opponents moyo is enormously
|
|
# undervalued (that is why patterns like LE14--safe jump towards
|
|
# oppoents moyo--need to have a fixed minimum value, which often leads
|
|
# bad moves)
|
|
# 2. O5 considerably weakens the influence of N5. Maybe this is not weighted
|
|
# highly enough by GNU Go?
|
|
loadsgf games/arend/constructed3.sgf
|
|
38 reg_genmove white
|
|
#? [O5]
|
|
|
|
#CATEGORY=STRATEGY
|
|
# This is an example for part 1 of my comment of the previous problem.
|
|
# This slightly adapted from a game (to make the problem more specific).
|
|
loadsgf games/arend/constructed4.sgf
|
|
39 reg_genmove black
|
|
#? [G10|H10|F13]
|
|
|
|
#CATEGORY=ENDGAME_TUNING
|
|
# EB714 is necessary here; without it, GNU Go played B6, only for the
|
|
# reason of filling a liberty
|
|
loadsgf games/arend/constructed5.sgf
|
|
40 reg_genmove white
|
|
#? [B7]
|
|
|
|
#CATEGORY=ENDGAME_TUNING
|
|
# EB715 necessary
|
|
loadsgf games/arend/constructed6.sgf
|
|
41 reg_genmove white
|
|
#? [B6]
|