1 A list of games to put in the regression suite.
3 This list can be processed automatically:
5 sed -ne 's/^\* //p' <TODO | while read r c; do echo wget -O $(echo "$r" | sed 's#.*games/##' | tr / - | sed 's/-\([^-]\)-/-0\1-/g;s/-\([^-]\)-/-0\1-/') "$r"; done
7 * http://files.gokgs.com/games/2011/6/6/tyzef-pachi30s.sgf ladder
8 pachi tries to run away from a losing ladder against a 7k in even
9 game. The 7k, disgusted, passed then resigned.
11 * http://files.gokgs.com/games/2011/6/5/pachi2-SimpleBot.sgf ladder
12 pachi shouldn't play the broken ladders up to move 53
14 * http://files.gokgs.com/games/2011/6/5/Zen19-pachi2.sgf semeai falseeye
16 . move 185 S15 very dubious
18 . move 211 T18 pachi still thinks the semeai is undecided.
19 but white has a lot more liberties. How can the playouts
22 . move 227 S3: here are suggestions by Aja
23 If a false will become a real eye by capturing a diagonal group,
24 then capture the group instead of filling the false eye. So if
25 pachi wants to fill s3 it will capture at T1 instead. Then S3
26 will become an illegal move so pachi will evaluate this group
27 100% alive and won't play at T2 such stupid move.
29 * http://files.gokgs.com/games/2011/6/5/tazaki-pachi30s.sgf semeai falseeye
30 pachi doesn't see even one move ahead in a semeai, it expects
31 an answer outside the semeai. The score drops from 0.625 to 0.337:
34 [880000] best 0.624794 komi 0.4 | seq L19 G12 F11 P11 | can L19(0.625)
35 O19(0.623) N19(0.469) P11(0.523)
36 FINEST: Got successful response to command "genmove b": = L19
37 FINEST: Got successful response to command "play w j16": =
39 FINEST: Queued command sent to engine: genmove b
40 [1026644] best 0.336825 komi 0.2 | seq P11 T8 R3 M9 | can P11(0.337)
41 O11(0.336) O19(0.334) N19(0.340)
42 FINEST: Got successful response to command "genmove b": = P11
44 The game was lost already but pachi should have known this before.
46 move 244: T13 doesn't make sense. This is due to wrong estimate of
48 [1922010] best 0.492722 | seq T13 T4 D5 M6 | can T13(0.493)
49 S18(0.493) L4(0.493) E6(0.491)
50 FINEST: Got successful response to command "genmove b": = T13
52 At move 282 pachi still thinks white's group is undecided.
53 We must find a way of not filling the S7 eye in the playouts.
55 * http://files.gokgs.com/games/2011/5/29/snowmen-pachi2.sgf
56 169 B5 seems to be the losing move. This move doesn't make sense.
57 Pachi has been wrongly evaluating both bottom group around S4 (it's just
58 hairy for simulations) and top group around A19 - that one is a bit more
59 interesting, the problem is that black is double likely to play E19
60 (since it's C19 selfatari cousin), but after E19, white will not notice
61 the need to connect at C19. Not sure how to solve this yet, perhaps some
62 way to add cousins to last_move logic...
64 * http://files.gokgs.com/games/2011/5/28/pachi2-Rubel.sgf
65 won game lost: 212 H13 should be at J12.
67 * http://files.gokgs.com/games/2011/5/27/ManyFaces1-pachi2.sgf
68 had won game, move 135 F1 very bad. at end doesn't see ko at t7. (b t10
71 * http://files.gokgs.com/games/2011/5/23/pachi2-ManyFaces1.sgf
72 leading up to 90% then lost.
73 bad evaluation of 3 areas according to pasky.
75 * http://files.gokgs.com/games/2011/5/25/pachi2-Zen19S.sgf
76 lost but was leading according to ajahuang and Zen itself
78 * http://files.gokgs.com/games/2011/2/20/ramesses2-pachi2.sgf
79 good example of losing a won game.
81 * http://files.gokgs.com/games/2011/1/11/llopl-pachi2.sgf ladder
84 * http://files.gokgs.com/games/2010/12/5/pachi-CzechBot.sgf
85 big mistake at F9 didn't see snapback. Pasky comments:
86 The "does not seen snapback" bug is nasty. The problem is that
87 there is also a "global capture heuristic" if there're no fun
88 local moves, it looks for groups to capture and if white tenukis
89 and does not capture immediately, this heuristic kicks in. But
90 there is no "global atari heuristic" that would drive black
91 similarly to play at F1 and create the snapback if I test with
92 playout=moggy:capturerate=0, it reliably destroys the snapback
93 i.e. with capturerate=90, white is much more likely to come back
94 to the situation later even when playing tenuki first.
96 * http://files.gokgs.com/games/2011/1/15/pachi2-rollingon.sgf semeai
97 big semeai with easy counting of liberties misjuged
98 move 169 0.7322 -> move 171 0.2600
100 * http://files.gokgs.com/games/2011/2/18/pachi2-tf-3.sgf seki
101 lost because of unrecognized seki
103 * http://files.gokgs.com/games/2011/6/9/botkiller2-pachi30s.sgf falseeye
104 threw away a won game. pachi thought until the end that top left
105 white group was dead, so it made silly moves elsewhere.
106 The playouts must be prevented from filling the eye at A17.
108 * http://files.gokgs.com/games/2011/6/10/pachi30s-samba-2.sgf falseeye
109 exactly the same problem as botkiller2-pachi30s.sgf:
110 pachi considers the top left black group as unsettled
111 although it is clearly alive. So it makes silly moves
112 elsewhere. We have to fix this bug, it is too frequent.
114 * http://files.gokgs.com/games/2011/6/9/jinen-pachi30s-2.sgf ko
115 move 165 G5 again playing the ko threat before the ko.
116 move 187 R18 is not a ko threat
118 At this point pachi still considers the top right as
119 unsetled, although it is is clearly lost for B. So pachi's
120 estimate of winning by 61% is way too optimistic.
122 * http://files.gokgs.com/games/2011/6/14/pachi2-Regenbogen.sgf
123 pachi doesn't see that bent four in top left corner is dead.
124 Still thinks it is winning 79% at move 250 G4.
126 * http://files.gokgs.com/games/2011/6/14/pachi2-cdocdo-3.sgf semeai
127 pachi loses every semai by a large number of liberties
128 but still considers itself winning 57% up to move 222 G16.
130 * http://files.gokgs.com/games/2011/6/13/pachi30s-Jep.sgf ladder
131 moves 102..108 playing a broken ladder
133 * http://files.gokgs.com/games/2011/6/18/dorabon-pachi2-4.sgf semeai falseeye
134 pachi considers the bottom left semeai won by black (but black
135 loses it) and the bottom right as undecided (white is clearly
136 alive). At move 270 L4 it still thinks it is leading by 80%
137 although the game is lost.
138 The bottom right is the same problem as
139 http://files.gokgs.com/games/2011/6/10/pachi30s-samba-2.sgf
140 and http://files.gokgs.com/games/2011/6/9/botkiller2-pachi30s.sgf
141 This corner shape is very frequent and pachi always gets it wrong.
143 * http://files.gokgs.com/games/2011/6/18/pachi2-ultimakang-4.sgf
144 pachi was winning up to move 230 Q14 big mistake. Win ratio dropped
145 from 60% to 25% at next move.
147 * http://files.gokgs.com/games/2011/6/18/akkannbee-pachi2.sgf seki
148 pachi totally confused by very simple seki.
150 * http://files.gokgs.com/games/2011/7/6/xiaosugi-pachi2.sgf seki
151 pachi thinks until the end that it wins by 80% because it
152 doesn't understand the seki in lower right. It should have lost
153 by 42.5 points but white mistakenly agreed to treat the seki as dead.
155 * http://files.gokgs.com/games/2011/7/5/pachi2-azlan7-2.sgf
156 move 84 E10 & 86 F10 big mistakes, leave white group dead.
157 (amazingly pachi recovered and azlan7 escaped in the end).
159 * http://files.gokgs.com/games/2011/7/5/fafahhhh-pachi2.sgf
160 threw away won game, probably because of complex corner situations
161 won by time but game was lost on board
163 * http://files.gokgs.com/games/2011/7/3/pachi2-oasis7.sgf
164 Passed even though a dead opponent group was marked alive.
166 * http://files.gokgs.com/games/2011/7/28/pachi2-Novicer.sgf falseeye
167 Pachi makes many stupid moves because it thinks the top left
168 corner is unsettled although it is clearly alive. This shape
169 (4 in a row) is so simple that it should get it right.
172 W 264 E19 doesn't make sense, but still 69%
173 W 266 H19 doesn't make sense, but still 61%
174 W 268 T9 should be T8
175 W 292 A17 50%, W 294 R6 37% shows that top left was not
177 Also false eye evaluation problem in bottom left.
179 * http://files.gokgs.com/games/2011/7/28/pachi2-pkunzip-2.sgf ladder
180 W 132 C9, W 134 B8, W 136 A8 plays a broken ladder but win ratio 52%
181 so it should not play desperate moves.
182 Pachi thought it was winning 72% at W 198 M3 but lost the game.
184 * http://files.gokgs.com/games/2011/7/16/ihatebots-pachi2.sgf semeai
185 pachi threw away the game because it didn't evaluate the
186 semeai in bottom left correctly.
188 * http://files.gokgs.com/games/2011/7/31/pachi2-Drogba987.sgf
189 B 28 H4 should be at F1 or G1 to save E2 & E3 and the game.
190 Pachi usually finds this so I'm surprised it didn't get it here.
192 * http://files.gokgs.com/games/2011/8/9/pachi2-BlueSpark.sgf ladder
193 move 150 R19: pachi extends a losing ladder
195 * http://files.gokgs.com/games/2011/8/8/Dallas-pachi2-2.sgf ladder
196 move 121 L19: pachi extends a losing ladder
198 * http://files.gokgs.com/games/2011/8/9/somrak-pachi2-2.sgf ladder
199 move 143 N19: pachi extends a losing ladder
200 This ladder bug seems more frequent since e9df2b6 (even though
201 this change made Pachi 40 elo stronger).
203 * http://files.gokgs.com/games/2011/8/10/jerryfish-pachi2.sgf semeai
204 lost even game against 3k. Win ratio drops from 61% at 117 N1
205 to 49% at the next move. So pachi didn't evaluate the semai correctly.
207 * http://files.gokgs.com/games/2011/8/12/xiaosugi-pachi2.sgf falseeye
208 Pachi hallucinated again on bottom right corner.
209 This is the same problem as
210 http://files.gokgs.com/games/2011/6/18/dorabon-pachi2-4.sgf
211 http://files.gokgs.com/games/2011/6/10/pachi30s-samba-2.sgf
212 http://files.gokgs.com/games/2011/6/9/botkiller2-pachi30s.sgf
213 This corner shape is very frequent and pachi always gets it wrong.
215 * http://files.gokgs.com/games/2011/8/24/StoneGrid-pachi2.sgf ladder
216 Move 51 onwards: pachi plays a broken ladder, which almost cost the game.
217 This might be the same bug as
218 http://files.gokgs.com/games/2011/8/9/somrak-pachi2-2.sgf
219 http://files.gokgs.com/games/2011/8/9/pachi2-BlueSpark.sgf
220 http://files.gokgs.com/games/2011/8/8/Dallas-pachi2-2.sgf
221 http://files.gokgs.com/games/2011/7/28/pachi2-pkunzip-2.sgf
222 http://files.gokgs.com/games/2011/6/13/pachi30s-Jep.sgf
223 http://files.gokgs.com/games/2011/6/6/tyzef-pachi30s.sgf
224 http://files.gokgs.com/games/2011/6/5/pachi2-SimpleBot.sgf
225 http://files.gokgs.com/games/2011/1/11/llopl-pachi2.sgf
227 * http://files.gokgs.com/games/2011/9/4/pachi2-stv.sgf ladder
228 Move 48 onwards: pachi plays a broken ladder, which cost not
229 just the game but also the first place in the tournament.
230 See also broken ladders in
231 http://files.gokgs.com/games/2011/8/24/StoneGrid-pachi2.sgf
232 http://files.gokgs.com/games/2011/8/9/somrak-pachi2-2.sgf
233 http://files.gokgs.com/games/2011/8/9/pachi2-BlueSpark.sgf
234 http://files.gokgs.com/games/2011/8/8/Dallas-pachi2-2.sgf
235 http://files.gokgs.com/games/2011/7/28/pachi2-pkunzip-2.sgf
236 http://files.gokgs.com/games/2011/6/13/pachi30s-Jep.sgf
237 http://files.gokgs.com/games/2011/6/6/tyzef-pachi30s.sgf
238 http://files.gokgs.com/games/2011/6/5/pachi2-SimpleBot.sgf
239 http://files.gokgs.com/games/2011/1/11/llopl-pachi2.sgf