1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
|
Phalanx is a chess playing program
Copyright (c) 1997, 1998, 1999, 2000, 2014 Dusan Dobes
LICENSE AND WARRANTY
- Phalanx is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2, or (at your option)
any later version.
- Phalanx is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
- You should have received a copy of the GNU General Public License
along with Phalanx; see the file COPYING. If not, write to
the Free Software Foundation, 59 Temple Place - Suite 330, Boston,
MA 02111-1307, USA.
WHERE TO GET PHALANX
- http://sourceforge.net/projects/phalanx/
- ftp://sunsite.unc.edu/pub/Linux/games/strategy/ (Only sources)
Please let me know if there are others, i will update this list.
Feel free to put Phalanx on your ftp site.
COMPILING, PORTABILITY
Compiling is simple, at least under Linux. Just type `make'. You might want
to edit the optimization flags in makefile to produce faster binary. Phalanx
is developed under GNU C, GNU Debugger, and GNU Make. If your system does
not support GNU exensions (e.g. long options), remove '-DGNUFUN' from DEFINES
in makefile. If you have incompatible 'make', try this:
$ cat *.c > allphalanx.c; cc -O allphalanx.c -o phalanx
INTERFACE, COMMAND LINE OPTIONS
Phalanx is xboard compatible. Running with xboard: 'xboard -fcp phalanx'.
Note that permanent brain (pondering) is off by default. Newer versions of
xboard set it on with the 'hard' command. If this does not work, try
'xboard -fcp "phalanx -p+"' or (for <4.0.0 versions of xboard) change your
initString (see Xboard documentation for details). It's better to stop
permanent brain in both programs, when playing Phalanx against another
program on a machine with one CPU.
It's also possible to run phalanx without xboard. Do "phalanx -h" to get
a list of command line options. One important command of phalanx's ASCII
interface is "help".
MORE ABOUT INTERFACE
I'm trying to write an interface that fits following three requirements:
- Xboard compatibility. For best results, get the latest version of Xboard.
- shell-like interface that allows running commands in a batch. It's very
useful for testing. Example: look into the file test.fin. It's a set
of chess problems and solutions. You can simply send this file to
Phalanx's stdin:
$ phalanx -c+ -o- -b- -f10 < test.fin | tee result
( Where: -c+ .... use cpu time
-o- .... don't use polling input
-b- .... no opening book
-f10 ... fixed time 10 seconds per move )
Watch how it works.
- Acceptable ASCII interface.
COMMAND LINE OPTIONS
Usage:
phalanx [options] [<moves> <minutes> [<increment in seconds>]]
phalanx [options] [<seconds per move>]
phalanx bench
phalanx bcreate [book creation options]
Options:
-t <transposition table size in kilobytes>
Transposition table really needs to fit in the RAM memory, it should
never be swapped to disk. Maximum size is your total RAM minus space
needed for OS (8-20 MB) minus space for other (possible) applications.
-f <fixed search time in seconds>
-x <+/-> xboard mode on/off default: on
"Xboard mode off" is designed for the ascii interface. In -s+ mode it
shows also the move that is currently searched.
-p <+/-> permanent brain on/off default: off
Phalanx ponders on a predicted move with the permanent brain on. It is
sometimes called pondering. Xboard sets it on with the 'hard' command
regardless of this option, just use xboard menu to set it on/off.
-s <+/-> show thinking on/off default: off
Phalanx shows depth, evaluation, time used in centiseconds, nodes
searched, and principal variation (best line) when searching. Xboard
overrides this option with 'post' and 'nopost' commands, just use
xboard menu instead of this.
-c <+/-> cpu time default: off
This one defaults to the "wall clock" real time. It's better to use
the CPU time when running test suites.
-o <+/-> polling input default: on
Use -o- for running test suites. Phalanx reads the positions from its
standard input and polling input makes it stop thinking after almost
zero seconds on each position.
-b <+/-> opening book default: on
-l <+/-> learning on/off default: off
-r <resign value in centipawns> default: 0 (no resigning)
-e <easy level 0...99> default: 0 (best play)
1 is the hardest and 99 is the easiest easy level. Phalanx tries to
emulate human-like blunders, the higher the number the more blunders it
plays. It also adds more randomness with the easy levels, repeating
games should be impossible. Easy levels set hashtable size to zero,
pondering and learning to off. Since version XXIII, Phalanx uses
the time normally, it does not respond immediatelly. NPS is lowered
to 100-300, unless overriden by the -n agrument. Root moves randomizing
is used here as well, between 10 to 60 centipawns, unless overriden
by the -z argument.
-z <random evaluation in centipawns> default: 0 (best play)
Randomize play. Add pseudo-random values to root moves evaluations.
The numeric value limits the random range, the interval of random
evaluations to be added is [-N/2 ... N/2].
-z <random evaluation in centipawns>:<N>
Like above, but only randomize first N moves. This is to avoid repeating
lines in opening, while keeping almost the same playing strength. For
example, -z 20:10 will randomize first 10 moves in the game by
20 centipawns.
-n <nodes per second> default: 0 (no limit)
Limits the speed to weaken the engine and to use less resources: The
speed of the machine does not matter here, it uses usleep() during
the search, so with low NPS it does not raise the machine load.
-v
Print version and exit.
-P <primary book directory>
-S <secondary book directory>
-L <learning file directory>
-g <log file name>
Book creation options:
-b <buffer size in cells>
Stage 1. One cell is 10 bytes. Twice the space is needed for sorting
the buffer, use no more than 1/3 of your total RAM.
-p <maxply>
Stage 1. Stop parsing the game at given ply (halfmove). Default value is
70, it's 35 full moves.
-c <max length of comment>
Stage 1. PGN files have comments in {}. Phalanx skips the comments when
parsing. Big PGN files also have lots of errors, sometimes '{' does not
have matching '}'. So, if the comment goes over the maximal length, we
suspect the text is no more the comment and continue parsing. Default
value is 4096 bytes.
-w <winning move value>
Stage 1. We are adding the [position,move] to the database. The side to
move won the game. The appropriate database entry gets <winning move
value> bonus. Default is 5 points. Use values in 1-10 range.
-d <drawing move value>
Stage 1. Same as the previous except that the game was drawn. Default
value is 2 points.
-l <losing move value>
Stage 1. Same as the previous except that the side to move lost the
game. Default value is 1 point.
-u <unknown result move value>
Stage 1. Same as the previous except that there is no result of the game
in the PGN file, sad and common case. Default value is 1 point.
-v <min value of move to add>
Stage 2. Moves that have less total points than this value are not
added into the sbook.phalanx. Bigger numbers create smaller books.
The smallest reasonable number is 10. Default is 15.
-g <min value percentage of best move to add others>
Stage 2. Use values in 50-100 range. The 2nd, 3rd, and other moves are
stripped if their value is less than G% of the value of the best move.
The G is decreased in positions where the 1st move value has high enough
value to give more variability in frequently played opening, e.g. we
want a bit more than just sicilian, maybe 1. ... e5 is also good.
50 gives lots of variability and maybe dubious moves, 100 gives almost
no variability and only the highest valued moves. Default is 80.
Note that the book creation does not work with DOS end-of-lines in the PGN
input, just use dos2unix filter if needed.
Examples: phalanx -c+ -s+ -o - -x- -f 60 -t4000
xboard -fcp "phalanx -r800"
xboard -fcp 'phalanx -e 99'
ASCII INTERFACE COMMANDS
about about Phalanx, show version, copyright info, and settings
bd display position (same as 'd')
bk show book moves (xboard) - shows ECO code/name and all moves
from both primary and secondary opening books. The last line
is the text string that is used to identify the position in
the pbook.phalanx. Primary book moves are shown along with
their probabilities. There is no such probability info for
the secondary book moves, as all secondary book moves have
equal probability.
book enable/disable opening book
both machine plays both
depth set search depth in plies (xboard). Search will be stopped
at given depth, no timing info is used.
fen display position in FEN
force user plays both (xboard)
go switch sides, start computing (xboard)
history show game moves in full notation
level N M I set level to N moves in M minutes, increment I seconds.
Phalanx needs its time info have updated from xboard via
the 'time' command to make this level work well.
level N set level to fixed time N seconds per move
new new game (xboard)
post show thinking (xboard)
remove take back last move, two plies (xboard)
nopost do not show thinking (xboard)
quit quit (same as 'exit' or end of file character)
score show static evaluation
time <N> remaining time is N/100 s (xboard)
undo undo last ply (xboard; same as 'u')
<FEN position> set test position, start search, show result
# comment
OPENING BOOK
From version VI, there are two book files - primary (pbook.phalanx), and
secondary (sbook.phalanx). A position is first searched in pbook.phalanx.
Only if it's not found there, sbook.phalanx is searched. You can specify
book directories via command line (-P, -S) or use environment variables
PHALANXPBOOKDIR and PHALANXSBOOKDIR. Otherwise Phalanx tries to find its
book files in current directory (./book.phalanx, ./sbook.phalanx) and finally
in compiled-in directory (/usr/local/lib/phalanx). You can change the
compiled-in directory in makefile.
- pbook.phalanx is 'hand'-written, text book. One line per position, sorted.
This time, it's bigger than really needed, because it was the only book
file till version V. The size will be smaller and the line format might
change to EPD+SAN in future.
- sbook.phalanx is binary book, generated from large PGN files. Six bytes
per move (4 hash key, 2 move). You can generate your own sbook.phalanx
with 'phalanx bcreate <options>', like this:
$ ./phalanx bcreate < manyGMgames.pgn
Book creation has two stages. First stage reads and parses PGN from
standard input and creates the 'rbook.phalanx' file. If rbook.phalanx
already exists, the first stage is skipped. Second stage filters positions
and moves from the rbook.phalanx to sbook.phalanx. First stage needs about
30 times more time than the second one. Look at 'book creation options' for
more details on creating your sbook.phalanx.
ECO DATABASE
From version XX, Phalanx has ECO database. Command 'bk' can also print ECO
code and opening/variation name. This works with xboard 4.0 via
"Help"/"Book" menu item. If you want to activate this feature, you must type
the 'bk' command (or click the "Help"/"Book" menu item in xboard) once per
session in the initial position. Then, the ECO index is created (that might
take few seconds on a slow machine). The ECO index is based on positions
rather than on moves so transpositions between openings can take place. The
eco.phalanx file is taken from Gnuchess distribution (originally eco.pgn, but
this is not real pgn).
INSIDE THE MACHINE
Phalanx uses (traditional) 10x12 board implementation. There are three
often used board implementations: "8x8" (GNU Chess), "bitboard" (Crafty),
and "10x12" (Nimzo, Phalanx). In short, "10x12" is easy to implement and
the code and basic data structures are small ( == fast on PC). The engine
uses many well known techniques: PVS (principal variation search),
transposition/killer table, static-eval cache, history killers, SEE (static
exchange evaluator), null move pruning, forward pruning, internal iterative
deepening, chess-specific extensions.
AUTHOR
Dusan Dobes
|
