1 This is brief developer-oriented overview in Pachi structure.
3 Pachi is completely Go-specific (c.f. Fuego; though e.g. atari go support
4 should be easy to add), but fairly modular. It has been built with focus
5 on MonteCarlo-based play, but it can in principle be used for other
12 Pachi consists of the following components:
15 +------+ +--------+ +---------+ +------------------+
16 | core | -- | engine | -- | playout | -- | tactical library |
17 +------+ +--------+ +---------+ +------------------+
23 * "core" takes care of the program's lifetime, GTP interface and basic
24 fast Go board implementation
26 pachi.c global initialization and the main loop
27 version.h current version information
28 debug.h debugging infrastructure
29 random.[ch] fast random number generator
30 gtp.[ch] GTP protocol interface
31 network.[ch] Network interface (useful for distributed engine)
32 timeinfo.[ch] Time-keeping information
33 stone.[ch] one board point coloring definition
34 move.[ch] one board move definition
35 board.[ch] board definition and basic interface
37 * "aux library" provides extra functions like static tactical evaluation
38 and pattern matching; it is somewhat interwound with "core" component
40 mq.h "move queue" data structure
41 stats.h "move statistics" data structure
42 probdist.[ch] "probability distribution" data structure
43 ownermap.[ch] simulation-based finalpos. "owner map" data structure
44 pattern3.[ch] fast 3x3 spatial pattern matcher
45 pattern.[ch] general multi-feature pattern matcher
47 * "tactical library" provides extended interfaces for the go board,
48 most important non-trivial tactical information
52 * "engine" receives notifications about opponent moves and is asked
53 to generate a move to play on given board
55 engine.h abstract engine interface
56 random/ example "random move generator" engine
57 replay/ example "playout move generator" engine
58 montecarlo/ simple treeless Monte Carlo engine, quite bitrotten
59 uct/ the main UCT-player engine, see below
60 distributed/ "meta-engine" for distributed play by orchestrating
61 several UCT engines on different computers
62 patternscan/ auxiliary engine for harvesting patterns from
64 joseki/ auxiliary engine for generating joseki patterns
66 * "playout" policy is asked to generate moves to play during the Monte Carlo
67 simulations, and to provide rough evaluation of moves feasibility for
70 playout.[ch] abstract playout policy interface,
71 Monte Carlo simulation execution
72 playout/light uniformly random playout policy
73 playout/moggy rule-based "Mogo-like" playout policy
75 * Also, several ways of testing Pachi are provided:
77 t-unit/ interface for writing unit-tests for specific
78 functionality, mainly tactics
79 t-play/ interface for testing performance by playing games
80 against a fixed opponent (e.g. GNUGo)
81 t-predict/ test prediction rates of various components
87 The UCT engine (the proper name should be MCTS as it does not have that
88 much common with classic UCT now) has non-trivial structure by itself:
90 +-------------+ +-----+ +-------------------+
91 | node policy | -- | UCT | --- | node prior-hinter |
92 +-------------+ +-----+ +-------------------+
98 * "UCT" is the core of the engine
100 uct.[ch] engine initialization, public interface
101 internal.h internal state and data structures
102 tree.[ch] minimax move tree with success statistics
103 walk.[ch] filling the tree by walking it many times
104 and running MC simulations from leaves
105 slave.[ch] engine interface for the distributed engine
107 * "node prior-hinter" assigns newly created nodes preliminary success
108 statistics ("prior values") to focus the search better
110 prior.[ch] variety of methods for setting the priors
112 * "node policy" mainly chooses the current node's child to descend
113 through during the tree walk, based on the already recorded statistics;
114 it must balance exploration and exploitation well during the selection
116 policy/ucb1 the old-school original simple policy
117 policy/ucb1amaf the AMAF/RAVE-based policy gathering statistics rapidly
119 * "dynkomi driver" dynamically determines self-imposed extra virtual komi
127 The infrastructure is optimized for speed to make it well suited
128 for bruteforce engines, however tradeoffs are made to make it useful
129 for heavier MonteCarlo playouts as well (e.g. real liberties are
130 tracked instead of pseudoliberties). If you are looking for raw
131 light playout speed, libEGO is better choice.
133 In general, arbitrary board sizes are supported; however, board sizes
134 smaller than 9x9 have not been tested and board sizes larger than 25x25
135 are not supported by GTP - also, in theory some buffer overflows might
136 happen with board sizes larger than 19x19. The engine parameters are
137 primarily tuned for 19x19 play.
142 While the Pachi engines generally play according to Chinese rules,
143 internally, Pachi uses Tromp-Taylor rules because they are simple,
144 fast and universal; they are very close to the New Zealand rules.
145 That means, it simply counts the number of stones and one-point eyes
146 of each color on the board, plus komi and handicap correction.
148 Tromp-Taylor rules also mean that multi-stone suicide is allowed! If you
149 do not like that (basically if you want to pretend it plays according
150 to Chinese rules), you need to rule that out in your engine, currently.
151 The provided engines DO avoid multi-stone suicide, though it is allowed
152 in the playouts for performance reasons (perhaps we should re-visit that
153 decision in light of heavy playouts).
155 Tromp-Taylor rules have positional superko; the board implementation
156 will set a flag if it is violated, but play the move anyway. You need
157 to enforce the superko rule in your engine.
163 ...is a very sad hack. ENSURE that only trusted parties talk to Pachi's
164 GTP interface, as it is totally non-resilient to any kind of overflow
165 or bad input attacks and allowing arbitrary input to be entered within
166 is a major security hole. Yes, this needs to be cleaned up. Also, currently
167 engines cannot plug in their own commands and there is no GoGui interface.
169 Pachi supports only few GTP commands now. Most importantly, it does not
170 support the undo command. The final_status_list command requires engine
174 General Pattern Matcher
175 =======================
177 Pachi has in-development general pattern matcher that can match various
178 sets of features (spatial and others), inspired by the CrazyStone and
179 van der Werf - de Groot - Stern pattern model. Please see pattern.h
180 for detailed description of the pattern concept and recognized features.
182 To harvest patterns, use 'pachi -e patternscan' (see patternscan/patternscan.c
183 for available options). The output of the pattern scanner are two data
184 structures: The matched patterns
186 (feature1:payload feature2:payload ...)
188 and spatial dictionary. "Spatial" feature represents a particular
189 configuration of stones in a circle around the move-to-play; each
190 configuration has its own record in the dictionary and the spatial
191 feature references only the id in the dictionary; so you need to keep
192 both the patterns and the "patterns.spat" file. Normally, 'patternscan'
193 will only match already existing dictionary entries, but you
194 can pass it "gen_spat_dict" so that it appends all newly found spatial
195 features to the dictionary - use "spat_threshold" to limit recording
196 only to frequently occuring spatial features; to start the dictionary
197 from scratch, simply remove any existing "patterns.spat" file.
199 There are few pre-made scripts to make the initialization of the pattern
202 * tools/pattern_byplayer.sh: Sorts out patterns from given SGF collection
203 by player names, one player per file in a dedicated directory. This is
204 useful if you want to use the patterns to e.g. recognize games of a
205 player by characteristic patterns. Spatial dictionary is autogenerated
208 * tools/pattern_spatial_gen.sh: Initializes spatial dictionary by spatial
209 features found at least N times in given SGF collection. This is useful
210 for further gathering of general pattern statistics while keeping
211 the amount of spatial features manageable.
213 * tools/pattern_spatial_show.pl ID: Shows spatial pattern of given id
216 * tools/pattern_bayes_gen.sh: Generates a Bayesian probability table
217 for each pattern in a SGF collection. The computed number is the empiric
218 probability of playing a given pattern when it is available on board.
220 * tools/pattern_bayes_merge.sh: Merges Bayesian probability tables
221 of patterns. This is useful for parallel pattern mining - run multiple
222 pattern_bayes_gen (perhaps remove the counts >= 2 test) and then merge
223 the generated tables.
225 * tools/pattern_getdrops.pl: Processes game logs to determine moves that
226 lead to large drop in Pachi's evaluation and extracts patterns that
227 represent these moves. This might allow Pachi to "learn from its past
228 mistakes" and recognize these moves in the future. A full pipeline
231 cat kgsGtp.log | tools/kgslog2gtp.pl |
232 tools/pattern_getdrops.pl | tools/pattern_bayes_gen.sh - |
233 tools/pattern_bayes_merge.sh patterns.prob - >patterns.prob2
239 The UCT engine allows external plugins to be loaded and provide external
240 knowledge for various heuristics - currently just biasing the MCTS priors,
241 but more can be added easily. The plugins should follow the API in
242 <uct/plugin.h> - see that file for details.
248 Joseki patterns can be generated from variations of a SGF file. Josekis
249 are matched per-quadrant, no other move must be within the quadrant. See
250 joseki/README for details on usage of the auxiliary engine and a full
251 recipe for making Pachi play according to joseki sequences extracted from
258 The UCT engine can "pre-read" the starting board position and
259 dump the core of the built tree to a file, loading it later. This is
260 called a 'tbook' (as in "tree book") and can be generated using the
261 tools/gentbook.sh script. The newly generated file is automatically
262 used by the UCT engine when found.
264 Alternatively, there is a support for directly used opening book
265 (so-called fbook, a.k.a. "forced book" or "fuseki book"). The book
266 is stored in a text file in Fuego-compatible format and can be loaded
267 using the ./pachi -f parameter. A naive way to build such a book
268 based on shell-based, twogtp-based UCT is available through the
269 tools/autobook/ framework.
275 This is a mostly unique idea in Pachi, currently in development;
276 it does not work too well yet, but Pasky has great hopes for it in
279 A local tree is a tree comprising of (CFG-topologically) local
280 sequences, built in parallel with the real game tree; there is
281 a separate tree for black-first and white-first sequences. E.g if
282 the game tree (on empty board) consists of D4-C6-D6-Q16-O17-D7,
283 the coresponding local tree sequences are (black) D4-C6-D6, (white)
284 Q16-O17 and (white) D7. The simulation result is then recorded as
285 usual in the game tree, but also in the corresponding local trees.
287 The goal of this is to dynamically build a cache of various
288 resolutions of local situations. This is to overcome cases where
289 the proper resolution of a given situation is easily found in
290 the tree, but improper heuristical bisaes keep muddying the
293 Exactly what kind of values to store in the local tree nodes is
294 still an open question. The original approach has been to simply
295 use simulation results directly or biased on base "value temperature";
296 now, we are experimenting with survival rate of the "base stone"
297 of the branch (the first move in the branch sequence). We also intend
298 to integrate criticality information with local tree data.
300 The local trees are descended in parallel with the main tree.
301 The values stored in local trees are added to the RAVE term
302 from the main tree during node selection. We also wish to use
303 the information from local trees in the simulations, but this is
304 still subject to research.
306 To enable local trees usage, pass local_tree=1 to the UCT engine.
307 There are many further knobs to twiddle, too.