gpsshogi — GPSShogi for XShogi and the CSA protocol


Search options:

-i [ --initial-limit ] arg (=400)
initial limit of iterative deepening
--deepening-step arg (=200)
step of iterative deepening
-l [ --search-limit ] arg (=1400)
maximum limit of search
-L [ --record-limit ] arg (=200)
minimum depth left of storing record
-n [ --node-limit ] arg (=16000000)
maximum number of search nodes (tested at each iteration)
--node-count-hard-limit arg (=18446744073709551615)
maximum number of search nodes (tested at every nodes)
-t [ --table-size ] arg (=4800000)
number of entries in table
--check-table-size arg (=-1)
ignored. just for backward compatibility.
--ignore-node-counts arg (=1)
ignore {node,checkmate} counts and table size and continue search until memory exhausts
--draw-coef arg (=-1)
preference for draw (prefer if positive, escape if negative), relative to pawn value
-N [ --num-cpus ] arg (=4)
num cpus for parallel search
--memory-use-percent arg (=100)
percentage for memory use (normally 100)

Evaluation options:

-e [ --eval-type ] arg (=test)
evaluation function (test or progress
--eval-data arg
filename for evaluation function
--progress-file arg
filename for NewProgress data
--eval-randomness arg (=0)
add random value generated by normal distribution of a given standard deviation, to evaluation values

Match options:

-b [ --black ] arg (=sente)
name of black player
-w [ --white ] arg (=gote)
name of white player
-T [ --time-left ] arg (=1500)
total time left
-B [ --byoyomi ] arg (=0)
seconds for each move (byoyomi)
-o [ --output-filename ] arg (=gpsshogi.csa)
filename of game record
-f [ --initial-position ] arg
csa file of initial position
-R [ --no-table-reuse ]
no effect (preserved only for backward compatibility)
--save-pv arg (=1)
use pv of previous search results
-c [ --csa ]
use csa protocol, instead of gnushogi compatible mode
-a [ --alphabeta ]
use alphabeta, instead of mtd(f)
-r [ --random ]
random playing
-s [ --sente ]
computer play first (second if not specified)
-P [ --opponent-time ]
search while the opponent is thinking
-x [ --move-with-comment ]
send move with comment (this is a private extension to CSA protocol)

Opening options:

-K [ --kisen-file ] arg (=/usr/share/gpsshogi-data/data/kisen/01.kif)
kisen file for opening book (unused unless kisen-id is specified)
-k [ --kisen-id ] arg (=-1)
id of kisen record for opening book (enabled if positive)
--csa-file arg
csa file for opening book
--opening-file arg
file for opening book (used for normal games)
do not use opening book
-m [ --book-moves ] arg (=30)
maximum depth traced in book
--enable-determinate-book arg (=0)
if a specified value (N) > 0, a move is selected among the top N weighted moves with a probability propotional to each move's weight (default 0 [disabled])
--weight-initial-coef arg (=16)
weight coefficient to the top moves, which is used to filter moves only for the initial one.
--weight-coef arg (=10)
weight coefficient to the top moves, which is used to filter moves except the initial one.

Other options:

test search performance
test search performance
--benchmark-more arg (=0)
number of problems for benchmark
--sleep-at-exit arg (=0)
seconds of sleeping at exit
test whether data files are properly placed
-h [ --help ]
produce help message
-v [ --verbose ]
produce verbose output
show version info