def policy_moves_cmd(self, args):
moves = []
probs = []
if self.go_engine.random_simulation:
color = self.board.current_player
empty_pts = self.board.get_empty_points()
legal_pts = [
move for move in empty_pts if self.board.is_legal(move, color)
]
probs = [
str(round(1 / len(legal_pts), 3))
for _ in range(len(legal_pts))
]
for mv in legal_pts:
coord = point_to_coord(mv, self.board.size)
moves.append(format_point(coord).lower())
mv_str = ' '.join(sorted(moves))
prob_str = ' '.join(probs)
self.respond(mv_str + " " + prob_str)
else:
m = PatternUtil.generate_pattern_moves(self.board,
self.pattern_table)
mvs, vals = PatternUtil.calc_probabilities(m)
for mv in mvs:
coord = point_to_coord(mv, self.board.size)
moves.append(format_point(coord).lower())
sorted_mvs, sorted_vals = (list(i)
for i in zip(*sorted(zip(moves, vals))))
probs = [str(round(num, 3)) for num in sorted_vals]
mvs_str = ' '.join(sorted_mvs)
sorted_vals = ' '.join(probs)
self.respond(mvs_str + " " + sorted_vals)
def policy_moves_cmd(self, args):
# if the game is already end
if (self.board.check_game_end_gomoku()[0]):
self.respond()
return
# check if still have the empty points
if len(GoBoardUtil.generate_legal_moves_gomoku(self.board)) == 0:
self.respond()
return
# use copy board
cboard = self.board.copy()
# check policy
if self.policy_type == "random":
moves = GoBoardUtil.generate_legal_moves_gomoku(self.board)
moves_string = []
for move in moves:
move_coord = point_to_coord(move, self.board.size)
move_as_string = format_point(move_coord)
moves_string.append(move_as_string)
sorted_moves = ' '.join(sorted(moves_string))
self.respond("Random {}".format(sorted_moves))
else:
policy, moves = PatternUtil.generate_policy_moves(cboard)
moves_string = []
for move in moves:
move_coord = point_to_coord(move, self.board.size)
move_as_string = format_point(move_coord)
moves_string.append(move_as_string)
sorted_moves = ' '.join(sorted(moves_string))
self.respond(policy + " " + sorted_moves)
def simulate(self, board, move, to_play):
"""
Run a simulate game for a given move.
"""
cboard = board.copy()
cboard.play_move_gomoku(move, to_play)
opp = GoBoardUtil.opponent(to_play)
return PatternUtil.play_game(cboard,
opp,
random_simulation=self.random_simulation)
def simulate(self, board, move, toplay, table):
cboard = board.copy()
cboard.play_move(move, toplay)
opp = GoBoardUtil.opponent(toplay)
return PatternUtil.playGame(cboard,
opp,
limit=self.limit,
random_simulation=self.random_simulation,
use_pattern=self.use_pattern,
table=table)
def policy_moves_cmd(self, args):
"""
Return list of policy moves for the current_player of the board
"""
policy_moves, type_of_move = PatternUtil.generate_all_policy_moves(
self.board, self.go_engine.use_pattern,
self.go_engine.check_selfatari)
if len(policy_moves) == 0:
self.respond("Pass")
else:
response = type_of_move + " " + sorted_point_string(
policy_moves, self.board.size)
self.respond(response)
def simulate(self, board, move, toplay):
"""
Run a simulate game for a given move.
"""
cboard = board.copy()
cboard.play_move(move, toplay)
opp = GoBoardUtil.opponent(toplay)
return PatternUtil.playGame(cboard,
opp,
komi=self.komi,
limit=self.limit,
random_simulation=self.random_simulation,
use_pattern=self.use_pattern)
def simulate(board, move, toplay):
"""
Run a simulate game for a given move.
"""
cboard = board.copy()
cboard.play_move(move, toplay)
opp = GoBoardUtil.opponent(toplay)
return PatternUtil.playGame(cboard,
opp,
komi=0,
limit=100,
random_simulation = True, #implement a way to change this accordingly
use_pattern = False, #implement a way to change this accordingly
check_selfatari = False) #implement a way to change this accordingly
def policy_moves_cmd(self, args):
if self.go_engine.random_simulation == True:
empties = self.board.get_empty_points()
color = self.board.current_player
legal_moves = []
for move in empties:
if self.board.is_legal(move, color):
legal_moves.append(move)
gtp_moves = []
probabilities = []
for move in legal_moves:
coords = point_to_coord(move, self.board.size)
gtp_moves.append(format_point(coords).lower())
probabilities.append(str(round(1 / len(legal_moves), 3)))
sorted_moves = ' '.join(sorted(gtp_moves))
probability = ' '.join(probabilities)
self.respond(sorted_moves + " " + probability)
else:
moves, values = PatternUtil.generate_pattern_moves(self.board)
new_values = PatternUtil.normalize_val(values)
dict = {}
#sorted_moves = sorted(moves)
new_sorted_values = []
format_moves = []
for move in moves:
#new_sorted_values.append(str(round(dict[move], 3)))
coords = point_to_coord(move, self.board.size)
format_moves.append(format_point(coords).lower())
for i in range(len(format_moves)):
dict[format_moves[i]] = new_values[i]
sorted_format_moves = sorted(format_moves)
for move in sorted_format_moves:
new_sorted_values.append(str(round(dict[move], 3)))
sorted_moves_str = ' '.join(sorted_format_moves)
probability = ' '.join(new_sorted_values)
self.respond(sorted_moves_str + " " + probability)
def _evaluate_rollout(self, board, toplay):
"""
Use the rollout policy to play until the end of the game, returning +1 if the current
player wins, -1 if the opponent wins, and 0 if it is a tie.
"""
winner = PatternUtil.playGame(board,
toplay,
komi=self.komi,
limit=self.limit,
random_simulation=self.simulation_policy,
use_pattern=self.use_pattern,
check_selfatari=self.check_selfatari)
if winner == BLACK:
return 1
else:
return 0
def policy_moves_cmd(self, args):
if self.go_engine.random_simulation:
#get legal moves
first_moves = GoBoardUtil.generate_legal_moves(
self.board, self.board.current_player)
moves = []
#convert points to formated
for move in first_moves:
#convert to coords
temp = point_to_coord(move, self.board.size)
#format version
temp = format_point(temp)
moves.append(temp)
moves.sort()
#reverse it
new_moves = []
column_letters = "ABCDEFGHJKLMNOPQRSTUVWXYZ"
#put format to point
for move in moves:
temp = str(column_letters.find(move[0]) + 1)
temp = coord_to_point(int(move[1]), int(temp), self.board.size)
new_moves.append((temp))
movelist = new_moves
p = []
for x in range(0, len(movelist)):
p.append(round(float(1) / float(len(movelist)), 3))
else:
movelist, p = PatternUtil.pattern(self.board)
result = ''
for move in movelist:
temp = point_to_coord(move, self.board.size)
temp = format_point(temp)
result += temp.lower() + " "
for i in p:
result += str(i) + " "
try:
if args[0] == "1":
return movelist, p
except:
self.respond(result)
def get_move(self, board, color):
"""
Run one-player MC simulations to get a move to play
"""
color = board.current_player
cboard = board.copy()
if self.random_simulation:
moves = GoBoardUtil.generate_legal_moves_gomoku(board)
else:
_, moves = PatternUtil.generate_policy_moves(board)
move_wins = []
for move in moves:
wins = self.simulate_move(cboard, move, color)
move_wins.append(wins)
return select_best_move(board, moves, move_wins)
def find_pattern_feature(features, board, point):
p = PatternUtil.neighborhood_33(board, point)
if p in pat3set:
Feature.set_feature(features, point,
patIndex[p] + NUM_SIMPLE_FEATURE, 1)
def get_move(self, board, toplay):
return PatternUtil.generate_move_with_filter(board, self.use_pattern,
self.check_selfatari)
def policy_moves_cmd(self, args):
signal.signal(signal.SIGALRM, self.handler) #added for timer
signal.alarm(int(self.timel) - 10) #added for timer
print("timer start")
try:
movelist = GoBoardUtil.generate_legal_moves(
self.board, self.board.current_player)
if self.go_engine.random_simulation:
"""#get legal moves
first_moves = GoBoardUtil.generate_legal_moves(self.board, self.board.current_player)
moves =[]
#convert points to formated
for move in first_moves:
#convert to coords
temp = point_to_coord(move, self.board.size)
#format version
temp = format_point(temp)
moves.append(temp)
moves.sort()
#reverse it
new_moves =[]
column_letters = "ABCDEFGHJKLMNOPQRSTUVWXYZ"
#put format to point
for move in moves:
temp = str(column_letters.find(move[0])+1)
temp = coord_to_point(int(move[1]),int(temp), self.board.size)
new_moves.append((temp))
movelist = new_moves"""
p = []
for x in range(0, len(movelist)):
p.append(round(float(1) / float(len(movelist)), 3))
else:
p = []
for move in movelist:
p.append(PatternUtil.pattern(self.board, move))
for y in range(0, len(p)):
p.append(round(float(p.pop()) / float(len(movelist)), 3))
try:
if args[0] == "1":
signal.alarm(0)
return movelist, p
except:
result = ''
for move in movelist:
temp = point_to_coord(move, self.board.size)
temp = format_point(temp)
result += temp.lower() + " "
for i in p:
result += str(i) + " "
signal.alarm(0)
self.respond(result)
except TimeoutException: #added for timer
#give best we have
if len(p) < len(movelist):
for x in range(0, (len(movelist) - len(p))):
p.append(1.0)
for y in range(0, len(p)):
p.append(round(float(p.pop()) / float(len(movelist)), 3))
signal.alarm(0)
return movelist, p
