def isInitialState(self):
"""Returns True if this is the first play of the game, i.e., all
playedCards entries are empty.
:returns: True if it's the first play of the game, False otherwise.
"""
return all(cards.empty(played) for played in self.playedCards)
def simulation(self, node):
# if agent's hand is empty return score
if cards.empty(node.hands[self.idx]):
return node.score / node.visits
else:
count = 0
for i in range(node.numPlayers):
if cards.empty(node.hands[i]):
count += 1
if count == node.numPlayers - 1:
return (node.numPlayers + 1)**-1
agents = [dummyAgent.DummyAgent for i in xrange(node.numPlayers)]
gm = game.Game(agents, node.hands, node.playedCards, node.whosTurn,
node.topCard, node.lastPlayed, node.finished)
results = gm.playGame()
return ((results.index(self.idx) + 1)**-1) / node.visits
def __init__(self,
playedCards,
whosTurn,
hands,
idx,
lastMove,
depth,
topCard=None,
lastPlayed=None,
finished=[],
parent=None,
score=0.):
self.lastMove = lastMove
self.depth = depth
self.visits = 1.
self.score = score
self.children = []
self.parent = parent
self.hands = hands
self.idx = idx
self.terminal = cards.empty(hands[self.idx])
self.turn = (self.idx == whosTurn)
super(mctsNode, self).__init__(playedCards, whosTurn, topCard,
lastPlayed, finished)
def simulation(self, node):
# if agent's hand is empty return score
if cards.empty(node.hands[self.idx]):
return node.score / node.visits
else:
count = 0
for i in range(node.numPlayers):
if cards.empty(node.hands[i]):
count += 1
if count == node.numPlayers - 1:
return (node.numPlayers + 1) ** -1
agents = [dummyAgent.DummyAgent for i in xrange(node.numPlayers)]
gm = game.Game(agents, node.hands, node.playedCards, node.whosTurn,
node.topCard, node.lastPlayed, node.finished)
results = gm.playGame()
return ((results.index(self.idx) + 1) ** -1) / node.visits
def isInitialState(self):
"""Returns True if this is the first play of the game, i.e., all
playedCards entries are empty.
:returns: True if it's the first play of the game, False otherwise.
"""
return all(cards.empty(played) for played in self.playedCards)
def __init__(self, playedCards, whosTurn, hands, idx, lastMove, depth,
topCard=None, lastPlayed=None, finished=[], parent = None, score=0.):
self.lastMove = lastMove
self.depth = depth
self.visits = 1.
self.score = score
self.children = []
self.parent = parent
self.hands = hands
self.idx = idx
self.terminal = cards.empty(hands[self.idx])
self.turn = (self.idx == whosTurn)
super(mctsNode, self).__init__(playedCards, whosTurn,
topCard, lastPlayed, finished)
def selection(self, root):
numDone = len(root.finished)
if root.children == []:
# player to play has no cards, but game isn't finished
if (not root.turn) and cards.empty(root.hands[root.whosTurn]) and root.numPlayers > numDone:
root.addChild(agent.PASS)
return self.selection(root.children[0])
elif (root.turn and root.terminal) or root.isFinalState():
return root
else:
curr_state = state.State(root.playedCards, root.whosTurn,
root.topCard, root.lastPlayed, root.finished)
testagent = agent.Agent(root.whosTurn, root.hands[root.whosTurn])
actions = testagent.getAllActions(curr_state)
root.addAllChildren(actions)
selected_child = random.choice(root.children)
return selected_child
else:
return self.selection(self.bestChild(root.children))
def selection(self, root):
numDone = len(root.finished)
if root.children == []:
# player to play has no cards, but game isn't finished
if (not root.turn) and cards.empty(
root.hands[root.whosTurn]) and root.numPlayers > numDone:
root.addChild(agent.PASS)
return self.selection(root.children[0])
elif (root.turn and root.terminal) or root.isFinalState():
return root
else:
curr_state = state.State(root.playedCards, root.whosTurn,
root.topCard, root.lastPlayed,
root.finished)
testagent = agent.Agent(root.whosTurn,
root.hands[root.whosTurn])
actions = testagent.getAllActions(curr_state)
root.addAllChildren(actions)
selected_child = random.choice(root.children)
return selected_child
else:
return self.selection(self.bestChild(root.children))