def train(qTables, numGames, alpha, tryHard):
tryHardGrowth = (1 - tryHard) / numGames
for i in tqdm(range(numGames)):
board = tt.genBoard()
movesLeft = True
winner = False
player = 2
keysSoFar = []
movesSoFar = []
computersPlayer = random.randint(1, 2)
while (movesLeft and not winner):
if player == computersPlayer:
bestMove = pickBestNextMove(qTables, keysSoFar, board, player,
computersPlayer, tryHard)
movesSoFar.append(bestMove)
tt.applyMove(player, bestMove, board)
else:
moves = tt.listEmpties(board)
randomMove = random.choice(moves)
tt.applyMove(player, randomMove, board)
player = tt.togglePlayer(player)
winner = tt.getWinner(board)
movesLeft = not tt.noMoreMoves(board)
score = scoreEndBoard(board, winner, computersPlayer)
updateQTable(score, qTables, keysSoFar, movesSoFar, alpha)
tryHard = tryHard + tryHardGrowth
def playGame():
board = tt.genBoard()
movesLeft = True
winner = False
player = 2
computersPlayer = 2 #random.randint(1,2)
turn = 0
print("NEW GAME")
if computersPlayer == 2:
print("COMPUTER GOES FIRST...")
while (movesLeft and not winner):
if player == 2:
print("X's Turn")
else: # player == 1
print("O's Turn")
tt.printBoard(board)
if player == computersPlayer:
board = pickBestNextBoard(board, player, computersPlayer)
player = tt.togglePlayer(player)
elif player == tt.togglePlayer(computersPlayer):
validMove = False
while validMove == False:
move = input("input move of form 'y x' ")
y = int(move[0])
x = int(move[2])
# validate move
if board[y][x] is not 0:
print("!!!INVALID MOVE!!!")
continue
else:
validMove = True
board[y][x] = tt.togglePlayer(computersPlayer)
player = tt.togglePlayer(player)
turn += 1
winner = tt.getWinner(board)
movesLeft = not tt.noMoreMoves(board)
tt.printBoard(board)
if winner:
if winner == 2:
print("WINNER: X")
else: # winner == 1
print("WINNER: O")
else:
print("TIE")
def test(mct, numGames, numSims):
numWins = 0
numTies = 0
numLosses = 0
for i in tqdm(range(numGames)):
board = tt.genBoard()
movesLeft = True
winner = False
player = 2
computersPlayer = random.randint(1, 2)
while (movesLeft and not winner):
if player == computersPlayer:
simulateChildren(mct, board, player, computersPlayer, numSims)
bestBoard = pickBestNextMove(mct, board, player)
# print("################")
# tt.printBoard(board)
# tt.printBoard(bestBoard)
# print("BESTMOVE")
board = bestBoard
else:
moves = tt.listEmpties(board)
randomMove = random.choice(moves)
tt.applyMove(player, randomMove, board)
player = tt.togglePlayer(player)
winner = tt.getWinner(board)
movesLeft = not tt.noMoreMoves(board)
if winner == computersPlayer:
numWins += 1
elif winner == tt.togglePlayer(computersPlayer):
numLosses += 1
else: # tie
numTies += 1
return numWins, numLosses, numTies
def test(qTables, numGames, tryHard=1.0):
numWins = 0
numTies = 0
numLosses = 0
for i in tqdm(range(numGames)):
board = tt.genBoard()
movesLeft = True
winner = False
player = 2
keysSoFar = []
movesSoFar = []
computersPlayer = random.randint(1, 2)
while (movesLeft and not winner):
if player == computersPlayer:
bestMove = pickBestNextMove(qTables, keysSoFar, board, player,
computersPlayer, tryHard)
movesSoFar.append(bestMove)
tt.applyMove(player, bestMove, board)
else:
moves = tt.listEmpties(board)
randomMove = random.choice(moves)
tt.applyMove(player, randomMove, board)
player = tt.togglePlayer(player)
winner = tt.getWinner(board)
movesLeft = not tt.noMoreMoves(board)
if winner == computersPlayer:
numWins += 1
elif winner == tt.togglePlayer(computersPlayer):
numLosses += 1
else: # tie
numTies += 1
return numWins, numLosses, numTies
def playGame():
saveQTables = False
fileName = 'qTables.pickle'
qTables = {}
keysSoFar = []
movesSoFar = []
tryHard = 0
alpha = 0.9
numTrials = 1000000
if saveQTables:
train(qTables, numTrials, alpha, tryHard)
f = open(fileName, 'wb')
pickle.dump(qTables, f, pickle.HIGHEST_PROTOCOL)
f.close()
else:
f = open(fileName, 'rb')
qTables = pickle.load(f)
f.close()
numWins, numLosses, numTies = test(qTables, 1000)
print("VS RANDOM OPPONENT...")
print("numWins:" + str(numWins))
print("numLosses:" + str(numLosses))
print("numTies:" + str(numTies))
quit()
board = tt.genBoard()
movesLeft = True
winner = False
player = 2
computersPlayer = random.randint(1, 2)
print("NEW GAME")
if computersPlayer == 2:
print("COMPUTER GOES FIRST...")
while (movesLeft and not winner):
if player == 2:
print("X's Turn")
else: # player == 1
print("O's Turn")
tt.printBoard(board)
if player == computersPlayer:
bestMove = pickBestNextMove(qTables,
keysSoFar,
board,
player,
computersPlayer,
tryHard=1.0,
verbose=True)
movesSoFar.append(bestMove)
tt.applyMove(player, bestMove, board)
player = tt.togglePlayer(player)
elif player == tt.togglePlayer(computersPlayer):
validMove = False
while validMove == False:
move = input("input move of form 'y x' ")
y = int(move[0])
x = int(move[2])
# validate move
if board[y][x] is not 0:
print("!!!INVALID MOVE!!!")
continue
else:
validMove = True
board[y][x] = tt.togglePlayer(computersPlayer)
player = tt.togglePlayer(player)
winner = tt.getWinner(board)
movesLeft = not tt.noMoreMoves(board)
tt.printBoard(board)
score = scoreEndBoard(board, winner, computersPlayer)
updateQTable(score, qTables, keysSoFar, movesSoFar, alpha)
for key in keysSoFar:
pprint(key)
pprint(qTables[key])
if winner:
if winner == 2:
print("WINNER: X")
else: # winner == 1
print("WINNER: O")
else:
print("TIE")
def playGame():
mct = {}
board = tt.genBoard()
player = 2
computersPlayer = 2
numSimsPreGame = 100000
numSimsOnline = 100
saveMCTree = False
fileName = 'mct.pickle'
if saveMCTree:
simulateChildren(mct,
board,
player,
computersPlayer,
numSimsPreGame,
verbose=True)
f = open(fileName, 'wb')
pickle.dump(mct, f, pickle.HIGHEST_PROTOCOL)
f.close()
quit()
else:
f = open(fileName, 'rb')
mct = pickle.load(f)
f.close()
# mct = {}
# numTrials = 100
# numWins, numLosses, numTies = test(mct, numTrials, numSimsOnline)
# print("VS RANDOM OPPONENT...")
# print("numWins:" + str(numWins))
# print("numLosses:" + str(numLosses))
# print("numTies:" + str(numTies))
# quit()
# w 0.6, t 0.11, l 0.3
board = tt.genBoard()
movesLeft = True
winner = False
player = 2
computersPlayer = random.randint(1, 2)
print("NEW GAME")
if computersPlayer == 2:
print("COMPUTER GOES FIRST...")
while (movesLeft and not winner):
if player == 2:
print("X's Turn")
else: # player == 1
print("O's Turn")
tt.printBoard(board)
if player == computersPlayer:
simulateChildren(mct,
board,
player,
computersPlayer,
numSimsOnline,
verbose=True)
bestBoard = pickBestNextMove(mct, board, player)
board = bestBoard
elif player == tt.togglePlayer(computersPlayer):
validMove = False
while validMove == False:
move = input("input move of form 'y x' ")
y = int(move[0])
x = int(move[2])
# validate move
if board[y][x] is not 0:
print("!!!INVALID MOVE!!!")
continue
else:
validMove = True
board[y][x] = tt.togglePlayer(computersPlayer)
player = tt.togglePlayer(player)
winner = tt.getWinner(board)
movesLeft = not tt.noMoreMoves(board)
tt.printBoard(board)
score = scoreEndBoard(board, winner, computersPlayer)
if winner:
if winner == 2:
print("WINNER: X")
else: # winner == 1
print("WINNER: O")
else:
print("TIE")
newData.append(data[int(sampleIndex)])
sampleIndex += intervalSize
assert (len(newData) == maxWindowSize)
return newData
verbose = True
logInterval = 100
plotInterval = 20
if train:
for i in trange(numTrains):
# if i > 1000:
# makeRandomMoves = False
board = tt.genBoard()
movesLeft = True
winner = False
player = 2
computersPlayer = random.randint(1,2)
playerOneTrainingSessions = []
playerTwoTrainingSessions = []
if verbose and i % logInterval == 0:
print("NEW GAME")
while(movesLeft and not winner):
if verbose and i % logInterval == 0:
if player == 2:
print("X's Turn")
else: # player == 1