def get_reward(state, stateprime, player):
# TODO
'''
REWARD RULESET
1. 1 point for every 2 in a row
2. 2 points for every 3 in a row
3. -1 point for every opponent 2 in a row
4. -2 points for every opponent 3 in a row
5. 9999999? for four in a row
6. -9999999? for opponent 4 in a row
'''
scorecard = {2: 10, 3: 100, 4: 99999999}
# brute force
arr = []
for column in state:
tern = ternery(column)
while len(tern) != 6:
tern = '0' + tern
arr.append(list(tern))
p1_rows = findInARow(arr, '1')
p2_rows = findInARow(arr, '2')
state1_score = 0
for i in p1_rows:
state1_score = state1_score + scorecard[len(
i)] if player == '1' else state1_score - scorecard[len(i)]
for i in p2_rows:
state1_score = state1_score - scorecard[len(
i)] if player == '1' else state1_score + scorecard[len(i)]
arr = []
for column in stateprime:
tern = ternery(column)
while len(tern) != 6:
tern = '0' + tern
arr.append(list(tern))
p1_rows = findInARow(arr, '1')
p2_rows = findInARow(arr, '2')
state2_score = 0
for i in p1_rows:
state2_score = state2_score + scorecard[len(
i)] if player == '1' else state2_score - scorecard[len(i)]
for i in p2_rows:
state2_score = state2_score - scorecard[len(
i)] if player == '1' else state2_score + scorecard[len(i)]
return state2_score - state1_score if player == '1' else state1_score - state2_score
def transition(self, column_choice, player):
column = self.state[column_choice]
tern = ternery(column)
if len(tern) == 6:
return False
new = str(player) + tern if tern != '0' else str(player)
self.state[column_choice] = int(new, 3)
return True
def pBoard(b):
for i in range(6):
for j in b.state:
t = ternery(j)
while len(t) < 6:
t = '0' + t
print(t[i], end='')
print()
def isFull(self):
l = 0
for i in range(7):
column = self.state[i]
tern = ternery(column)
if len(tern) == 6:
l += 1
if l == 7:
return True
else:
return False
def neural_network_choose(board, player, model): inputs = [] for col in board.state: tern = ternery(col) if col != ' ' else '' a = [int(n) for n in list(tern)] for n, i in enumerate(a): if i == 2: a[n] = -1 while len(a) != 6: a.append(0) inputs += a array = np.array([inputs]) t = model.predict(array, batch_size=10).tolist() board.transition(t.index(max(t)), player) return board
def gameOver(self):
arr = []
for column in self.state:
tern = ternery(column)
while len(tern) != 6:
tern = '0' + tern
arr.append(list(tern))
p1_rows = findInARow(arr, '1')
p2_rows = findInARow(arr, '2')
if 4 in [len(n) for n in p1_rows]:
self.winner = '1'
return True
if 4 in [len(n) for n in p2_rows]:
self.winner = '2'
return True
elif self.isFull():
return True
else:
return False