def test_get_value_and_state_returns_current_value_if_state_known(self):
state1 = get_board_state_tuple("X--|---|---")
state2 = get_board_state_tuple("XO-|---|---")
self.player.values[state1] = 0.6
self.player.values[state2] = 0.3
self.assert_get_value_and_state_is(0.6, "X--|---|---", None, Board.X)
self.assert_get_value_and_state_is(0.3, "XO-|---|---", None, Board.O)
def test_get_move_chooses_random_best_available_move_if_random_gte_epsilon_and_multiple_bests(
self, random_mock, choice_mock):
random_mock.return_value = 0.1
choice_mock.side_effect = MockRandom(1).choice
self.player.values[get_board_state_tuple(
"X--|-XO|---")] = 0.501 # position 0
self.player.values[get_board_state_tuple(
"--X|-XO|---")] = 0.501 # position 2
# Symmetries for X--|-XO|---:
# -O-|-X-|X-- (Rotated by 90 degrees)
# ---|OX-|--X (Rotated by 180 degrees)
# --X|-X-|-O- (Rotated by 270 degrees)
# ---|-XO|X-- (Reflected horizontally) - position 6
# --X|OX-|--- (Reflected vertically)
# X--|-X-|-O- (Reflected on left diagonal)
# -O-|-X-|--X (Reflected on right diagonal)
#
# Symmetries for --X|-XO|---:
# XO-|-X-|--- (Rotated by 90 degrees)
# ---|OX-|X-- (Rotated by 180 degrees)
# ---|-X-|-OX (Rotated by 270 degrees)
# ---|-XO|--X (Reflected horizontally) - position 8
# X--|OX-|--- (Reflected vertically)
# ---|-X-|XO- (Reflected on left diagonal)
# -OX|-X-|--- (Reflected on right diagonal)
assert_get_move_is(self, self.player, self.board, 2, Board.X,
"---|-XO|---")
choice_mock.assert_called_once_with([0, 2, 6, 8])
def test_get_move_chooses_random_best_available_move_if_random_gte_epsilon_and_multiple_bests(
self, random_mock, choice_mock):
random_mock.return_value = 0.1
choice_mock.side_effect = MockRandom(1).choice
self.player.values[get_board_state_tuple("X--|-XO|---")] = 0.501
self.player.values[get_board_state_tuple("--X|-XO|---")] = 0.501
self.player.values[get_board_state_tuple("---|-XO|X--")] = 0.501
assert_get_move_is(self, self.player, self.board, 2, Board.X,
"---|-XO|---")
choice_mock.assert_called_once_with([0, 2, 6])
def test_set_reward_does_not_update_values_for_each_state_if_learning_disabled(
self):
self.player.disable_learning()
self.player.values[get_board_state_tuple("---|-X-|---")] = 0.6
self.player.values[get_board_state_tuple("-O-|-X-|---")] = 0.55
self.player.values[get_board_state_tuple("-O-|-X-|--X")] = 0.7
self.player.values[get_board_state_tuple("-OO|-X-|--X")] = 0.85
self.player.values[get_board_state_tuple("XOO|-X-|--X")] = 1.0
self.assert_values_after_reward_are([0.6, 0.55, 0.7, 0.85, 1.0], [
"---|-X-|---", "-O-|-X-|---", "-O-|-X-|--X", "-OO|-X-|--X",
"XOO|-X-|--X"
], Board.X)
def assert_get_value_and_state_symmetric_is(self, value, pieces,
symmetric_pieces, winner,
piece):
self.player.set_piece(piece)
state = get_board_state_tuple(pieces)
symmetric_state = get_board_state_tuple(symmetric_pieces)
current_value, new_state = self.player._get_value_and_state(
symmetric_state, winner)
self.assertAlmostEqual(value, current_value)
self.assertEqual(state, new_state)
self.assertIn(state, self.player.values)
if symmetric_state != state:
self.assertNotIn(symmetric_state, self.player.values)
def test_get_move_chooses_best_available_move_if_random_gte_epsilon(
self, random_mock, choice_mock):
random_mock.return_value = 0.1
choice_mock.side_effect = MockRandom(0).choice
self.player.values[get_board_state_tuple("---|-XO|---")] = 0.501
assert_get_move_is(self, self.player, self.board, 5, Board.O,
"---|-X-|---")
choice_mock.assert_called_once_with([5])
def assert_game_info_is(self, winner, winning_positions, turn, pieces, game_info):
expected_game_info = \
{
"winner": winner,
"winning_positions": winning_positions,
"turn": turn,
"board": list(get_board_state_tuple(pieces))
}
self.assertEqual(expected_game_info, game_info)
def test_get_move_chooses_best_available_move_if_learning_disabled(
self, random_mock, choice_mock):
self.player.disable_learning()
random_mock.return_value = 0.099
choice_mock.side_effect = MockRandom(0).choice
self.player.values[get_board_state_tuple("---|-O-|--X")] = 0.501
assert_get_move_is(self, self.player, self.board, 4, Board.O,
"---|---|--X")
random_mock.assert_not_called()
choice_mock.assert_called_once_with([4])
def assert_values_after_reward_are(self, values, pieces_list, winner):
self.player.set_piece(Board.X)
values_dict = {}
for pieces, value in zip(pieces_list, values):
set_board(self.board, pieces)
self.player.store_state()
values_dict[get_board_state_tuple(pieces)] = value
self.player.set_reward(winner)
self.assertEqual(sorted(values_dict.keys()),
sorted(self.player.values.keys()))
for key, value in self.player.values.items():
self.assertAlmostEqual(self.player.values[key], value)
def test_get_move_chooses_best_available_move_if_random_gte_epsilon(
self, random_mock, choice_mock):
random_mock.return_value = 0.1
choice_mock.side_effect = MockRandom(0).choice
self.player.values[get_board_state_tuple(
"---|-XO|---")] = 0.501 # Original
# Reflected horizontally
# Force other symmetric choices to be of lower value
self.player.values[get_board_state_tuple(
"-O-|-X-|---")] = 0.499 # Rotated by 90 degrees
# Reflected on right diagonal
self.player.values[get_board_state_tuple(
"---|OX-|---")] = 0.499 # Rotated by 180 degrees
# Reflected vertically
self.player.values[get_board_state_tuple(
"---|-X-|-O-")] = 0.499 # Rotated by 270 degrees
# Reflected on left diagonal
assert_get_move_is(self, self.player, self.board, 5, Board.O,
"---|-X-|---")
choice_mock.assert_called_once_with([5])
def test_get_move_values_returns_move_values_for_available_moves(self):
self.player.values[get_board_state_tuple("XOO|---|-X-")] = 0.7
self.player.values[get_board_state_tuple("XO-|O--|-X-")] = 0.65
self.player.values[get_board_state_tuple("XO-|-O-|-X-")] = 0.75
self.player.values[get_board_state_tuple("XO-|--O|-X-")] = 0.62
self.player.values[get_board_state_tuple("XO-|---|OX-")] = 0.72
self.player.values[get_board_state_tuple("XO-|---|-XO")] = 0.67
assert_get_move_values_are(self, self.player, self.board, {
2: 0.7,
3: 0.65,
4: 0.75,
5: 0.62,
6: 0.72,
8: 0.67
}, Board.O, "XO-|---|-X-")
def test_get_num_states_returns_correct_num_of_states(self):
self.player.values[get_board_state_tuple("---|-X-|---")] = 0.9
self.player.values[get_board_state_tuple("---|-X-|--O")] = 0.75
self.player.values[get_board_state_tuple("---|XX-|--O")] = 0.7
self.assertEqual(3, self.player.get_num_states())
def test_get_value_and_state_returns_current_value_if_symmetric_state_known(
self):
state1 = get_board_state_tuple("XO-|X--|---")
state2 = get_board_state_tuple("XO-|X--|O--")
self.player.values[state1] = 0.56
self.player.values[state2] = 0.45
# Original
self.assert_get_value_and_state_symmetric_is(0.56, "XO-|X--|---",
"XO-|X--|---", None,
Board.X)
self.assert_get_value_and_state_symmetric_is(0.45, "XO-|X--|O--",
"XO-|X--|O--", None,
Board.O)
# Rotated by 90 degrees
self.assert_get_value_and_state_symmetric_is(0.56, "XO-|X--|---",
"---|O--|XX-", None,
Board.X)
self.assert_get_value_and_state_symmetric_is(0.45, "XO-|X--|O--",
"---|O--|XXO", None,
Board.O)
# Rotated by 180 degrees
self.assert_get_value_and_state_symmetric_is(0.56, "XO-|X--|---",
"---|--X|-OX", None,
Board.X)
self.assert_get_value_and_state_symmetric_is(0.45, "XO-|X--|O--",
"--O|--X|-OX", None,
Board.O)
# Rotated by 270 degrees
self.assert_get_value_and_state_symmetric_is(0.56, "XO-|X--|---",
"-XX|--O|---", None,
Board.X)
self.assert_get_value_and_state_symmetric_is(0.45, "XO-|X--|O--",
"OXX|--O|---", None,
Board.O)
# Reflected horizontally
self.assert_get_value_and_state_symmetric_is(0.56, "XO-|X--|---",
"---|X--|XO-", None,
Board.X)
self.assert_get_value_and_state_symmetric_is(0.45, "XO-|X--|O--",
"O--|X--|XO-", None,
Board.O)
# Reflected vertically
self.assert_get_value_and_state_symmetric_is(0.56, "XO-|X--|---",
"-OX|--X|---", None,
Board.X)
self.assert_get_value_and_state_symmetric_is(0.45, "XO-|X--|O--",
"-OX|--X|--O", None,
Board.O)
# Reflected on left diagonal
self.assert_get_value_and_state_symmetric_is(0.56, "XO-|X--|---",
"XX-|O--|---", None,
Board.X)
self.assert_get_value_and_state_symmetric_is(0.45, "XO-|X--|O--",
"XXO|O--|---", None,
Board.O)
# Reflected on right diagonal
self.assert_get_value_and_state_symmetric_is(0.56, "XO-|X--|---",
"---|--O|-XX", None,
Board.X)
self.assert_get_value_and_state_symmetric_is(0.45, "XO-|X--|O--",
"---|--O|OXX", None,
Board.O)