def num_state_action_comb():
""" Estimate how many state-action combinations there are
to evaluate the feasibility of a table-based q-learning approach. """
empty = np.zeros(NUM_CARD_VALUES, dtype=np.int8)
total_comb = []
comb_ap = 1
for i in range(NUM_CARD_VALUES):
comb_ap *= (AVAILABLE_CARDS[i] + 1)
for _ in tqdm.tqdm(range(100000)):
# Already played can be anything
ap = np.zeros(NUM_CARD_VALUES, dtype=np.int8)
for i in range(NUM_CARD_VALUES):
ap[i] = np.random.randint(0, AVAILABLE_CARDS[i] + 1)
# Board must be a move that is already included in already played
comb_b = possible_next_moves(ap, empty).shape[0]
# Actions from other agents must be within the remaining cards
remaining = AVAILABLE_CARDS - ap
hands = random_initial_cards(cards_to_use=remaining)
comb_actions = 1
for hand in hands:
comb_actions *= possible_next_moves(hand, empty).shape[0] \
if not np.all(hand == 0) else 1
total_comb.append(comb_ap * comb_b * comb_actions)
# 68,319,447,356,160,000
print(max(total_comb))
# 138,538,758,758,400
print(min(total_comb))
def test_possible_next_moves_empty_board(self):
""" Tests possible moves on empty board. """
hand = np.array([0, 2, 0, 0, 0, 0, 0, 2, 0, 2, 0, 0, 2])
board = np.zeros(NUM_CARD_VALUES)
actions = possible_next_moves(hand, board)
only_passing = only_passing_possible(hand, board)
self.assertFalse(only_passing)
self.assertTrue(
np.all(actions == np.array(
[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 2],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 2],
[0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 1],
[0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 2],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 2],
[0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
[0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2],
[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2]])))
def do_step(
# Board state
self,
already_played,
board,
agents_finished,
# Possible states before the next move of this agent
list_next_possible_states=lambda ap, b: ([], []),
# Other parameters
always_use_best=False,
print_luck=False):
"""
Performs a (partial) step in the game.
Returns (Player finished,
Already played cards, New board,
Best decision made randomly)
"""
# Prepares the step to do
self.prepare_step()
# If player has already finished, pass
if has_finished(self.hand):
return True, already_played, board, False
# Possible actions; Pass if no possible play
possible_actions = possible_next_moves(self.hand, board)
if len(possible_actions) == 1 and \
np.all(possible_actions[0] == 0):
return False, already_played, board, False
# Decide action to take
(possible_qvalues, action_index, action_taken,
random_choice, best_decision_made_randomly) = \
self.decide_action_to_take(
already_played, board, always_use_best,
print_luck, possible_actions)
# Compute next state
next_hand = self.hand - action_taken
next_board = board if np.all(action_taken == 0) else action_taken
next_already_played = already_played + action_taken
# Process next state
self.process_next_board_state(
already_played, board, list_next_possible_states,
next_already_played, next_board, next_hand, possible_qvalues,
action_index, action_taken, random_choice, agents_finished,
always_use_best)
# Return next state
self.hand = next_hand
return (has_finished(self.hand), next_already_played, next_board,
best_decision_made_randomly)
def test_possible_next_moves_empty_board_lots_of_cards(self):
""" Tests possible moves on empty board with lots of cards. """
hand = AVAILABLE_CARDS
board = np.zeros(NUM_CARD_VALUES)
actions = possible_next_moves(hand, board)
only_passing = only_passing_possible(hand, board)
true_actions = pd.read_csv("./egd/game/tests/test-data/actions.csv",
header=None).values
self.assertFalse(only_passing)
self.assertTrue(np.all(actions == true_actions))
def test_possible_next_moves_non_empty_board(self):
""" Tests possible next moves. """
hand = np.array([0, 2, 0, 0, 0, 0, 0, 2, 0, 2, 0, 0, 2])
board = np.array([0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0])
actions = possible_next_moves(hand, board)
only_passing = only_passing_possible(hand, board)
self.assertFalse(only_passing)
self.assertTrue(
np.all(actions == np.array(
[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1],
[0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1]])))
hand = np.array([1, 2, 3, 1, 0, 0, 0, 3, 0, 4, 0, 0, 2])
board = np.array([0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 1])
actions = possible_next_moves(hand, board)
only_passing = only_passing_possible(hand, board)
self.assertFalse(only_passing)
self.assertTrue(
np.all(actions == np.array(
[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 3, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 1],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 2],
[0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 2],
[0, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
[0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2],
[0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2],
[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2]])))
def test_possible_next_moves_bug_jokers_only(self):
""" Tests that a historic bug does not occur again. """
hand = np.array([0, 1, 0, 0, 0, 1, 0, 0, 2, 4, 2, 4, 1])
board = np.array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0])
actions = possible_next_moves(hand, board)
only_passing = only_passing_possible(hand, board)
self.assertFalse(only_passing)
self.assertTrue(
np.all(actions == np.array([
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
])))