def expectimax_calc(move, old_board, depth, isChance):
board_to_check = util.execute_move(move, old_board)
if depth == MAX_DEPTH:
return heuristic_value(move, board_to_check, old_board)
if isChance == 1:
avg = 0
ecc = util.cell_of_type_count(0, board_to_check)
if ecc == 0:
possibility_multiplier = 1
else:
possibility_multiplier = 1 / ecc
for x in range(4):
for y in range(4):
if board_to_check[x][y] == 0:
board_to_check[x][y] = 2
avg += possibility_multiplier * CHANCE_2 * expectimax_calc(move, old_board, depth + 1, 0)
board_to_check[x][y] = 4
avg += possibility_multiplier * CHANCE_4 * expectimax_calc(move, old_board, depth + 1, 0)
board_to_check[x][y] = 0
return avg
else:
results = []
for m in move_args:
check_board = util.execute_move(m, board_to_check)
if util.board_equals(check_board, board_to_check):
results.append(0)
else:
results.append(expectimax_calc(m, board_to_check, depth + 1, 1))
return max(results)
def a_very_lauenchr_expectimax_calc(depth, new_board, isChance, old_board):
if depth == MAX_DEPTH:
return heuristic_value(new_board, old_board)
if isChance == 1:
avg = 0
chance_2 = 0.9
chance_4 = 0.1
ecc = util.cell_of_type_count(0, new_board)
if ecc == 0:
chance_cel = 0
else:
chance_sel = 1 / heuristics_empty_cells_count(new_board)
for x in range(4):
for y in range(4):
if new_board[x][y] == 0:
new_board[x][y] = 2
avg += chance_sel * chance_2 * a_very_lauenchr_expectimax_calc(depth + 1, new_board, 0, old_board)
new_board[x][y] = 4
avg += chance_sel * chance_4 * a_very_lauenchr_expectimax_calc(depth + 1, new_board, 0, old_board)
new_board[x][y] = 0
return avg
else:
results = []
for m in move_args:
results.append(a_very_lauenchr_expectimax_calc(depth + 1, util.execute_move(m, new_board), 1, new_board))
return max(results)
def heuristic_empty_tiles(board): ''' Returns the amount of empty tiles in the given board, weighted to have a value between 0 (no spaces) and 1 (1 being an empty board) ''' empty_tiles_count_weighted = util.cell_of_type_count(0, board) / 15 return util.round_nearest(empty_tiles_count_weighted, 0.01)
def _empty_cells_weighted(board):
'''
only used for expectimax calc
'''
count = util.cell_of_type_count(0, board)
if count == 0:
return 1
return count
def expectimax_calc(board_to_check, old_board, depth, isChance):
# TODO: kann man bei der Suche abbrechen, :higest_tile aus der Ecke getrieben wird?
# => welchen Einfluss hat dies auf Anfangsszenarien?
if depth == MAX_DEPTH:
return heuristic_value(board_to_check, old_board)
if isChance == 1:
avg = 0
empty_cells_count = util.cell_of_type_count(0, board_to_check)
if empty_cells_count == 0:
moves_to_check_count = 0
return 0
else:
possibility_multiplier = 1 / empty_cells_count
for x in range(4):
for y in range(4):
if board_to_check[x][y] == 0:
board_to_check[x][y] = 2
avg += possibility_multiplier * CHANCE_2 * expectimax_calc(
board_to_check, old_board, depth + 1, 0)
board_to_check[x][y] = 4
avg += possibility_multiplier * CHANCE_4 * expectimax_calc(
board_to_check, old_board, depth + 1, 0)
board_to_check[x][y] = 0
return avg
else:
results = []
for m in move_args:
results.append(
expectimax_calc(util.execute_move(m, board_to_check),
board_to_check, depth + 1, 1))
return max(results)
def heuristics_empty_cells_count(board):
return util.cell_of_type_count(0, board)
def heuristics_empty_cells_count(board):
return util.round_nearest(util.cell_of_type_count(0, board) / 15, 0.01)
def heuristic_low_tiles(board): ''' Returns the amount of 2 and 4 tiles on the given board. ''' return (util.cell_of_type_count(2, board) + util.cell_of_type_count(4, board)) / 8 + 0.1
