Refactored

a_player.py

@@ -1,4 +1,5 @@
 import numpy as np
+from state import f_state_slices
 
 
 class APlayer(object):
@@ -14,17 +15,11 @@ class APlayer(object):
 	def set_debug(self, with_debug):
 		pass
 
-	@staticmethod
-	def get_potential_moves(state: np.array) -> np.array:
-		st = state.reshape(state.size)
-		indices = [idx for idx, s in enumerate(st) if '-' in s]
-		return np.array(indices)
-
 	def move(self, state: np.array):
 		return state, False
 
-	def state_from_move(self, state: np.array, field):
+	def to_state(self, state: np.array, move):
-		state.reshape(state.size)[field] = self.mark
+		state.reshape(state.size)[move] = self.mark
 		return state
 
 	def reward(self, value):
@@ -33,39 +28,10 @@ class APlayer(object):
 	def new_game(self):
 		pass
 
-	@staticmethod
-	def f_state_slices():
-		nr = 3
-		nc = 3
-		result = []
-		# Create row finishing states
-		d1 = ()
-		d2_r = ()
-		for r in range(0, nr):
-			d1 += (r,)
-		for c in range(0, nc):
-			d2 = (c,) * nc
-			result.append((d1, d2))
-
-		# Create column finishing states
-		for c in range(0, nc):
-			d2 = (c,) * nc
-			result.append((d2, d1))
-		for c in range(0, nc):
-			d2_r += (nc - c - 1,)
-
-		# Create diagonal finishing states #1
-		result.append((d1, d1))
-
-		# Create diagonal finishing states #2
-		result.append((d1, d2_r))
-
-		return result
-
 	def has_won(self, state: np.array):
 		result = False
 		ref = [self.mark, self.mark, self.mark]
-		for f_state_slice in APlayer.f_state_slices():
+		for f_state_slice in f_state_slices():
 			if np.all(state[f_state_slice] == ref):
 				result = True
 				break

human_player.py

@@ -2,6 +2,7 @@ import numpy as np
 
 from a_player import APlayer
 from helper import to_state_string, create_empty_state, create_test_state
+from state import get_potential_moves
 
 
 class HumanPlayer(APlayer):
@@ -29,7 +30,7 @@ class HumanPlayer(APlayer):
 				continue
 
 			move = int(choice) - 1
-			if not move in self.get_potential_moves(state):
+			if not move in get_potential_moves(state):
 				print("Feld is bereits belegt!")
 				print("Versuche es nochmal")
 				continue
@@ -38,7 +39,7 @@ class HumanPlayer(APlayer):
 			has_moved = True
 			break
 
-		state_next = self.state_from_move(state.copy(), move)
+		state_next = self.to_state(state.copy(), move)
 		return state_next, has_moved
 
 	def reward(self, value):

machine_player.py

@@ -2,7 +2,7 @@ import numpy as np
 from a_player import APlayer
 from helper import sample
 from numpy.random import uniform
-
+from state import get_potential_moves
 
 class MachinePlayer(APlayer):
 	class Params:
@@ -56,7 +56,7 @@ class MachinePlayer(APlayer):
 		do_sample = False
 		values = np.array([])
 		# get possible move
-		moves = self.get_potential_moves(state)
+		moves = get_potential_moves(state)
 		if moves.size == 0:
 			return state, False
 
@@ -64,7 +64,7 @@ class MachinePlayer(APlayer):
 		best_value = -1
 		for move in moves:
 			# create hypothetical next state
-			state_next = self.state_from_move(state.copy(), move)
+			state_next = self.to_state(state.copy(), move)
 			# evaluate value
 			value = self.get_value(state_next)
 			if best_value < value:
@@ -91,7 +91,7 @@ class MachinePlayer(APlayer):
 		if self.state is not None:
 			self.state_last = self.state.copy()
 
-		self.state = self.state_from_move(state.copy(), next_move)
+		self.state = self.to_state(state.copy(), next_move)
 
 		# Learn
 		if not is_exp and self.state_last is not None:

state.py

@@ -0,0 +1,36 @@
+import numpy as np
+
+
+def get_potential_moves(state: np.array) -> np.array:
+	st = state.reshape(state.size)
+	indices = [idx for idx, s in enumerate(st) if '-' in s]
+	return np.array(indices)
+
+
+def f_state_slices():
+	nr = 3
+	nc = 3
+	result = []
+	# Create row finishing states
+	d1 = ()
+	d2_r = ()
+	for r in range(0, nr):
+		d1 += (r,)
+	for c in range(0, nc):
+		d2 = (c,) * nc
+		result.append((d1, d2))
+
+	# Create column finishing states
+	for c in range(0, nc):
+		d2 = (c,) * nc
+		result.append((d2, d1))
+	for c in range(0, nc):
+		d2_r += (nc - c - 1,)
+
+	# Create diagonal finishing states #1
+	result.append((d1, d1))
+
+	# Create diagonal finishing states #2
+	result.append((d1, d2_r))
+
+	return result