- added reward sum and reward mean

machine_player.py

@@ -18,6 +18,7 @@ class MachinePlayer(APlayer):
 		self.with_debug = False
 		self.values = values
 		self.episode_history = None
+		self.reward_sum = 0
 
 	def set_debug(self, with_debug):
 		self.with_debug = with_debug
@@ -44,31 +45,43 @@ class MachinePlayer(APlayer):
 
 		return result
 
-	def set_value(self, state: np.array, value):
-		self.values[self.to_key(state)] = value
+	def set_value(self, state: np.array, reward):
+		self.values[self.to_key(state)] = reward
+		self.reward_sum += reward
+		if self.with_debug:
+			print(f"{self.mark}: Rewarded {reward:0.3f}")
 
 	def end_game(self, state: np.array, value):
 		self.set_value(self.state, value)
 		if self.params.do_learn_from_history:
 			self.learn_from_history()
 
+		r_sum = self.reward_sum
+		r_mean = self.reward_sum/self.count
+
+		if self.with_debug:
+			print(f"Reward sum  : {r_sum}")
+			print(f"Reward mean : {r_mean}")
+
+		return r_sum, r_mean
+
 	def new_game(self, state: np.array):
 		self.state = state
 		self.episode_history = []
+		self.reward_sum = 0
+		self.count = 0
 
 	def learn_from_history(self):
 		rev_hist = list(reversed(self.episode_history))
 		for hist in rev_hist:
+			if not hist['is_exp']:
 				d = self.calc_value(hist['state'], hist['next_state'])
-				if not hist['is_exp']:
-					self.set_value(hist['state'], d)
+				self.set_value(hist['state'], d)
 
 	def calc_value(self, state: np.array, next_state: np.array):
 		v0 = self.get_value(state)
 		v1 = self.get_value(next_state)
 		d = v0 + self.params.alpha*(v1-v0)
-		if self.with_debug:
-			print(f"{self.mark}: Learned {d:0.3f}")
 
 		return d
 
@@ -121,6 +134,6 @@ class MachinePlayer(APlayer):
 			self.set_value(self.state, d)
 
 		self.state = next_state
-
+		self.count += 1
 		return next_state, True