Make the board prettier, provide some additional stats

examples/random_tic_tac_toe.py

@@ -1,62 +0,0 @@
-import gym
-import numpy as np
-import gym_tic_tac_toe
-import random
-
-
-def random_plus_middle_move(moves, p):
-    if ([p, 4] in moves):
-        m = [p, 4]
-    else:
-        m = random_move(moves, p)
-    return m
-def random_move(moves, p):
-    m = random.choice(moves)
-    return m
-
-env = gym.make('tic_tac_toe-v0')
-
-num_episodes = 2000
-num_steps_per_episode = 10
-
-collected_rewards = []
-for i in range(num_episodes):
-    s = env.reset()
-    print (s)
-    print ("starting new episode")
-    env.render()
-    print ("started")
-    total_reward = 0
-    done = False
-    om = 1
-    for j in range(num_steps_per_episode):
-        moves = env.move_generator()
-        print ("moves: ", moves)
-        if (not moves):
-            print ("out of moves")
-            break
-        if (len(moves)==1):
-            m = moves[0]
-        else:
-            if (om == 1):
-                m = random_plus_middle_move(moves, om)
-            else:
-                m = random_move(moves, om)
-        print ("m: ", m)
-#         a = env.action_space.sample()
-#         print (a[0])
-#         #sm = s['on_move']
-#         #print (sm)
-#         a = tuple((om, a[1]))
-        s1, reward, done, _ = env.step(m)
-        om = -om
-        env.render()
-        total_reward += reward
-        s = s1
-        if done:
-            print ("game over: ", reward)
-            break
-    collected_rewards.append(total_reward)
-    print ("total reward ", total_reward, " after episode: ", i, ". steps: ", j+1)
-print ("average score: ", sum(collected_rewards) / num_episodes)
-print("#########")

gym_tic_tac_toe/__init__.py

@@ -1,6 +1,6 @@
 from gym.envs.registration import register
 
 register(
-    id='tic_tac_toe-v0',
+    id='tic_tac_toe-v1',
     entry_point='gym_tic_tac_toe.envs:TicTacToeEnv',
 )

gym_tic_tac_toe/envs/tic_tac_toe_env.py

@@ -4,17 +4,18 @@ import numpy as np
 
 class TicTacToeEnv(gym.Env):
     metadata = {'render.modes': ['human']}
+    
+    symbols = ['O', ' ', 'X'];
 
     def __init__(self):
         self.action_space = spaces.Discrete(9)
-        self.observation_space = spaces.Discrete(512*512*2) # flattened
+        self.observation_space = spaces.Discrete(9*3*2) # flattened
     def step(self, action):
         done = False
         reward = 0
 
         p, square = action
         
-       # p = p*2 - 1
         # check move legality
         board = self.state['board']
         proposed = board[square]
@@ -54,14 +55,14 @@ class TicTacToeEnv(gym.Env):
     def render(self, mode='human', close=False):
         if close:
             return
-        print("on move: " , self.state['on_move'])
+        print("on move: " , self.symbols[self.state['on_move']+1])
         for i in range (9):
-            print (self.state['board'][i], end=" ")
+            print (self.symbols[self.state['board'][i]+1], end=" ");
-        print()
+            if ((i % 3) == 2):
+                print();
     def move_generator(self):
         moves = []
         for i in range (9):
-            
             if (self.state['board'][i] == 0):
                 p = self.state['on_move']
                 m = [p, i]

random_tic_tac_toe.py

@@ -1,6 +1,6 @@
 import gym
-import numpy as np
+#import numpy as np
-import gym_tic_tac_toe
+import gym_tic_tac_toe #noqa
 import random
 
 
@@ -14,49 +14,59 @@ def random_move(moves, p):
     m = random.choice(moves)
     return m
 
-env = gym.make('tic_tac_toe-v0')
+
+env = gym.make('tic_tac_toe-v1')
 
 num_episodes = 2000
 num_steps_per_episode = 10
 
 collected_rewards = []
+oom =1
 for i in range(num_episodes):
     s = env.reset()
-    print (s)
+    #print (s)
-    print ("starting new episode")
+    #print ("starting new episode")
-    env.render()
+    #env.render()
-    print ("started")
+    #print ("started")
     total_reward = 0
     done = False
-    om = 1
+    om = oom;
+    #run one episode
+    #print("starting player: ", om);
+
     for j in range(num_steps_per_episode):
         moves = env.move_generator()
-        print ("moves: ", moves)
+        #print ("moves: ", moves)
         if (not moves):
-            print ("out of moves")
+            #print ("out of moves")
             break
         if (len(moves)==1):
+            #only a single possible move
             m = moves[0]
         else:
             if (om == 1):
                 m = random_plus_middle_move(moves, om)
+                #m = random_move(moves, om)
             else:
                 m = random_move(moves, om)
-        print ("m: ", m)
+        #print ("m: ", m)
-#         a = env.action_space.sample()
-#         print (a[0])
-#         #sm = s['on_move']
-#         #print (sm)
-#         a = tuple((om, a[1]))
         s1, reward, done, _ = env.step(m)
         om = -om
-        env.render()
+        #env.render()
         total_reward += reward
         s = s1
         if done:
-            print ("game over: ", reward)
+            #print ("game over: ", reward)
             break
+    env.render()
+    total_reward *= oom;
     collected_rewards.append(total_reward)
-    print ("total reward ", total_reward, " after episode: ", i, ". steps: ", j+1)
+    #print ("total reward", total_reward, "after episode: ", i+1, ". steps: ", j+1)
-print ("average score: ", sum(collected_rewards) / num_episodes)
+    oom = -oom
+
+print ("after "+ str(i+1) + " episodes:");
+
+average = sum(collected_rewards) / num_episodes;
+print ("average score: ", average);
+print("percentage: ", round(100*(average+1)/2,1));
 print("#########")

setup.py

@@ -2,7 +2,7 @@ from setuptools import setup
 from setuptools import find_packages
 
 setup(name='gym_tic_tac_toe',
-      version='0.0.1',
+      version='0.0.2',
       install_requires=['gym'],
       url="https://github.com/nczempin/gym-tic-tac-toe",
       packages=find_packages()