NS separate agents. When illegal move is played, return a negative reward but let the agent retry
(didn't end the game). Move was greedily decided (Max)
REWARD_DRAW_OR_NOT_FINAL = tf.constant([0.], dtype=np.float32)
REWARD_ILLEGAL_MOVE = tf.constant([-0.001], dtype=np.float32)
FULLY_CONNECTED_LAYER_PARAMS = (4000,)
LEARNING_RATE = 1e-3
BATCH_SIZE = 64
REPLAY_BUFFER_MAX_LENGTH = 100000
DISCOUNT = tf.constant([0.99], dtype=np.float)
INITIAL_COLLECT_STEPS = 100
LOG_INTERVAL = 10000
SAVE_INTERVAL = 20
Side.NS won 603 games and Side.EW won 677 games
7:47:06.146054 elapsed

0:00:00: 1111111101
0:05:00.971831: 0101010111
0:07:48.771348: 1010101010
0:11:12.614469: 1111010101
0:14:33.110924: 1010001110
0:17:35.797175: 0100001110
0:20:49.740536: 1011101100
0:24:11.448512: 1010110100
0:26:53.865620: 0011101110
0:29:50.417799: 0100111011
0:32:12.378052: 0110100111
0:34:30.271034: 0011101100
0:36:38.833860: 1011001100
0:38:58.957863: 0110110101
0:40:55.120296: 1100010001
0:43:03.203635: 1110110001
0:45:14.473143: 0110011000
0:47:45.461070: 1110100100
0:50:12.602610: 1000100101
0:53:06.456467: 1100110011
0:55:54.197811: 0110101001
0:57:53.325608: 1110111001
1:00:25.709639: 0110011101
1:03:17.224100: 0101110100
1:05:34.363476: 1010001110
1:08:35.271761: 0011100100
1:10:51.897015: 1001110010
1:13:14.242905: 1101111001
1:15:32.656020: 1000111101
1:17:48.139378: 1001111100
1:20:04.533261: 1101001111
1:22:34.688367: 1001110111
1:24:29.242717: 1001100011
1:27:08.930540: 1110011010
1:29:02.145087: 0011110110
1:31:31.089281: 1101010110
1:33:48.451387: 1010101000
1:36:31.284616: 0010010101
1:38:58.686437: 0100011100
1:41:39.252454: 1111111011
1:43:39.894701: 1011011000
1:46:01.917200: 0110001111
1:48:10.889278: 0111001010
1:50:34.835449: 0111101010
1:53:15.348501: 0001111100
1:55:41.655765: 1111000011
1:57:33.291502: 0010001010
1:59:57.635708: 1000111010
2:02:09.920365: 1001010110
2:04:18.300589: 1011111010
2:07:00.344096: 1010010110
2:09:04.625193: 0010101011
2:10:53.093393: 0110100100
2:13:08.971044: 1010010100
2:15:03.818088: 1100001110
2:16:47.852043: 1001101010
2:18:50.720521: 0101010000
2:20:58.836941: 1000001100
2:22:44.923732: 0010011011
2:25:30.842505: 0101010101
2:27:34.522499: 1111001101
2:29:46.624307: 1001010010
2:31:56.370750: 0001010011
2:34:49.525006: 1010101001
2:37:16.146394: 1011110101
2:39:26.964914: 0111111100
2:41:29.703319: 1001110000
2:43:51.063756: 1110000101
2:45:53.795283: 0110001011
2:47:58.071451: 0101100111
2:50:42.290468: 1101100111
2:53:20.598148: 1010001100
2:56:35.461564: 1101100110
2:58:56.570320: 0101110001
3:01:04.691158: 0110010000
3:03:39.438092: 0110101000
3:05:28.242350: 0010101110
3:08:00.772334: 0111100111
3:10:33.274853: 0011110110
3:12:36.611303: 1010110000
3:14:47.231833: 1111011111
3:17:09.420464: 1000011001
3:20:03.552167: 0101111111
3:22:05.134713: 1011101011
3:24:24.813706: 0110011111
3:27:15.101075: 0100100101
3:29:14.795210: 1001100100
3:31:33.266237: 0011100101
3:33:34.037560: 1010100110
3:35:53.025941: 1110011000
3:37:39.859502: 1001111101
3:40:06.291279: 1101001010
3:42:22.486629: 1001110101
3:44:44.685988: 1101011000
3:47:28.183262: 0011010100
3:49:28.173152: 1011001110
3:51:28.496307: 1110111011
3:54:03.654214: 0000110111
3:56:15.117226: 0011000010
3:58:32.692443: 1101111000
4:00:43.035533: 1000100111
4:02:24.130286: 0100101000
4:04:42.964088: 1010101000
4:06:23.662103: 0101011101
4:09:04.578292: 0110010111
4:11:01.422205: 1111011011
4:13:07.795710: 1101000010
4:15:18.332279: 0011011001
4:17:54.803209: 0101000111
4:20:46.812665: 1101110110
4:22:25.298680: 1001011000
4:24:48.933477: 0010110110
4:27:06.256427: 0101100001
4:28:52.724730: 1100001011
4:31:45.962178: 0011001010
4:33:51.914240: 1101001111
4:36:29.944942: 1100011010
4:38:59.834199: 0100011011
4:41:23.254594: 0100001101
4:43:58.319486: 0010110100
4:46:06.120849: 0010001010
4:48:50.066704: 1111101010
4:51:02.472187: 1110011011
4:52:49.418448: 0111111101
4:55:26.827236: 1110010001
4:57:33.962298: 0001000110
5:00:27.930914: 0001100100
5:03:34.829447: 0010010000