Blockade

Python remake of the 1976 multiplayer snake-like game created in Python Arcade.

The game currently supports only 2 players. There are 5 types of players (1 human and 4 AI bots):

• Human - controlling with arrows or WSAD.
• Random - a bot that randomly selects one of the possible move directions (moves erratically but avoids immediate death until it's inevitable).
• Heuristic - a bot with moves heuristically scored by the difference of available area (calculated with flood fill algorithm) and Manhattan distance to the opponent (the bot always chases the opponent while dodging death). The move with the highest score is selected unless there is a tie, then it's chosen randomly out of the top-scored moves.
• Optimized - the best bot achieved in a custom competitive coevolution process, it's similar to the Heuristic bot but the behavior is modified by 4 weights (agoraphillic/agoraphobic, aggressive/elusive, evasive/ballsy, preferring straight lines/turns).
• Reinforcement learning - a bot trained on many games while receiving rewards after every step. The training environment is a custom ParallelEnv created using Gymnasium, PettingZoo and SuperSuit. The model is a tuned A2C from Stable Baselines 3 trained on 20 million steps.

The player which survives wins!

Controls

Action	Key (arrows / wsad player)
Move UP	🡅 / W
Move DOWN	🡇 / S
Move LEFT	🡄 / A
Move RIGHT	🡆 / D
Increase game speed	+
Decrease game speed	-
Mute/unmute sound	M
Exit	Escape

Requirements

For playing the game you only need:

• Python Arcade
• NumPy
• Stable-Baselines3 (at least 2.0.0)

Details in requirements.txt.

Running

python blockade.py [-h] 
                   [-p1 {arrows,wsad,random,heuristic,optimized,rl}]
                   [-p2 {arrows,wsad,random,heuristic,optimized,rl}]
                   [-a {10,11,12,13,14,15,16,17,18,19,20}]
                   [-t {15,20,25,30,35,40,45,50,55,60,65,70,75}]
                   [-s GAME_SPEED]
                   [-r RANDOM_SEED]
                   [-m]
                   [-w]
                   [-v]

Arguments:

Long name	Short name	Description	Options	Default value
--help	-h	Shows help message and exits the game.	[flag]	False
--player1	-p1	Type of the first player (green color).	arrows, wsad, random, heuristic, optimized, rl	arrows
--player2	-p2	Type of the second player (red color).	arrows, wsad, random, heuristic, optimized, rl	random
--arena-size	-a	Size of the square game arena (in tiles).	Integers between 10-20	10
--tile-size	-t	Size of a square game tile (in pixels).	Integers between 15-75	50
--game-speed	-s	Game speed, number of moves per second.	Floats between 1.0-60.0	2.0
--random-seed	-r	RNG initialization seed, controls random behaviors of bots.	Integer	42
--mute-sound	-m	Mutes game sound effects.	[flag]	False
--window-hidden	-w	Hides game window (sound and human players are not available in this mode).	[flag]	False
--verbose	-v	Verbosity switch, prints game info to the terminal.	[flag]	False

Note: two human players can't use the same input method at the same time.

Bot training

Optimized bot

Reinforcement learning bot

Bot performance comparison

Base results

Wins/draws/loses from the perspective of Player 1.

Player 2 \ Player 1	RandomBot	HeuristicBot	OptimizedBot	ReinforcementLearningBot
RandomBot	480/66/454	834/129/37	849/128/23	560/94/346
HeuristicBot	43/85/872	344/166/490	512/88/400	24/516/460
OptimizedBot	55/137/808	272/354/374	385/221/394	1/931/68
ReinforcementLearningBot	329/126/545	829/100/71	0/1000/0	386/371/243

Aggregated results

Bot type	RandomBot	HeuristicBot	OptimizedBot	ReinforcementLearningBot
Total wins	1767	4501	3390	1830
Total draws	831	1604	3080	3509
Total loses	5402	1895	1530	2661