Mahjong is a tile-based game with 4 players. It uses a deck of 136 tiles that is comprised of 4 identical sets of 34 unique tiles. The objective is to form 4 sets and a pair with the 14th drawn tile. If no player wins, no player receives a reward.

Our implementation wraps RLCard and you can refer to its documentation for additional details. Please cite their work if you use this game in research.

Observation Space

The observation is a dictionary which contains an 'obs' element which is the usual RL observation described below, and an 'action_mask' which holds the legal moves, described in the Legal Actions Mask section.

The main observation space has a (6, 34, 4) shape with the first index representing the encoding plane. The contents of each plane are described in the table below:

Plane	Description
0	Current Player’s hand
1	Played tiles on the table
2	Public piles of player_0
3	Public piles of player_1
4	Public piles of player_2
5	Public piles of player_3

Encoding per Plane

Plane Row Index	Description
0 - 8	Bamboo 0: 1, 1: 2, …, 8: 9
9 - 17	Characters 9: 1, 10: 2, …, 17: 9
18 - 26	Dots 18: 1, 19: 2, …, 26: 9
27	Dragons Green
28	Dragons Red
29	Dragons White
30	Winds East
31	Winds West
32	Winds North
33	Winds South

Plane Column Index	Description
0	Tile Set 1
1	Tile Set 2
2	Tile Set 3
3	Tile Set 4

Legal Actions Mask

The legal moves available to the current agent are found in the action_mask element of the dictionary observation. The action_mask is a binary vector where each index of the vector represents whether the action is legal or not. The action_mask will be all zeros for any agent except the one whose turn it is. Taking an illegal move ends the game with a reward of -1 for the illegally moving agent and a reward of 0 for all other agents.

Action Space

The action space, as described by RLCard, is

Action ID	Action
0 - 8	Bamboo 0: 1, 1: 2, …, 8: 9
9 - 17	Characters 9: 1, 10: 2, …, 17: 9
18 - 26	Dots 18: 1, 19: 2, …, 26: 9
27	Dragons Green
28	Dragons Red
29	Dragons White
30	Winds East
31	Winds West
32	Winds North
33	Winds South
34	Pong
35	Chow
36	Gong
37	Stand

For example, you would use action 34 to pong or action 37 to stand.

Rewards

Winner	Loser
+1	-1

Version History

• v4: Upgrade to RLCard 1.0.3 (1.11.0)
• v3: Fixed bug in arbitrary calls to observe() (1.8.0)
• v2: Bumped RLCard version, bug fixes, legal action mask in observation replaced illegal move list in infos (1.5.0)
• v1: Bumped RLCard version, fixed observation space, adopted new agent iteration scheme where all agents are iterated over after they are done (1.4.0)
• v0: Initial versions release (1.0.0)