MUR: Positions Over Moves
Mur is history's first open system abstract strategy game. Mur is an open system because it has no capturing and allows a player to re-enter game stones he was forced to withdraw. Moreover, Mur was designed to have a mobile center. These two aspects of the design allows its material to be sustained throughout play while allowing the game to be reset yet continued from the final position of the previous game allowing for over a billion possible arrangements for each new setup.
The design focuses more on the number of possible game positions rather than the number of moves possible during any given turn. The reason for this is to provide a massive challenge to any computer tree search routine; besides, recent history has shown us that AI is eventually fully dominant against humans no matter how many possible moves are thrown its way. China's ancient game of weichei is more complex than chess in terms of the number of moves and positions available per turn. In terms of moves weichei offers hundreds of possible moves during a turn; chess offers only twenty or thirty moves; and Mur offers even less than half that of chess.
In 2003, computer engineer Omar Syed invented a simple abstract strategy game that offered over seventeen thousand options for each move; however, increasing the number of move options still led to AI becoming better than humans within six years of the game being invented. The key to human dominance over AI then is not to be found in the number of possible moves but rather by increasing the number of positions. Mur achieves this increase in positions dramatically to say the very least. The setup for the first game is always the same in Mur; however, the second game--except for bringing the mur stone to the center of the board--uses the final position of the previous game as the setup for the next game. Mur has over a billion possible setup arrangements of the game stones for the second game with each branching off into another billion possible setups for the third game!
Mur was designed to be ultra pattern-sensitive so that even game positions with similar patterns would have wildly different outcomes which prevents any computer neural network from having any real strong advantage from identifying the most similar of patterns. Moreover, a neural network needs to be trained on a good data set. Alhpha Go--the AI which defeated human weichei players--designed a neural network which relied on the KGS Go server which has over thirty million positions on record; however, a data set from Mur games would not be as significant due to the ultra pattern-sensitive nature of the game and would really only be a drop in the ocean of possible game positions.