Macblut is an intelligent Tablut player written in Java and based on space state search. It follows a variation of the rules proposed in the work of John C. Ashton.
Specifically, Macblut uses a minimax algorithm with alpha-beta cuts, in particular the implementation provided by the AIMA library. The quality of cutoff states is determined through a custom heuristic, which differs based on whether the player is playing as black (attacker) or white (defender).
Further information (in Italian) can be found in “presentation.pdf”.
- Percentage of yet non-captured white pawns
- Percentage of captured black pawns
- Chances of the king escaping
- Distance of black pawns from the king
- King in castle
- King in danger
- Percentage of yet non-captured black pawns
- Percentage of captured white pawns
- “Rhombus formation”
- Distance of black pawns from the king
- King in danger