bysreg/bysai

bysai's Issues

truncate early when using minimax iterative deepening search

when all moves that visited during one of the iterative deepening depth are a winning move then we dont need to iterate deeper. minimax should just return the best move index from current depth

AI profile feature

currently, when we want to change the AI technique parameters, we have to set the param in the c++/java source code. this defeats the whole purpose of using lua technology as bysai use lua to prevent the recompilation of the game code.

the solution :
separate parameter of the AI into an external lua file, when the game loads, the game will instantiate the AI with parameter from the lua ai profile.

spec of lua ai profile :
-ai technique
-other params..

more monte carlo parameters

(duplicate)
options to tweak monte carlo :

1. maximum tree size stored
2. UCT implementation

more monte carlo parameters

add more monte carlo parameters :

1. maximum tree size
2. selection strategy