06-21-2022, 01:31 PM
It would help a lot in the testing of strategies and manual tuning of A/B AI if Ludii could define and use a separate set of Heuristics for each player.
(ai { { <role> (bestAgent... ) (heuristics ...) } { <role> (bestAgent... ) (heuristics ...)}... } )
Separately, it might be useful to have variable transforms such as (counter)/(boardSize); (emptyCount)/(boardSize); (MoverCount)/(boardSize); (pieceCount/(boardSize); etc
These would allow the heuristics to change over the course of the game.
I don't know how much they could be optimized so as not to reduce AI depth too much though.
(ai { { <role> (bestAgent... ) (heuristics ...) } { <role> (bestAgent... ) (heuristics ...)}... } )
Separately, it might be useful to have variable transforms such as (counter)/(boardSize); (emptyCount)/(boardSize); (MoverCount)/(boardSize); (pieceCount/(boardSize); etc
These would allow the heuristics to change over the course of the game.
I don't know how much they could be optimized so as not to reduce AI depth too much though.