Enter Your Reply

Perhaps this is a bit off-topic in this place, but just out of curiosity: can you elaborate a bit on this 'digital implementation'? Does this include an AI, and if so, is this an AlphaZero-type self-learning, or a conventional alpha-beta search (with NNUE or hand-crafted)? It would be very nice if you could create a strong player for games with promotion-on-capture and contageon, such as Maka Dai Dai Shogi, and its shrunken versions. As it is not clear to me what would be a good strategy for such games.

It's already available, you can read about it and find a link here: https://drericsilverman.com/2021/05/13/ancient-shogi-revival-part-ii-the-big-ones/

Everything up to Tai Shogi is now present. Numerous options are available to attempt to encompass the various opinions on rules/piece movements in some of the large Shogis. Where relevant I also included modern variants, like your own Nutty/Cashew/Macadamia series.

There is an AI, but it's basic alpha-beta and not optimised at this point; online play is also available. The Ai Ai software is a general-purpose game-playing package primarily built around MCTS, and the Chess engine is a relatively new addition, so there's still lots of work to do on that. The priority in the Shogi arena right now is having the large Shogis all functional and bug-fixed, then we'll work on the AI. I'm currently working on Taikyoku Shogi.

Reinforcement learning is coming, but not soon. Lots to do until we're ready for that.