Chess Programming

15 July 2020 Hobbies / Career and Technology 2 min read

I’ve recently starting trying to put together my own chess engine. Obviously on some level I knew how complex a task this was beforehand, but I still didn’t anticipate quite how many difficult problems would need to be considered. There is;

Board representation, either piece centric or square centric. I am most likely going to use the latter, but then there are considerations of which model to use, with a choice of 8x8 Board, 10x12 Board, Vector Attacks, and 0x88.
Actually implementing the rules, which sounds easier than it is. 50-move and threefold repetition draws require some persistent knowledge of previous positions in the game.
Move generation, which can be legal, pseudo-legal, chunked or staged.
Evaluation… insert my whole knowledge of the game here. And I have to work out how to teach that to some silicon.
Search, which can be depth-first, alpha-beta or best-first (including the very fashionable Monte-Carlo tree search). All of these have a bunch of different sub categories, and most of them have ludicrous memory requirements.
Connection to GUI, either through implementation of the Universal Chess Interface or the Chess Engine Communication Protocol. Then I also have to pick a suitable GUI, or write my own.
Tablebases and opening books - to at least make it a real engine.
Performance considerations, because C# wouldn’t be anybody’s first choice of language to write a highly-performant chess engine in. The language is… slow… so some kind of optimisation is a must, and that is doubly true given it will be running on my local machine and I’m not going to give it half a terabyte of memory to play with.

It’s going to be difficult, but these are all super, super fun problems to solve. For me, anyway.