It's not like I'm abandoning the decades of development. I'm starting with a fork of Stockfish 10(pre-NN, and using too recent a version could make it hard to submit it to TCEC due to rules about similarity to existing participants.) and I've been digging into the code and algorithms therein for months now. My approach is an extension of existing techniques.
I actually started with a greenfield implementation, but found myself reading more and more Stockfish code anyway as it's an excellent reference for the state of the art. And I realised it was far too much work to reimplement all this stuff, so that's why I switched to a fork.
And I wanna learn linalg and deep learning anyway.
I actually started with a greenfield implementation, but found myself reading more and more Stockfish code anyway as it's an excellent reference for the state of the art. And I realised it was far too much work to reimplement all this stuff, so that's why I switched to a fork.
And I wanna learn linalg and deep learning anyway.