Hoy to all!

I need advice for NNs.

I wanna make a chess engine that uses an NN. (see collabs)

I want the NN to output a probability for each move.

However, i dont know what data to train the NN with.

If i train it with the moves from an engine, it simulates an engine and that sucks.

If i train it from games, i only see one move, and thus, cant train the other moves.

My thougt is, that maybe, i let the NN play an entire game and if it loses, it adjusts itself somehow. But how?

If you wanna contribute, email me :wittmaxi@outlook.de