How do I create an engine that plays like me?

Hi,

I want to create a chess bot that plays like me (minus the blunders).

I have downloaded a lot of my games and I filtered out variants. I'm planning on using stockfish to check if my move was a blunder or not, but maybe it won't play very humanly if I do that.

Also, because I'm not a chess engine, I don't have an evaluation function. So how can I make my neural network adapt to the moves I make? I have read a bit about move adaptation on chessprogramming.org, but it's still unclear to me how to do it in practice. Also, I think it matters if the selected move has the lowest cost by just a little bit, or if it crushes all other moves completely.

I have another problem: because I have less than 1000 usable games, even with data augmentation (white/black mirroring, flipping board if no one has castling rights) and filtering out blunders, it would be only tens of thousands of samples. Also, i have to sacrifice some of the training data to test my model. So what would you recommend when training with a small amount of data like this, to prevent overfitting?

it might be possible if ya code an engine and make it play da openings ya play but i doubt its possible but ya never know

Adapting the openings to mine will be easy. Just using an opening book of my creation will do the trick.

I am thinking of doing something similar for a class I am taking. I am looking into using auto encoders.

What are auto encoders?

@Checkmate6659 autoencoders are a class of unsupervised or semi-supervised neural networks that attempt to reconstruct your inputs as outputs.  Essentially, you try to feed them data and they in turn try to output what you fed them. If you trained an autoencoder on a sequence for input it might be able to output a likely follow up

the problem is that it could spit out an illegal move

what im imagining is that the nn gives a cost for every successor of the position and takes the one with the lowest cost

we could very well take the positions for white only when training, and if the nn is playing black then we mirror the board before passing it in to the nn

"I want to create a chess bot that plays like me (minus the blunders)."

Then it will not be playing like you.

i want it to have my playing style basically

but i don't want it to blunder

I'm no computer expert, but in general, maybe get a computer, load Stockfish and then throw the computer out the window.