Genetic algos

Discussion in 'Automated Trading' started by angrytrader, May 11, 2022.

  1. I am going to give this library a go:

    Does anyone have experience with genetic algorithms? I have a feeling that until I find the right "question", the answer that I will be trying to evolve to can mislead me pretty easily.

    Apart from being really fancy at beer gatherings when someone talk about it, are these genetic algos providing any of you any edge over fail-and-repeat good'ol backtesting?
    SimpleMeLike likes this.
  2. rb7


    I've tried GA like 27 years ago. I wasn't able to find any useful edge with them.
    I still have my app that I've developed along with NNs.
    I was also using them to optimize my NNs.
    I'd say that they are (GA) a good way to curve fit a strategy.
    jetbird, Craig66, ET180 and 1 other person like this.
  3. ET180


    Not really. I spent a while working on genetic algorithms, genetic programming, and other metaheuristic approaches about a decade ago. Given enough free parameters, it's easy to find some settings that fit whatever data you have wonderfully. The problem is that it won't generalize to new data and you end up curve fitting. If I was going to go down that path again, I'd look at creating random forests of simple algorithms.
    rb7 and angrytrader like this.
  4. I did have the recommendation to go for Random Forest algos from someone else, and I also had the comment that GA might help to optimize an existing algo that came from NNs.

    Very interesting, I should then switch to Random Forest as a first attempt.

    Thank you!
  5. stochastix


    Nothing besides stochastic models will work, appropriately specified
    angrytrader likes this.
  6. Stochastic processes might fly over my head at the moment. Unless I get into some deep reading before I code anything I won't be able to hit the right track.

    Can you recommend any books on the subject?
  7. stochastix


    The v
    The volatility surface by Jim Gatheral , . It's hard for me to recommend any because I started out with measure theory and integration by MM Rao 15 years ago . The options and futures book by John Hull is a classic by a master. I know a guy who helped him publish the book awkwardly during the crash of 2008
    angrytrader likes this.
  8. stochastix


    If you want some tips I spend most of my time alone wouldn't mind having someone to help develop this volatility trading model with. It's hard to psyche myself up for it because I've done most of it in my head so far . I've mostly implemented the quadratic rough heston model via Monte Carlo simulation but I'm working on fourier pricing now
  9. I see, I can definitely help with the coding bit but I will have to follow your lead on the model for a while. If you don't mind having someone at a learning stage, I can help.
  10. stochastix


    I don't mind. Do you do Java ?
    #10     May 11, 2022