RenTech: using artificial intelligence to morph one trading strategy into another

Discussion in 'Strategy Development' started by losemind, Aug 10, 2011.

  1. losemind


    RenTech: using artificial intelligence to morph one trading strategy into another

    Hi all,

    I met with a FX high speed trading guy (not necessarily super HFT) and he told me how they use artificial intelligence in trading ... and he hinted that RenTech is using AI in similar approaches...

    He said they don't use AI to forecast prices directly; instead, they use AI to morph one trading strategy into another according to market conditions, for example, morphing momentum into mean-reverting according to market regimes.

    I am guessing he's referring to using Genetic Programming... and I am guessing that the FX quant trading guys have the tendency of using GP or AI to morph strategies...

    Anybody knows how to do it? How might be the performance? Please shed some lights on me! Thanks!
  2. emg


    it is secret. nobody will shed a light
  3. AI????

    So a simple program that detects changes in market characteristics that YOU define and switches strategies accordingly is AI??? Erm, I don't think so.
  4. nitro


    My research is leading me in the same direction, although not necessarily with AI. I have thought about this for years and finally resigned myself that you have to have some sort of algorithm that changes based on regime. But this is not so easy - easy to say, not so easy to do...I know how I would do it, I don't know how they do it.

  5. jem


    Essentially you have to find what is working and pound the hell out of it. (quote to me from a very top of the line consultant who made a famous fund profitable)

    Is it really that hard to figure out.

    Lets say you have a pull back from the high strategies... a class of break out strategies. a class of vol breakout strategies and multiple time frames. Some reversion type of stuff and you also have really fast ones.

    You turn on ones which are currently making money in the market and when blended together reduce drawdown.

    Over time you develop sets of strategies which work but go into drawdown, and then you vol pump em.

    In the end your sill would be about blending and vol pumping and find those strategies which can handle your size. The strategies themselves can come from anyone who trades profitable and understands a running market vs a maintenance market.

    Thats what I would do... if I were trading enough money and had some programmers and tech guys.
  6. You can home-brew a faux version of this. It'll take some horsepower to see it develop in realtime, but it can be done.

    Take a very large universe of moving average crossover*** indicators. Very large. Like everything from 1 vs 2 all the way up to (1 vs. 200 ... 199 vs 200). I'm talking 1000s of them.

    Define going long and going short as the usual crossover in each strategy. Start keeping track of the running outcome from every combination. Plot the results in a kind of "heat map".

    Make up lots of simulated market data. Run it all through the contraption. View the heat maps.

    Pretty soon you will see how a trend develops across timeframes, and how it comes undone. Now you can write some heuristics, and voila, you have an "AI" to move you between market regimes.

    And you can take this to the next step - a running "second derivative" of the outcomes go into a second heat

    Etc etc etc.

    ***Moving average crossovers is just one example - there are a million ways to differentiate trend or churn - pick whatever you like best, it's all pretty close to the same in the end.
  7. poyayan


    I always think skynet will show up in the commercial arena first..:)
  8. schizo


    LMAO. Good one. :D
  9. I would say its a combination of multiple strategies say 1000 ? and the system's ability to learn and recognize markets and which strategy will work or not, and thus, the morphing and adaption.

    btw, if anyone had the details to tell u might as well ask them to give you $10 million to your bank a/c =).
    #10     Sep 4, 2011