What platform do you use to run multiple intraday strategies?

Discussion in 'Automated Trading' started by fan27, Oct 30, 2017.

  1. fan27

    fan27

    Yes....that is something I will be looking at down the road. Really what I am trying to do is determine which platforms I should target my service towards. I am working on a Machine Learning implementation that can test ideas really fast with the goal of being able to generate code which can run in an already built platform (i.e. NinjaTrader, TradeStation, etc).
     
    #11     Nov 4, 2017
  2. traider

    traider

    python for research
    something else for production
     
    #12     Nov 4, 2017
    fan27 likes this.
  3. Simples

    Simples

    In my limited view, no platforms perform the way I want them to, and I got tired of trying to shoehorn my ideas into existing designs that imposed artificial limitations. It's a learning curve and can be great for learning, but to go beyond, I think you need to free yourself and start from scratch.

    I see you're also doing it ass backwards. Trying to find a platform to target.. After the fact. Now it may seem like that'll be your answer in the end, but the process is exactly backwards of what you should be focusing on: Who are your users/customer base? What are their problems, what do they need? What are they willing to pay? I'm assuming you're basing your income on future clients, and not pure edge to be exploited privately in the markets?

    Yes, this is even more work! But may save you many dead-end attempts that lead nowhere other than as a learning tool. Also, if you can't face your customers, then you're not going to have much business relationship with them later either.

    When you know your customer, then you may discover they prefer certain tools, but what if they don't want to use those tools? Making great tools is more about people, than the tools themselves.

    If you make a backtester, you're already 70% there to make a live performing system. Also, how much do you want to marry external tools and their inner workings? Finding the right APIs might be preferable to finding the right platforms.
     
    #13     Nov 6, 2017
  4. fan27

    fan27

    I don't disagree about the limitations of existing platforms. What I am attempting to do is to solve a very specific problem which is finding profitable trading strategies efficiently. In my view, this is the biggest pain point for people looking to auto trade. I will have a prototype soon that I can get in front of actual users to confirm my thesis. My technology will be able to run independently of any platform...it's just that the final step will be to generate platform specific code. Will keep you posted!
     
    #14     Nov 6, 2017
  5. traider

    traider

    Why don't you solve the problem of finding strategies first? You might be surprised at how hard it is before you worry about the second part which is easy
     
    #15     Nov 6, 2017
    Simples likes this.
  6. fan27

    fan27

    I already have a prototype that can test over 1000 strategies per second, can discover strategy candidates on sample data and then further test strategies that have user defined characteristics on out of sample data. The project is not complete but I feel confident what I currently have is of significant value.
     
    #16     Nov 6, 2017
  7. traider

    traider

    How do you know what you have isn't just a huge data mining exercise?
     
    #17     Nov 6, 2017
  8. fan27

    fan27

    It finds strategy candidates in sample data and then retests those same strategies in out of sample data. If both runs have positive results then the likely hood of curve fitting is minimized.
     
    #18     Nov 6, 2017
  9. userque

    userque

    At this speed, I assume your prototype doesn't test machine learning algo strategies that can take much much longer than a second to generate one signal.?
     
    #19     Nov 6, 2017
    Simples likes this.
  10. fan27

    fan27

    I have a custom machine learning implementation. I'll post a more in depth overview tomorrow.
     
    #20     Nov 6, 2017