100 % wins over 240 days

Discussion in 'Automated Trading' started by Spectre2007, Mar 12, 2016.

  1. Being honest is one thing. Being silly is another.
     
    #41     Mar 14, 2016
    Alex27 and dartmus like this.
  2. All the simulated result in the world worth less that one cent.
     
    #42     Mar 14, 2016
    Spectre2007 likes this.
  3. Yep, till I see it with forward testing not worth anything.
     
    #43     Mar 14, 2016
    indicator777 likes this.
  4. Simples

    Simples

    What you need may not just be infrastructure, but an unattended execution environment. If it's custom, it may need custom development and maintenance. Myself, I use after market-hours to maintain SL-triggers, so have no requirements for live operation. It's a very easy stepping-stone for automation, and leaves intraday for the junkies. For live, you'd need to investigate different options. It can't be easy though, because there's many things that can go wrong, from interpreting analysis-data to managing trade orders. I don't think there's any quick fix other than either do it yourself and weed out the inevitable bugs, or get someone to do it for you, either a partner or an employee. I know there are automation solution offered by some brokers / companies, but they're all very proprietary and specific, so might not be applicable to your requirements. It maybe worth checking out though. This is assuming you can't trust or monitor Ninjatrader. Might be options for that as well in there. Alerts etc., but it will be limited too and not so unattended as you'd like. If you're OK with it, maybe lease a server somewhere to run NT remotely?

    Tick data.. Is that data really coming in the right order, or can it be messed up somehow? Have you checked how you will get fills in that particular environment and what the cost of spread might be? Do you have a sound logic behind reasoning of why the trade should work, or is it a find that just show up in statistics? Have you tried the scenario yourself, or would you need to depend on automation to handle it for you?

    In short, part of your edge will be in the robustness of the unattended automation, which will require significant time to develop and refine. What might look good on paper, might have different issues in real operations that you would normally avoid manually.
     
    #44     Mar 14, 2016
    Spectre2007 likes this.

  5. Thank you for your reply,.. I think a retail platform that holds OCO orders on server best bet.
     
    #45     Mar 14, 2016
  6. botpro

    botpro

    Look in the mirror, dude!
    When developing and testing systems then simulated market data is the best.
    No need to pay for useless data which covers just a single one of the many possibilities the market can take.
    With simulated market data 1000s of market situations can quickly be tested at no costs.
     
    #46     Mar 14, 2016
  7. remind me to lay low next time ... one of my trading comps is hacked to death while connected to network.
     
    #47     Mar 14, 2016
  8. dartmus

    dartmus

    There's often large differences in tick data between different platforms. Just one more thing that has to be considered.
     
    #48     Mar 14, 2016
  9. U gotttttt nutttttiiiiiinnnnnnn, lol... Some partial code.. How desperate can you be to hack a comp..

    Doing a clean reinstall.. Of OS.. Windows 10..
     
    #49     Mar 15, 2016
  10. The secret to the 100% success of the above systems is .. If you wait long enuf, any random entry in the direction of the long term trend will go positive if given time to.

    If you look at above systems all they did was go long..the key is "the implied long term trend" and MAE, mean adverse excursion. If your MAE is too high most systems will be turned off or even if statistically you can group your MAE's to a cluster. A single MAE greater than the cluster can be protected against without compromising the take profit point. With a system protective stop loss.
     
    Last edited: Mar 15, 2016
    #50     Mar 15, 2016
    VPhantom likes this.