Seeking feedback on my rules-based system

Discussion in 'Automated Trading' started by morganpbrown, Apr 12, 2019.

  1. IAS_LLC


    Fortran 90...I used to use that a minute ago when I was working on "bleeding edge" aerospace products...u aerospace?
    #11     Apr 13, 2019
    morganpbrown likes this.
  2. Thanks, and a good point. Most of the ETFs that I actually trade have only been around for 9-10 years, as you noted. However, during the model training phase, I will actually analyze securities that have been around longer. For instance, I train the model using ES futures (which go back to 1997, I believe), but actually trade UPRO (around since 2010?).

    I'd certainly prefer to be backtesting the actual security that I'm trading, but I would rather trade a slightly flawed model with a richer set of statistics than vice versa.
    #12     Apr 13, 2019
  3. Haha, definitely a lot of F90 in your community! My day job is actually in energy.
    #13     Apr 13, 2019
  4. danielc1


    Control of past data is not an indicator of future results. If you get results with this, it is not because of what you think you are doing. It will be because of your money handeling strategies with winner and losers.
    #14     Apr 13, 2019
  5. For the S&P, I analyzed 103 indicators, testing 2615 intervals. From this, I extracted 150 testable strategies and chose the "best" 6 that I'm currently trading. So I'm definitely spanning a large model space, but I'm quite sure that the correlation between many of these strategies is pretty strong. Ideally, I'd like to only trade strategies that are sufficiently "different". Other than looking at the backtest graph for each one, I haven't put any meat on how to actually measure "different".
    #15     Apr 13, 2019
  6. I have to say that I'm humbled that you guys would take the time to read what I wrote and provide this thoughtful feedback. I'm just a middle-aged guy fumbling in the dark through a sea of bad/disinformation. It's good to see that I'm not totally lost, but I'm still waiting for the other shoe to drop. As my wife said, "a lot of smart people are trying to solve this problem, why are you better?"
    #16     Apr 13, 2019
  7. Just to be clear, I wasn't laughing at the choice. Just saying that a specific personality would choose F90 and that seems to be you very well.
    #17     Apr 13, 2019
    morganpbrown likes this.
  8. The software language is just one of the many tools you use to get to the ultimate goal (i.e. make some money). It doesn't matter which tool you choose, as long as it helps you to achieve the goal.
    #18     Apr 13, 2019
    morganpbrown and nooby_mcnoob like this.
  9. Sounds like a variation of that old canard, "Risk management is the only real edge in trading". I'm sorry but but this is just untrue and shows a fundamental misunderstanding of trading. There are only 2 ways to make money in trading: 1) Luck, 2) Having an edge. It is a mathematical fact that no "money handling strategy with winners and losers" in the world can expect to make money trading a zero mean random walk.
    #19     Apr 13, 2019
    morganpbrown likes this.
  10. Sprout


    Consider transitions between ‘states’ as events. The market functions as a sequence of events. Use this sequence of events as the independent variable and time as the dependent variable.
    The sequence of events signal both entries, exits, early entries, early exits, waits and reversals.
    Every event has what came before, the event, what must come after.

    ‘Must’ = understanding the price volume relationship. In a nutshell, Volume leads price.

    How one goes about defining events makes all the difference.

    The challenge is to turn noise into signal, and what is/are the prerequisite/s necessary to accomplish this.
    #20     Apr 13, 2019
    morganpbrown likes this.