Choosing mechanical systems

Discussion in 'Automated Trading' started by prophet, Aug 17, 2003.

  1. prophet

    prophet

    Which is preferable to trade?

    A system that enhances existing market trends, but will require a secondary system (discretionary or mechanical) to determine it’s direction:
     
  2. prophet

    prophet

    … or a system that determines the longer term direction, but has higher shorter term risk:
     
  3. prophet

    prophet

    … or a system that does well in some situations, breakeven (or doesn’t trade) in others:
     
  4. prophet

    prophet

    Or all of the above, trading all systems simultaneously, allocating equity between them?

    Or should I simply work longer to develop a system that combines the strengths of these three? This may be possible, though very difficult. It’s taken six months to develop these three. These systems are purely single market (NQ currently, ES possible), and use proprietary indicators, competing subsystems, and trained networks (constrained linear least squares).
     
  5. Allocate ...
     
  6. Dear Prophet,

    You raise some good questions. Of the three systems you propose none seem to be satisfactory (sufficiently profitable) if I understand you correctly. So the first question you should ask yourself is whether they truly qualify as valid "systems"?

    In my experience, the combination of three "poor" systems leading to a sufficiently profitable system is a very difficult task. It may in fact be much more difficult than the time it took you to arrive at the present systems.

    Don't give up though! But be prepared to put in 6 * 6 more months if necessary.
     
  7. prophet

    prophet

    Thanks!
    I agree. Mechanical or discretionary?

    I realize that allocating capital between systems can be itself implemented as a mechanical system to be optimized and tested. That first system I showed might do really well with a mechanical performance trend determination.

    Because my systems already employ competing subsystems, I’d end up with three layers of systems, and the implementation and testing headaches grow exponentially! Being only one person, I’m concerned if my efforts are better spent improving the systems I have (say producing one great system) or trading/allocating equity between multiple systems.
     
  8. PetaDollar

    PetaDollar Moderator

    I'd go with this one on multiple markets.

    This way you only worry about one system,

    and the markets might take turns printing money for you.
     
  9. prophet

    prophet

    I’m not satisfied because I feel I should be able to combine the advantages of these three (and more) without having to add extra complexity, over optimization or trade multiple markets. I want a system to work well on many markets. However, I don’t want the single-market trading risk to be so great that I am forced to trade multiple markets.

    These systems are works in progress. I suppose their validity depends on statistical significance of the methods and one’s risk preference.
     
  10. PetaDollar

    PetaDollar Moderator


    The system will capture market action that appears which it is designed to capture.

    We cannot help or make that market action appear, no matter how badly we want it. An we cannot design a system which captures all market action.

    A system which is b/e or out most of the time is perfect to trade on multiple markets for this reason.
     
    #10     Aug 17, 2003