Will this continue to work?

Discussion in 'Strategy Development' started by Bowgett, Jul 14, 2005.

  1. Bowgett

    Bowgett

    This is backtesting result of trading on IWM for year+. The problem is I found a bunch of similar strategies with not a bad returns but I need to pick just one to go with it with real trading. How do you pick a strategy that not only worked before but will also potentially continue working in the future?
     
    • iwm.jpg
      File size:
      68.2 KB
      Views:
      374
  2. oh, please, where do I send my check?
     
  3. Bowgett

    Bowgett

    I had situations like the one I attached. Eventually performance recovers but it sucks right after you start live trading. Have anybody tried trading strategies that performed poor recently but good before?
     
    • qqqq.jpg
      File size:
      74.9 KB
      Views:
      226
  4. monaco2

    monaco2


    lol ap
     
  5. This strategy testing thing is one of the best gimmicks they have come up with yet.

    If you throw enough strategies into a backtester you are bound to come up with something that looks great in retrospect. The law of averages. Talk to Nitro or Abogan or some of these other guys who work with systems for a living and see whats really involved.
     
  6. fader

    fader

    i have gone through a backtesting exercise recently - your equity curve could be (a) a random result (b) curve-fitted - if it is either of the two, there is no reason why it should continue to work - what you need is to test that your curve is not (a) or (b) - there are threads here on ET that discuss this - all the best.
     
  7. Bowgett

    Bowgett

    It is not a random result because there are more like that one with similar parameters and returns but I am afraid that some of them are curve-fitted. I will search for curve-fitting on ET. Thanks for help.
     
  8. Will this continue to work?
    If you have to ask, it won't.
     
  9. Bowgett

    Bowgett

    What makes you think so?
     
  10. vikana

    vikana Moderator

    Will it continue to work?

    Without any details, who knows.

    Here's my answer: if you have more than 3 parameters, probably not. With 2 or 3 params, maybe. Less than 2 would be promising.
     
    #10     Jul 14, 2005