When a system spectacularly fails

Discussion in 'Strategy Building' started by jcl, May 8, 2012.

  1. ssrrkk

    ssrrkk

    What I meant by "signal" is the useful information in the data as opposed to the noise. What I meant by long term signal is this: a finite support signal can only contain wavelengths up to the window length and will lack frequencies lower than that. So a low pass filter on a windowed signal really is a band pass with the lower frequency cutoff dictated by the window length. A sudden loss in profitability could be construed as a change in the characteristics of the price distributions (non-stationarity) OR it could be that there were longer term signals that were missed in the original time series due to the finite training window length.
     
    #11     May 9, 2012
  2. "Simple trend following" doesn't work...
    But it sells boatloads of sexy, glossy color books to idiots.

    If you cannot explain...
    Precisely why your methodology makes money...
    Exactly what gives you a Competitive Advantage...
    Against experienced traders at all the major banks, etc...
    Then you have nothing.
     
    #12     May 12, 2012
  3. Indeed it does, if you go where no one else is. You haven't seen everything.
     
    #13     May 12, 2012
  4. jcl

    jcl

    Simple trend following does certainly work, however not with the strategy from that book, which is a little "too simple".

    Anyway, this is the same strategy with a switch-off mechanism such as proposed by logic_man:

    [​IMG]

    The blue curve below the main window is the lowpass filtered weekly profit. It it goes below zero, the strategy is switched in a mode where trades are not executed, but just simulated. This happens in 2011. From that point on the equity curve is a flat line until the strategy becomes profitable again, or until eternity, whichever happens first.
     
    #14     May 13, 2012
  5. ssrrkk

    ssrrkk

    Still couldn't avoid the huge drop PL in the fall of 2008...
     
    #15     May 13, 2012
  6. Markets do two things.

    They mean revert, and they trend. This means by default you have to have two systems running all the time. The mean reverting system is the dominant one. Which, by the way, is one reason why most traders lose, they think it is the other way around.

    Now when the market starts to trend, the mean reverting system must be shut off. The question is, when to do it?:)
     
    #16     May 13, 2012
  7. +1

    boatloads. and they don't even have to be glossy.
     
    #17     May 13, 2012
  8. Technically, you don't have to beat the "experienced traders at all the major banks", you just have to beat the other losing traders to whom those bank-based traders will be selling or from whom they'll be buying.

    The niche between the "smart money" and the dumbest "dumb money" is profitable. Basically, you take a piece of the middle 50% of a trend. Do that enough times and it adds up.

    Here's my take on it:

    http://www.elitetrader.com/vb/showthread.php?s=&threadid=241147
     
    #18     May 13, 2012
  9. Well, if 70% of trading today is Algorithmic...
    All that, by definition, is done by very advanced, well capitalized people...
    Everyone working in that space is paid well into 6 figures...
    These guys have advanced degrees (you know, the smartest guy in your class 10-20 years of HARD work later)...
    They are not reading slick nonsense about "simple trend following".

    When you add more traditional expert traders and hedge fund managers...
    Your "dumb money" might shrink to 10-20% of all trading.

    Basically, ET is Dumb Money Central.

    But it's not so bad...
    You don't have to BEAT Smart Money head-to-head...
    You just have to know how to AVOID direct trading against it.

    It's like poker...
    The most basic skill imaginable is picking an "easy table"...
    Which most people don't seem to realize is also a BASIC SKILL in trading.
     
    #19     May 13, 2012
  10. No, I'm highly specialized... which you MUST be to succeed in this business.

    Theoretically, since EXECUTION is 90% of the ball game...
    If you are a great trader with fabulous infrastructure...
    You can probably make money with entirely random entry points...
    But that would be sub-optimal.
     
    #20     May 13, 2012