Trading based on patterns

Discussion in 'Strategy Development' started by ProgrammerGuy, Sep 18, 2007.

  1. IMHO, I have always thought of basing strategies on Triangles, H&S, double bottums was complete BS. None of my profitable trading models has anything to do with this.

    But because I'm always looking for new ideas, I'm willing to give it a shot, program the thing, and backtest to see if I can produce a profitable enough model to trade. I was wondering if anyone anyone has developed any strategies based on this "garbage"?

    Thanks
     
  2. There have already been studies and books published on this, and it was not very promising...
     
  3. MGJ

    MGJ

    Might look into "Breakouts from Congestion" patterns where congestion is something like (unusually narrow Bollinger Bands) or (horizontal sloping Donchian Channels) or (abnormally low fractal efficiency) or (ADX has been falling for a very long time).
     
  4. There are probably three things you can do to well exceed the CO (Conventional Orthodoxy) testing results.

    1. Front run the triangle BO's.

    2. Break the H&S into 6 parts.

    3. Break the double bottoms and tops into 4 parts.

    The general schema for this is to use bit masks containing data sets. Always include the indicator leading signals of price and do this in pairs of vector binary data. By incorporating Linda's three count AND a multiple fractal view of this, the pairs will appear.

    If you read Joe Ross are able to throw away his orientation on trend and congestion stuff (and particularly the ratios he banties about), then you get to see that making money is always available since you are continually detecting patterns instead. Patterns ripple from fastest fractal to slowest fractal and then become undetectable.

    Use Kahn's (See Barron's web pages; he is editor) "sizing" recommendations to be able to "detect" by using the bit masks.

    It is very strange that the financial community hasn't ben able to capitalize someone like you to get the job done scientifically. They didn't it turns out. What they did finance was dead end studies by people who knew nothing about the markets. That happens.

    I guess there is a lot of computer science and IT oriented work that has surfaced. The first requirement is to "feed" these guys tasking assignments. All tasking must come from modelling. And, More important, to get a checker on their production to verify that they actually did the development of the modelling.

    Here the key to doing the check out is making the model "roll". I think it is now called real time testing. But the fact is that this approach will not checkout the development since the guys using the real time testing do not know how the markets work up to this point and they are short on the coding verification techniques.

    You are correct that what is viewable by the public and the people who paid out the money is "garbage". People believe the "garbage" and do not throw out the "garbage producers". These believers have earned the consequences.

    Where I you I would skip doing the work since you would only back test it. At some point you may wish to set up a procedure for testing your coding instead.
     
  5. ggoyal

    ggoyal

    forget all that. just look at the trend. look at the immediate trend and look at the trend of the past few days. im sure you r a short term trader. so look at trend and ride it.

    now obviously you have to look at other things like the lowest support price and when it looks like it has gone up or down too much.

    gettin an idea?