Discussion in 'Trading' started by vanilla2, Jul 23, 2003.

  1. I'm very interested in bringing signal processing techniques into my strategy writing. It's more of a mid-term goal for me since I know very little about it right now.

    I find myself often trying think of ways to quantify a dominant frequency in the instruments I'm trading. I currently trade a mech system on two timeframes, and on days like today when both are out of sync I think it would be interesting to trade a spectrum of four or five frequencies and perhaps use a dominant frequency calculation for money management purposes. If I could weight my net position from all respective frequencies toward the dominant one, I might squeeze some more expectancy out of my system.

    I've done some preliminary random web searches for formulae to start out with, but usually what I find is extremely dense material that I don't understand. Can anyone reccommend an entry level text in the DSP field that could get me started on this?

  2. I'm sorry for blatanly bumping this thread to the top of the stack, but was hoping somebody may have a suggestion for introductory reading on this topic. Thanks in advance.
  3. xygeek


    Doesn't amplitude figure into this somehow too? :p
  4. From a completely lay point of view, I would think so, yes.

    One rudimentry way I was considering measuring this would be to track the distance between pivots which is basically measuring amplitude. I think this is oversimplified however, because pivots of various strengths are happening all the time.

    Would it make more sense to measure the distance between the most recent high and low pivots within a spectrum of pivot lengths and somehow weight which frequencies are increasing, and which are decreasing in a given time period?

    I believe the proper way to measure dominant frequency would incorporate amplitude in a much more sophisticated way than I mentioned above.
  5. xygeek


    When I said amplitude I was actually picturing direction and degree of trend slope. I think phase is the dsp term for direction. I'm quite interested in the degree of trend slope but don't know if there's a dsp term or function to measure it. It's my opinion the frequency aspects of dsp don't perform well with market prices.

    You mention a spectrum of pivot lengths but it's not clear if you meant pivot strength. Regardless I'm leaning toward lengths that ignore pivot cycles but a spectrum of lengths is appealing. I like, 1, 2, 3, 20, (plus maybe 10, or 50). It's dependent on the timeframe. But what functions can I use them with?

    Here's a book you can download. ... for free.