Best Trending Timeframe

Discussion in 'Technical Analysis' started by Corso482, May 6, 2003.

  1. What's the timeframe that trends the most? I guess the answer to this question changes over time. But how would you quantify the answer if you wanted to know today?

    For those who use trend-based systems, answering this question is essential.
  2. Two possible ways come to mind:

    1) Measure the percentage of time ADX is above 25 (pick a number) and increasing for each time frame.
    2) Calculate the effiency ratio on each time frame, taking net movement and dividing by the gross (up and down) movement. I think Kaufman did some work with this when looking at variable MA's.

    I did the former study eons ago when I traded Fidelity Select mutual funds. I wanted to focus only on the 'trendy' funds. The results sure seemed to match up with the visual and trading feedback I got.

    Good luck.
  3. *******************
    Corso 482------------------3 points, none random;

    [1]Part of it depends ,in a practical sense,with the personality of the trader. I usually didn't hold any as long in a down trend bear market since profits are given and TAKEN or lost faster.

    [2] Usually hold longer in an uptrend [like today !]] since patterns , profits may be more orderly.

    [333] Good but complex question. Looking on :cool: time frames [plural] from 30 minutes to 9.333 + years .Whether you figure in a sideways trend , a 20 day moving average or not; it pays me to include longer time frame charts.
    ''That which hath been is NOW ; and that which IS to be hath already been '' Solomon-Ecclesiastes 3;15
  4. These are hard questions. I would set up a trend following system on the time frames I was comfortable in assuming risk. Then, I would compare on a point basis the P/L from these periods divided by the number of trades, giving me the average gain per trade. Next, I would divide the P/L per period by the last price of the instrument for the day. Adding these two numbers together gives you the 'score' for that instrument for each time frame. Finally, divide the 'score' by the total cost of trading resulting, typically, for stocks in a very small fraction of 1%. Plotting this number on the x-axis of a scatter chart along with the number of trades to achieve it on the y-axis produces a picture like this. It then becomes possible to filter out the most profitable system results for various time frames and/or instruments. The picture looks at the results of a trading system applied to the following stocks and is a couple of years old: AMCC, AMZN, BEAS, BRCD, FLEX, ITWO, JNPR, NXTL, OPWV, PMCS, RFMD, SANM, YHOO. Better results are toward the left-hand side of the picture and toward the top.

  5. Here is a second picture showing the best performing stocks in that group based upon the parameters discussed in the previous post. They are BEAS, JNPR, PMCS, RFMD, and SANM.

  6. As there is a little interest in this, I want to finish my thoughts toward setting up a way to measure the trend. I've included a picture by way of explanation. G/L is the same as P/L. Sc/Cf means: Sc(ore) divided by C(ost) f(actor). The instruments can then be filtered using Excel's Autofilter feature on the last two columns of the spreadsheet. However, with a little thought you can adapt this basic design to accomodate different time frames using just one instrument. To create the chart use the last two columns, create a scatter chart connected with lines and reverse the y-axis to the right side.

    One final thing: in the first picture I posted I neglected to include the last row of the spreadsheet in the chart. That's why you can only see twelve dots in the picture -- YHOO was inadvertently left out although its exclusion didn't change the overall picture much and didn't change the actual final outcome at all.