Fast OR Slow ?

Discussion in 'Technical Analysis' started by TSOKAKIS, Mar 28, 2003.

  1. Here is my basic thought on the well known delay problem.
    A fast [and smart] indicator will catch the trend ASAP, but it will leave the trend prematurely.
    Another slower [and more wise] indicator will loose the early birds, but it will be rewarded with the delayed exit and enjoy the sweetest part of the movement.
    Both methods have their plus and minus.
    The ideal combination of two [or more] speeds would include all the advantages in one strategy.
    Cond0 is a fast trend detector and gives an entry at A0 and follows untill B0.
    Cond1 is a slower detector, the entry is +1 delayed at A1, but the exit is +3 delayed at B1.
    Which is better ?
    There will be many opinions, but any conclusion is related to the time frame the market moves.
    Trends duration is not always the same and, unfortunately, nobody knows it in advance !!
    Just after an 1-bar breath [the white dotted line], the market decides to move higher again for 6 bars.
    My slow indicator gives now a quick entry at A11, the "fast" is not that fast anymore and begins at A00.
    [Fast and slow are not absolute words, as you probably agree...]
    Both indicators agree for a final exit at B00.
    What is the optimal solution ?
    We often read in various T/A books or articles that this is the destiny of any T/A attempt.
    They seem to agree that the "nature" of an indicator is covered by the postulate "fast in/fast out OR slow in/slow out".

    I do not like it and I will try to break it.

    I need a smart trend detector and a smarter way to COMBINE the fast and the slow rate.
    Do I ask something unrealistic ?
    I would appreciate your opinion.

    Don't over analyze your strategy.

    Trade it...real money (reduced size) or via a realtime simulator (normal size).

    After a few months of this...

    our opinions won't matter because you'll have realtime results.

    More'll develop a feel or intuition about your strategy in certain market conditions.

    Make sure you document well all the key market internals for that day because the market will often repeat situations at a later date...

    having a good memory via reguraly reviewing your trading notes about that particular strategy...

    will be an exploitable edge.

    Once again...don't over analyze it because you already seem to have a good grasp about the workings of your strategy and our opinions truly is unimportant especially after you have realtime results.

    P.S. If you already have realtime trading results about this strategy...

    why didn't you post it?

  3. Every indicator has a look-back, large look-back gives a smooth indicator but gives late signals, late in...late out. Check out this site, years ago he was the leader in low lag stuff, plenty of info there. Hope it is helpful.
  4. Do your thing; you can figure out stuff it looks like.

    Here are three aids for you:

    1. Get on the correct chart.

    Do a yellow brick road for NQ. (Attached). See where your efficiency is and see where the market efficiency fails. You can redo all the data entries to see how much money you make. You look like you are running slow over all.

    2. Make use of market data that is there.

    By not using volume, you denign yourself a leading indicator of price. It is all a good thing to corroberate any signals you use as well.

    3. Look ahead occasionally.

    If you wish you can always look ahead to see what is coming. Once you have calibrated your indicaotrs and their respective signals to the participants in the market, the setting work throughout the range of possible coonsideration (on any acceptable candle duration.)

    You are making one fundamental error vis a vis appreciating capital if you view making money by applying it to a variety of applications. Rotating capital through multiple applications has different entry/exit criteria than you are using. For your presentation, skipping this aspect is practical for now.
  5. Some uptrends are sustainable for more than 10-20 days.
    A fast trend detector, like DEMA(StochD(40),20) would buy at B0 and sell at S0.
    The very next days it would force the trader to follow the undesirable whipsaws W1, W2 and probably loose
    a part of the B0/S0 profits.[see the black line]
    A slow trend follower like DEMA(StochD(40),50) would start at B1[missing the interesting first 7 bars]
    and stop at S1[6 bars after the actual end of the trend]. The use of the slow period helps to avoid whipsaws but expose the profits
    to a probable dramatic decrease after the actual peak. If the start of the trend was sharp with the well known gap ups
    then the remaining profits would not be similar to this great trend.[see the yellow line].
    The technical analyst wish would be
    1. to begin ASAP,
    2. avoid the whipsaws and
    3. get out as closer to the actual peak.
    It is the basic T/A problem with almost any indicator and here is a solution proposal : the variable smoothing.
    Begin with the fast indicator and, as the trend goes on, increase smoothing factor adding +1 delay per bar.
    If the trend holds for 20 bars, the beginning would be per=20 and today you would run a per=40.
    If the trend stops for any reason, sharply back to per=20 to begin another, bearish cycle.
    The result is the ENTRY/EXIT of the att. ^NDX gif
    For more details here is the AFL* solution:
    Paste in your IB the

    /*TREND DETECTOR WITH VARIABLE PERIOD, by D. Tsokakis, Apr2003*/
    t=20;// fast period
    Plot(x,"",1,8);// fast trend detector
    tA=50;// slow period
    Plot(xA,"",7,8);// slow trend detector
    Plot(50*Cond1,"",5,2);// the start of the fast uptrend
    Plot(50*Cond2,"",4,2);// the end of the fast uptrend
    k1=BarsSince(Cond1);// uptrend bar counter
    k2=BarsSince(Cond2);// downtrend bar counter
    t1=IIf(k2>k1,t+k1,t+k2);// the variable period
    x1=DEMA(StochD(40),t1);// the variable trend detector

    The method accepts further calibration if you replace t1 with
    coeff=0.5; // delay calibration, set coeff from 0.5 to 1.5
    t1=IIf(k2>k1,t+coeff*k1,t+coeff*k2);// the calibrated variable period
    according to your needs.

    *AFL, Amibroker Formula Language from
  6. devadas


    This afl is giving syntax errors .Can anyone rectify it ?
  7. prash454


    The corrected and complete AFL is attached herewith.

  8. devadas


    Prashanth, Thanks for your post