the best moving average

Discussion in 'Technical Analysis' started by stephenszpak, May 4, 2008.

  1. Hi

    Since this subject has certainly been covered, maybe
    someone has some links to threads or whatever.

    I'm NOT really interested in the best time interval to use for a moving average.

    I'm looking for what the best moving average actually is.
    I guess it should be very smooth (to eliminate false triggers)
    yet track the data quickly when it moves quickly and slowly
    when it moves slowly.

    pre-smoothing of data>??

    A long long time ago I was playing around in wealth-lab
    and seemed to determine that when using a time period
    of 40 or 50 bars or so, the only pre-smoothing that could
    be done that would not introduce delay was a 2 period
    weighted moving average. Even a 2 period simple introduced
    delay. I was analyzing the charts visually.

    I guess it has to be VIDYA (applied to pre-smoothed data).
    use a short time frame and transfer to a longer>??

    Does it make sense to apply a very long moving
    average (perhaps 200,300,500 whatever)
    to 5 second or even 1 second bars
    and drag and drop this onto
    bars, of let's say, 3 minute. Wouldn't this result in a very
    smooth average?

    Can all this be done visually, by trial and error?
    Can it be done in Ninja (or perhaps QuoteTracker)?
    How does one know how far off from 'perfect' (hee, hee)
    one is when done?

  2. The field of digital signal processing seems to make use of many of the more advanced smoothed adaptive moving averages.

    A proprietary one is the Jurik moving average. Other free ones to consider aside from VIDYA are a gaussian moving average and also Alan Hull's moving average.
  3. 50 period ema
  4. Yes.

    You seem to want maximum smoothing for minimum lag. In which case Jurik is best followed by either Hull or T3 depending on what you like. Jurik costs and the other two are widely available.
  5. Thanks for those that mentioned the Hull moving average.
    It might be shown below in an attachment.
    • hull.png
      File size:
      49.9 KB
  6. If the simple and the Hull are compared for the same
    time period one can see the Hull is more responsive
    yet also causes more false positives. Therefore one
    has to increase the time factor of the Hull. I'm not sure
    if it will always be this way or not, that is, visual
    confirmation is necessary.

    One wonders how to make progess. There are an infinite
    number of smoothing and data manipulating processes.
    It's hard for someone like my self, not a math person, to
    even know when progess is being made.

    Thanks for the replies though.

  7. i always like anything that can be divided by 8
  8. #10     May 4, 2008