Self adjusting Indicators

Discussion in 'Strategy Building' started by Guitar-Fire, Nov 21, 2003.

  1. Hi everybody,

    I' d like to ask you, if there's somebody who has an idea how to create a self adapting indicator.

    What I mean is, if you use an simple indicator like moving average crossover to create a pure reversal system, then you'll find certain times where it works and others where not.

    I believe the problem is the fixed parameter used in those indicators. They only work if they fit to the current market behaviour.

    Hope I made myself clear, any suggestions or questions ?

    Regards, Ralf
  2. Have you heard of KAMA kaufman adaptive Moving Averages. I suggest you read Perry Kaufman and Tushar Chande books. There are many ways of doing this.
  3. Ehlers could be another one. :confused:
  4. Yes. I missed that.
  5. CalTrader

    CalTrader Guest

    Actually, the best bet is to read a good book on applied statistics and program your own indicator. Its easy to design an "adaptive" statistic. Usually the "adaptation" is related to eliminating outlier data that once removed reveals a more accurate picture of the statistically significant information. That is, information that is statistically significant to you.
  6. gms


    The closest idea to what you may be seeking could be to assess each market's cycle length, then use that length as the parameter.
  7. thanks for your posts.

    gms: do you know a mathematical method to find the cycle-length ?

    easyguru: I tried to use adaptive MA, but I feel they're of very limited use

    Regards, Ralf
  8. ramora


    For TradeStation, Clyde Lee at has some very good tools that are adaptive and use cycle length. His work on Hurst is very good. There is a lot of info on his site even if you are not a TradeStation user.

    I have no conntection with Lee other than I respect his work.

  9. gms


    Draw channel lines on a chart (highs and lows) and see what the width is (points wise), if it's substantial in that you'd be able to grab a profitable piece of it then note the length of time it takes from trough to peak. On further thinking, your parameter may be better if you make it half the cycle. Backtest to see.
  10. They


    If you do you can play with rccameron's Dynamic Bands or Variable Bands.

    They adjust as the day develops - 5 min chart starts out and the period is set to 1. After the second bar is completed the period is set to 2, etc...

    It contains five different bands (bollinger, stoller, keltner etc.) and works off sma, ema, wma, lr, etc.
    #10     Dec 11, 2003