[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Band-its...?



PureBytes Links

Trading Reference Links

Ian,

>           Yes, that was my first thought - MAs and BBands but I can't 
>get an MA to cut as smoothly through such varying data. The Sigma bands 
>are possibly/probably BBands but the central 'trend line' as it's called 
>is the puzzle.

Well, it says the length period is 250 days.  Given that, did you see
what happens with a simple moving average, an exponential moving
average, or something more exotic like the T3 moving average?

I note also that visually, this trend plot seems like it may be
lag-corrected.  Lag correction is something one can do with very
little overshoot for data with a slowly-changing trend (short-term
market data doesn't qualify but long term might).  A double
exponential moving average (EMA) is an example of a lag-corrected
moving average, using a trend forecast to correct the lag:

Single EMA (what we're all used to): EMA = w*x + (1-w)*EMA[1];

Double EMA:  DEMA = w*x + (1-w)*(DEMA[1] + T)
             T = v*(DEMA-DEMA[1]) + (1-v)*T[1]

where x is the data to smooth, w is weighting factor for the
exponential moving average, v is a second weighting factor for the
trend forecast T.

Don't try to use this to make a lag-free moving average on
short-term data of a few bars; it overshoots pretty wide on
dramatic changes in trend.  I gather that Mark Jurik's JMA uses
this technique in an adaptive fashion, likely with a better-quality
underlying filter than EMA, to eliminate overshoot and obtain a
smooth curve with low lag.

-- 
  ,|___    Alex Matulich -- alex@xxxxxxxxxxxxxx
 // +__>   Director of Research and Development
 //  \ 
 // __)    Unicorn Research Corporation -- http://unicorn.us.com