RE: Re[2]: JMA (was: Hull Moving average)

["Adrian Pitt" <apitt@xxxxxxxxxxxxx>]

Thanks for the constructive response Bob. Are you using them as a substitute
for price to smooth other indicators as Mark suggests or in more traditional
ways? 

Adrian

-----Original Message-----
From: Bob Fulks [mailto:bfulks@xxxxxxxxxxxx] 
Sent: Friday, 12 January 2007 2:39 AM
To: omega-list@xxxxxxxxxx
Subject: RE: Re[2]: JMA (was: Hull Moving average)


At 08:51 AM 1/11/2007, Adrian Pitt wrote:

>I'm a bit confused.  While the JMA may have some complex maths behind 
>it, aren't we ultimately looking for an average (in whatever way we 
>choose to use it) that shows the least lag and the least noise, and 
>Mark Jurik has basically shown this to be the case with his JMA.

The T3 average is very smooth with a very good trade-off between smoothness
and lag. I use it and many variations of it of my own design. It is linear,
meaning that:  
 
   T3Average(A + B) = T3Average(A) + T3Average(B)

The Jurik average is adaptive in that the parameters vary based upon the
price values. This makes it respond faster hence less lag for comparable
smoothing. However, this makes the average not linear:

   JMA(A + B) <> JMA(A) + JMA(B)

Linearity is a useful property for some work.

But you cannot simply substitute either average for more commonly used
averages and expect your system to magically do better. You need to design
around them from the start. 

I always use better averages than the traditional "simple moving average" or
"exponential moving average". There is no comparison, in my opinion.

Bob Fulks

-- 
2:52 PM

-- 
3:33 PM