|
PureBytes Links
Trading Reference Links
|
Just some interesting information about the memory effect of
exponential moving averages.
Say a simplified version of XMA is defined as,
XMA = Price x Factor + XMA previous x (1 - Factor)
Then you can shuffle the XMA terms a bit to get the # of terms before
the Price will no longer have effects on your data.
For example, say, the current Price = 100, Factor = 0.1
Then, in current bar it carries the weight of 100 x 0.1 = 10.
Then, in next bar it carries the weight of 10 x (1 - 0.1) = 9.
So, here it goes,
Bars forward Weight remains
0 10
1 9
2 8.1
3 7.29
...
Say most of the time we are only interested in values greater than 0.01,
then you can summarize the "fade out" period N as
Price x Factor x (1 - Factor)^(N - 1) <= 0.01
So for the example above, you need to move forward by 66 plus terms.
Notice that this information also tells you that if the XMA is applied
to a series that moves very fast in terms of rate of change, the xma
will never be stable and always produces different results if the starting
point of the price series has changed.
One good example - XMA on rate of change :)
=====
Lawrence Chan http://www.tickquest.com
Transform market data into opportunities
|