|
PureBytes Links
Trading Reference Links
|
Dans un courrier daté du 04/10/98 17:59:52 , vous avez écrit :
<<
i agree with what you say 'cause i think i said it before regarding
sensitivity. now you expounded on the other end of the problem:
stabilization.
question1: if one use blau's momentum (or any velocity or acceleration
function), what are the minimum number of bars for stabilization to
occur?
>>
Depends on the lookback periods of the various averages.
This case is difficult , because it relies on the initialisation of imbricated
Xaverages.
You can determine by experience:
Build two charts , one starting before the other.
Insert the Blau's code in both (as an indicator) and see when both match (and
continue without exception) on both charts.
Examining the bar where dispalyed values are the same will give you an idea of
the stabilization period from the shortest chart.
<<
question2: how can one write the code so as not to produce these
errors when dealing with momentum? i seem to recall mark jurik and bob
fulks talking about this stuff many months ago.
>>
One solution is to initialize the Xaverage with a regular average (same
length), or a linear regression estimate.
This will solve the initialisation quicker, and takes the advantage of MBBack.
The drawback is that you cannot easily use this with imbricated Xaverages,
because the initialisation of the 2nd can be done only after the first one.
The solution, in this case, is to build the initialisation code in a general
user function that uses all the xaverages (as raw code with their own
initialiation when the correct value is known for sure).
This is not a TS code problem, this is a programming problem that is due to
the recursive calculation.
<<
i plot close to 13,000 tick bars per chart, so i really don't worry
about that problem since it's 13,000 bars in the past. now i'm
wondering should i worry?
>>
In my opinion, only the very first 100 bars could be affected. Certainly
less, maybe more, but never 1000 or 13000!
You do not have to worry about if you use 13000 bars.
Do the dual chart experiment to estimate the stabilization window size, then
you will know when having to worry about or not.
<<
TJ
just trying to learn, man, just trying to learn
>>
Just trying to explain, mister, just trying to explain.
Sincerely,
Pierre Orphelin.
|