PureBytes Links
Trading Reference Links
|
> -----Message d'origine-----
> De : MikeSuesserott [mailto:MikeSuesserott@xxxxxxxxxxx]
> Envoye : dimanche 29 juillet 2001 20:35
> A : pierre.orphelin@xxxxxxxxxxxxxx; omega-list@xxxxxxxxxx
> Objet : AW: TradeStation Precision - Summary
>
>
> Hi Pierre,
>
> (un?)fortunately, I have long since removed TS from my computers,
> so I have no way of trying out any EL example for you, nor to improve on
yours.
So, you had a good software and you removed it????
>
> But I am glad you suggest to do some testing; this is in the true
> spirit of the scientist who wants to find out the truth, not necessarily
be right.
> Unfortunately, the test you describe, Pierre, is not sufficient to prove
> anything at all, though it may impress the unsuspecting public - epater le
> bourgeois, isn't that the phrase?
>
Should be the phrase if it was my intention.
But this was not the case.
First, I used to average a constant because I do not use any other
software,and that the comparizon was straightforward
Second, the xaverage function below, shows that XAverage[1] is used with no
doubt, so if the error exist
after the first bar, it will propagate because we use XAverage[1] to
calculate XAverage[0].
(1 - Factor) * XAverage[1] obviously does not carry a noticeable error.
Even if price varies from bar to bar, it will carry the same order of
magnitude error ( price and price [1] are close
{*******************************************************************
Description: Exponential Average
Provided By: Omega Research, Inc. (c) Copyright 1999
********************************************************************}
Inputs: Price(NumericSeries), Length(NumericSimple);
Variables: Factor(0);
If Length + 1 <> 0 Then Begin
If CurrentBar <= 1 Then Begin
Factor = 2 / (Length + 1);
XAverage = Price;
End
Else
XAverage = Factor * Price + (1 - Factor) * XAverage[1];
End
> I am sure you are aware of the difference between repetition and
> iteration. Your test just *repeats* the same brief iteration with the same
numbers
> 35000 times. Not much enlightenment is to be expected from such a
> test, and you yourself correctly named it "debile". <g>
>
This is YOUR xaverage example. You said that EMA ( xaverage) propagates the
error because of the lack of double in TS.
I say no, it does not ( and this is not due to the constant serie). The
cumulative effect of the error has the same order of magnitude with constant
series or different price in the serie,and it goes always in the same
direction, adding (1 - Factor) * XAverage[1]
> Let me suggest a different experiment. Calculate a 200-bar EMA over that
> 35000 bar chart, not using a constant, but real EOD values instead, for
> instance from the SPX chart, and then repeat the very same
> calculation with the very same data in an arbitrary-precision environment
such as
> Mathematica or Maple. I think you will be surprised at the results.
I do not use these software; so I cannot.
Maybe they are dead wrong, why not ?
Because they were not developped by Omega, hmmm?
Sincerely,
Pierre Orphelin
www.sirtrade.com
TradeStation Technologies representative in France
|