PureBytes Links
Trading Reference Links
|
I've said it before and I'll say it again: This debate over precision is
ridiculous.
It is ridiculous for two reasons: first, because the input data is not
precise; and second, because extra precision in calculations is not needed
for trading.
First, the input data is imprecise. Prices are set by a bunch of screaming
banshees in a "pit". Do we need to go out to the nth decimal place to
decide what the price is? Or is a good approximation (which is all that the
price spotters can give us anyway) good enough? I say that it makes no
sense to run calculations out to the nth decimal place when the input data
is +/- 0.1 points.
Second, extra precision is not needed for trading. What's needed for
trading is not more precision, but less. We need "fuzzy" approaches, not
precise ones. Carrying calculations out to the nth decimal place in a fuzzy
environment is ridiculous.
The Omega Man
----- Original Message -----
From: Mark Brown <markbrown@xxxxxxxxxxxxx>
To: Omega List <omega-list@xxxxxxxxxx>
Sent: Saturday, June 19, 1999 5:55 PM
Subject: Easy Language Math Precision Wrong?
> Read this old post from the TradeLab list to see.
>
> > >Even though it is possible to print more precise-looking numbers on the
> screen, TS4 and TS5 both perform mathematical calculations using what is
> called "Single Precision" in Visual Basic. Easy Language function
arguments
> are passed as Single Precision values. Results are returned as Single
> Precision values.
>
> James F. Mazzulla asked:
>
> > Thanks for the clarification, Bob. Can you tell us if this is
> =particularly= problematic and, if so, what, if any, course of action
needs
> be taken? TIA
>
> It is not problematic. It is certain.
>
> There is no reasonable way a user can improve Easy Language mathematical
> precision, short of writing high precision mathematical language
extensions.
> However, Easy Language program code can be designed to reduce end-result
> errors due to poor mathematical precision in some cases. There are
several
> ways that can be accomplished, depending on the nature of the calculations
> being performed.
>
> That is a very big subject books have been written about. The methods and
> related considerations are too extensive to explain here, but you will
find
> books on the subject in university computer science libraries. It was an
> especially important topic years ago when most computers had poor
> mathematical precision, so you might find especially interesting ideas in
> books published in the 1960's and 1970's.
>
> -Bob Brickey
> Scientific Approaches
> sci@xxxxxxxxxx
>
>
> Chris Baker wrote:
>
> > Besides single precision, TradeStation 4.0 cannot subtract correctly.
> >
> > Actual example from Easy Language:
> > X = LS1 - LS;
> >
> > Using Print statement:
> > LS1 = 868.90000
> > LS = 869.10000
> > X = -0.19995100
> >
> > Since X now has 6 significant digits instead of 1, when I subtracted it
> from another variable in the same format as LS and LS1, it lead to a
> TradeStation floating point error which required the server to be
restarted.
>
> That error is due only to lack of precision.
>
> -0.2 = 868.90000 - 869.10000
>
> Easy language returned -0.19995100, which is in error by only:
>
> -.000049 = -0.2 - (-0.19995100)
>
> -Bob Brickey
> Scientific Approaches
> sci@xxxxxxxxxx
>
>
|