|
PureBytes Links
Trading Reference Links
|
At 11:32 PM +0200 7/28/01, pierre.orphelin wrote:
>Maybe there is something else on this list who understand the precision
>calculus allowed display concept.
>
>I cannot believe that it is only a french specificity...
Please read "The Perils of Floating Point" at:
<http://www.lahey.com/float.htm>
Perhaps it explains it better than I have been able to.
This is simply the way the floating point representation works. That
is a fact in any language.
TradeStation uses single-precision floating point representation so
it has these problems.
Excel uses double-precision floating point representation (probably)
so it has these problem too, but they are 10 million times smaller
than the problems in TradeStation because it uses more precision, so
these errors are usually not noticeable in Excel.
Even with this extra 10 million X in precision, the people who wrote
the Excel functions knew about these issues and made sure the
algorithms they used were not compromised by the accuracy issues even
as small as they were. The total Excel product (complete with it's
function library) can be counted on to provide accurate answers.
Other programming systems such as MatLab also go to great lengths to
assure that the accuracy of their product with its function libraries
give very accurate answers. That is what the world has come to expect
in such products.
The problems are VERY NOTICEABLE in TradeStation so people who write
functions and programs for TradeStation have to be aware of the
issues and "work-around" them. The people who wrote the TradeStation
functions either were not aware of the problems or didn't care about
the accuracy so the functions and indicators shipped with the product
have noticeable errors. That is a fact - not my opinion.
My initial post in this thread tried to point out how to "work-around"
the issues, because unfortunately, Omega doesn't seem to want to fix
them.
Bob Fulks
|