|
PureBytes Links
Trading Reference Links
|
"Patrick Gamble" <pgamble@xxxxxxxxx> writes:
>Any software developer, when looking at the level of
>precision they need in their programme, first must
>look at the data that the application will be handling.
She must also look at the potential mathematics that data
may be subjected to. This question will have much more
influence on the precision requirements.
>If the data is accurate to, say, two decimal places, then
>it is useless functionality to be able to deal with that data
>to six places. The data that Tradestation deals with is
>often not even accurate to two places.
That might suffice if all the user will do with the data is
add up a few sums, or compare a few simple patterns on the
raw data. It's wildly insufficient if the user might try
slightly more sophisticated algorithms. A measure of some
process---an "indicator" if you like---can require MUCH more
precision than the data fed to its input.
>It is also wrong from a system building perspective, as it
>suggests an entirely incorrect level of detail is attainable.
>
>So the demand that Tradestation should be accurate to
>more decimal places is not only stupid, in that it focusses
>attention on an 'improvement' that has no bearing on the
>job the software is designed to do, but is also a symptom
>of a state of mind that entirely fails to understand the real
>level of accuracy that is attainable. These traders will fail.
This sort of thinking restricts Tradestation to nice simple systems
on the order of "is the price approaching this trendline I placed
on the chart by hand?" One can argue that sophisticated math
is a distraction, and one really should just watch price, and for
those folks, Tradestation indeed forces this simple methodology---
if the poor user realizes the limitation exists. If he doesn't
realize it, he's liable to get a "system" with a mind of its
own, fueled by instabilities in TS's arithmetic.
Jim
|