|
PureBytes Links
Trading Reference Links
|
This is to clarify a few points that came up in a barrage of public and
private emails.
1. TS is a good program, why use Mathematica?
Please note that it is not my intention to knock TS. The software has served
me well, for charting purposes as well as for simple systems, ever since
1989 when I first bought System Writer. However, for some of my applications
TS has simply proved too limited, and so I moved on to Mathematica. Just as
an example, it is very easy in Mathematica to plot a 1-sigma line (which is
shaped like a parabola) with the most recent pivot high as its vertex; I
didn't find any way to do this in EL (now don't write to tell me I can use
Bollinger Bands - this is something completely different). Yes, I know I can
always create a DLL, but the point is that in Mathematica this plot can be
done effortlessly from *within* the language. There are many things like
this in option analysis for which the absolute freedom of Mathematica is a
God-sent.
2. Who cares about precision when data quality is the real problem?
Yes, Omega data have their problems, but why should this keep one from
researching just how reliable a tool TS really is, and how it behaves under
numeric stress.
3. What's the use comparing the precision of TS with Mathematica? Naturally
Mathematica will turn out to be more accurate because it was written for
better accuracy.
When you suspect a measuring tape to be inaccurate, you need to compare it
to a measuring device that can be depended upon to be correct. As you know,
even double-precision calculations are only approximations because they are
limited by the floating-point unit of the processor. In contrast,
Mathematica can be told to do a calculation to any desired degree of
precision - even to thousands of decimal places. Sure, it would not make any
sense at all to build this accuracy into TradeStation, and even Mathematica
will use it only on demand. But for *testing* the accuracy of TS we need a
yardstick that can be depended upon to be accurate.
4. The EMA tests done so far don't seem to show any anomaly.
Right now, I have no way of checking whether these results are correct. It
may also be that the EMA is not a good enough example because it may be
programmed in a way that does not entail the kind of feedback processes that
lead to numeric instabilities. I hope, in a few days, to post an example
that will show exactly what can happen. Please be patient, I have no time
now, but I'll be sure to look into this.
5. Concentrating on these precision questions will not make one a successful
trader.
Quite true. Reminds me of the story about that millionaire - when asked to
what business method he owed his success, he replied, "I buy widgets at one
dollar apiece, and sell them for two dollars, and from that one percent
difference I have made all my money." We can only speculate if it would have
hurt his career if he had also learned some basic math...
Best wishes,
Michael Suesserott
|