[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: CIS Ugly Duckly Fooey!



PureBytes Links

Trading Reference Links


> It usually annoys me when comments are made about those who trade
> within 3 day windows are 'gurus' who make excuses. These statements
> are usually made by those who have not found a way PERSONALLY to use
> them.
> 
> If you want to know about trading time days, you ask those who
> specialize in them. If someone has to make excuses for a window that
> grows bigger than +/- one day off, then that person specializes in
> something else, not pure time days.
> 

I took a look at your bias page, and there is very good, solid 
material here for market analysis. A possible next step for traders 
is to "integrate" these results into the same kind of analysis done 
on the next-higher timeframe. Best of all, this approach lets the 
market itself speak to the trader. Now the trader's job is to listen 
and observe, and release oneself from the need for the trader's 
analytical expectation to come true. This last part is most 
important, and the essence of good trading. 

When one develops a method they believe will forecast pivot points in 
time, defining them as being a bar where the bar either side of it 
does not exceed the target bars extreme excursion in the direction of 
the pivot point, one must be careful that statistical chance does not 
nullify apparent good results. I'm not trying to be fancy with words 
here. If one takes a price series of a tradable, and tests each and 
every day for the properties of the "pivot", one will find a high 
number of bars that fit. Adding a plus-or-minus one or two bars, and 
one will find many, many bars that fit.

This will cause the pivot-point forecast method to look "very good" 
and have a high degree of apparent accuracy. However, the statistical 
tools to establish confidence (Z-score, chi-test, etc.) will show a 
high degree of cause by "chance", since a high number of bars 
satisfied the pivot conditions.

This is where the danger comes in for traders. To have a _real_ edge 
over markets, it must be proven that apparent edges are real. All 
traders should not overlook the danger of letting any trading signal 
be "plus-or-minus one day" or whatever bar length. Look at simple 
math: there are 5 days in a trading week. If you allow any signal to 
be defined "right" at plus-or-minus one day, you have just used three 
of the five days. You will be "right" 60 percent of the time just by 
random chance. If two different trading signals occur and they are 
four days apart, you have covered a 6-day period completely!! So, the 
trader's "signal" is "right" 100 percent of the time, and it sure is 
impressive to display all the many occurrences where the signal was 
"right". But, of course, the trader is fooling themselves with their 
percieved success.

I hear you have Tradestation now. This is excellent for a programmer 
such as yourself. Your "bias" approach is an excellent candidate for 
a screen/filter within Tradestation, and I do something similar 
myself, using multiple timeframes, swing concepts and price excursion 
distances, to assist in my trading.

I would urge you to test FDATES on a rigorous basis. Not for accuracy 
at pivot points, but for the establishment of tradable conditions 
that can be exploited by the observance of rules and the 
establishment of action conditions.

It is one thing to have indicators or whatever that are "right" (in 
our minds only, I might add) a high percentage of times, and quite 
another thing to trade profitably from the same signals.

And trading profitably is the bottom line.

Cliff