|
PureBytes Links
Trading Reference Links
|
List,
Many moons ago, I posted a little linear regression system that featured
the Forecast Oscillator. The response was surprising (lots of it) and
today, I still communicate with many of the original respondents. I've
continued to use the same "framework" for my testing. In an earlier post
today (a private email that made it to the list...I'm a little dingy
tonight...had to get up a 5 am to trade cocoa), I alluded to using the CMO.
I've used many indicators in these tests (i.e., Forecast Oscillator, a
modified Time Series Forecast, MACD Histogram, Bollinger Band Histogram,
CMO, & others).
Before I explain the method to my madness, please read the following
sentences carefully. Backtesting systems is very dangerous. The act of
backtesting is not the dangerous part...believing that the results can be
duplicated in the future is very dangerous. Let's face it, we are "best
fitting" circumstances to static prices etched in stone. So please, I'd
prefer not to hear the lectures about the folly I pursue. I've been system
testing since 1975 and I've made a bazillion mistakes (and a little chump
change) over the years. I'm still looking for the holy grail. So, here's
the outline:
1. The basic formula:
Enter Long:
Cross(opt1,ForecastOsc(CLOSE,opt3))
Close Long:
Cross(ForecastOsc(CLOSE,opt3),opt2)
Enter Short:
Cross(ForecastOsc(CLOSE,opt3),opt2)
Close Short:
Cross(opt1,ForecastOsc(CLOSE,opt3))
You can substitute any standard formula for the ForecastOsc or you can put
in a custom formula (just remember that custom formulae need to look like:
fml("Karnack's SuperSecret") It's in your manual.
2. opt3:
In this search "opt3" represents the number of days inserted into the
forecast oscillator. I usually use three (3) to ten (10) for the forecast
oscillator, but if I'm using a custom formula, sometimes I don't even need
opt3, because I using a fixed set of parameters within the custom formula.
3. opt1:
Opt1 is the numeric value below a zero basis line that will trigger a long
position and close out the short. Yes Virginia, in my secular little
world, I prefer stop and reverse trading. The parameters for this option
depends on the commodity (and yes, it does work on stocks) you're trading.
One must eyeball the forecast oscillator to see how far it swings above and
below zero. For the forecast oscillator, I usually use 0 to -3.
4. opt2:
Opt2 is the numeric value above a zero basis line that will trigger a short
sale. Zero to 3 seems to work for this formula.
5. Steps:
I step opt3 using whole numbers to represent days. With Opt1 and Opt2, I
use: .1 steps.
6. Other indicators:
When substituting the CMO (or any indicator) for the Forecast Oscillator,
one must be aware of the terrain the indicator travels over. It would be
ridiculous to us zero to 3 (as the optimizing numbers) if the mid point is
50 and the indicator traverses between +10 (on the downside) and +90 (on
the upside).
The overall theory behind this test is that many indicator oscillate from
positive to negative and back again (duh). The trick is not to trigger
action when the indicator turns in a new direction (if you're interested,
I've been down that road and I'm still wearing a neck brace from the
whiplash). The theory is that once an indicator extends to a certain
level, the market is either overbought or oversold.
In downtrending markets (can you spell deflation?), the short sale trigger
(opt2) is going to be closer to the zero basis line than opt1. Please see
the attachment. What will happen when the grains, cocoa, crude, and damn
near everything else starts to go up? Good question Steve! The system
will not perform as well if you continue to use the same parameters. In a
perfectly sideways market, one would assume that the trigger points would
be equal distance from zero. As in many markets, this system works better
when things trend indefinitely.
I hope this post will help others who have inquired about the linear
regression system. Attached is the original system, using the Forecast
Osillator, for March Crude Oil. In this example, opt3 is set to 8 (number
of days in the forecast oscillator); opt2 is .1 (sell signal); opt1 is -2.3
(buy signal).
To quote R.N. Elliot: "Even though we many not understand the cause
underlying a particular phenomenon, we can, by observation, predict the
phenomenon's recurrence."
To qoute Karnack (my alter ego): "I got knocked down seven times and got
up eight".
Finally, from a trader on the realtraders forum: "Futures trading involves
financial risk, lots of it".
Sweet dreams,
Steve Karnish
CCT
Attachment Converted: "c:\eudora\attach\crude1.gif"
|