|
PureBytes Links
Trading Reference Links
|
Was on vacation, please excuse the late post...
I have used OmniTrader and ARM2 for couple years now, both "out of the box"
and constructing some custom systems for realtime data.
Here's the crux of my conclusions which vary for EOD vs. RealTime. In short,
Omnitrader can be somewhat useful, but not in the way I expected... natch...
1) If you use Ed's suggestion of "confirming" ARM2 signals with a simple
double MA cross AND an MACD cross, you will probably be right more often
than wrong on EOD signals only (!!).
What I found reasonably successful was to create a "Master Profile" for
ARM2, then use confirming columns of all the other basic Profiles. Then I'd
sort by Master signal, then Ed's Magic, then Breakout signals.
I'd also filter the list quite severly for high signal strength and high
APR.
Those stocks that bubbled to the top (and that showed lots of confirming
signals in all 5 columns) would then be examined for the MACD and MA
confirms. If you got all that to agree, then as an EOD entry, it was not too
shabby.
HOWEVER, you have to wonder which was the real "signal"... the ARM2, or the
simple MACD and MA confirms. Interestingly enough, there is an undocumented
statistical listing in the Trading Systems window that rates every one of
the 120 possible signals on your stock baskets.... and of course, MACD and
MA crosses consistently had some of the highest "backtest hit rates" on a
percentage basis, whether optimized or not.
2) As for the Realtime systems, I personally found that ARM2 fell flat on
it's face most of the time. And I didn't much care for the non-ARM reatime
profiles either.
What I ended up doing was to use the previously mentioned statistical
studies to create my own modules from scratch. If you played with the
optimization ranges and the overall strength of the voting process filters,
you could create a realtime system that would be faily good on a basket of
stocks, but rather inconsistent on any one particular symbol.
Obviously, OmniTrader was aptly demonstrating the danger of overoptimization
in a wonderful way!! For example, JNPR would leap to the top of the realtime
sort list for about one week, then disappear off the map the next. But as
stated below, this fact itself might be useful in TS.
3) What I DID find of value in OmniTrader Realtime was the ability to, in a
very broad sense, test the overall usefullness of these 120 indicators.
As stated, the MACD and double MA cross generated consistently more accurate
entries than say, pure oscillators (at least those included with OT). Of
further interest was that the so-called pattern recognitions (candle
patterns) were extremely accurate, but generated very few signals.
Other intriguing standouts were Step and Reverse (parabolic in TS) and
longer range trendline breakouts.
Lastly, it seemed the simpler the system (fewer indicators), the more
reliable the voting process. FWIW, I would force the core indicators to
vote, and let the rest fight it out as whether they were "significant"
enough to add to the overall decision.
You can also learn a lot about exit strategies by again creating your own
rather than using the included packages.
4) Obviously, in the end, I came to TS because the backtesting was just not
rigorous enough in OmniTrader, and the rather simplistic fuzzy logic was too
undocumented to work effectively with.
However! As mentioned I still found a decent use for OT Realtime without the
ARM2 module. If you watch what stocks tend to consistently bubble up over
time, even with some optimization thrown in, you can find some nice
candidates for a more disciplined system design in TS. Sort of a volitility
filter of sorts.
Hope this helps!
Gene
----- Original Message -----
From: "Ian Waugh" <ianwaugh@xxxxxxxxxxxxxxxxxxx>
To: <omega-list@xxxxxxxxxx>
Cc: <ianwaugh@xxxxxxxxxxxxxxxxxxx>
Sent: Wednesday, May 09, 2001 9:15 AM
Subject: Nirvana, OmniTrader and ARM2
> Anyone using OmniTrader successfully? I think there's a lot of useful
> information here trying to get out, I'm just not sure how to get it out...
>
> However, I keep getting flyers from Ed Downs about ARM2, an (expensive)
> add-on which claims something like 70% accuracy in predicting signals.
>
> I'd really like to hear from anyone using OT successfully and particularly
> anyone who has more information about ARM2.
>
> Please email privately if you wish.
>
> Cheers,
> Ian
>
>
|