|
PureBytes Links
Trading Reference Links
|
Tom Cathey wrote:
> I had hoped to get a real rise out of the optimization guys on
> this list. You see, I hate to admit it, but I too, have spent 10's of
> thousands of hours trying every derivation of
> optimization/systems/indicators out there...What happens is that based
> on the normal bell curve you get data skewed in one direction and are right
> most of the time, but when the probabilities turn the other way (as they
> always do eventually) the optimization is wrong most of the time...big
> time. Over time, you are LUCKY to break even...but add human emotions and
> most likely this "break even" senerio and results in a big loss.
This is interesting. I'm not real big on optimization either. However, in the
above scenario, the problem may be due to using too large a database for
testing. Markets do change significantly over time, and you can easily find
yourself referencing data not appropriate for current conditions. Of course,
smaller databases aren't going to be statistically significant, but that
shouldn't be the goal here anyway, since I agree that this isn't really going
to be attainable anyway - a general feel for the way strategies have performed
lately with given charts may be all we can reasonably expect here.
> In my "humble" opinion, I submit that one should find 2-3 LOOSLY fitting
> indicators that work on all markets, and learn how they act in both
> trending and chopping markets.
This has been my experience as well. For instance, I follow hundreds of
stocks, and I once had the crazy idea to try to individualize each chart.
Worked fantastic on paper (i.e. the past), but didn't perform as well as the
much more simplified approach I had been using. It isn't hard to understand
why, when you stop to think - the more generic and simple the scheme, the less
likely the results would be due to idiosyncrasies. As far as the bottom line
goes, I only use the "results" as general guides - how a system or strategy
performs in real time is the sole real indicator of performance, of course.
> THEN develop a timing method independant of
> these indicators that is the MAIN source of recommendations. This "timing"
> method is key and the source of successful trading.
I'm not sure what you mean here, Tom. Is this a matter of intuitive "feel"?
If so, I do agree that this has its place. I'm only a beginner at system
writing, and there are factors, especially overall market conditions, which I
cannot account for, and i'm not even sure I'd what to (I still think all this
is still to a degree an art). As well, there is important info that can't be
programmed at all - for instance, the way something reacts after the signal is
generated. However, mechanical signals *are* timing methods, and still can be
very useful.
> The bottom line: To develop a so called "edge" you must come up with
> something that is used by the minority at any particular time. All the
> popular canned indicators are not unique and are used by the
> majority....and are all based on price, and are usually optimized for some
> particular time frame...poison.
This is probably true with futures, but with stocks, thankfully scalpers (a few
minutes to an hour or two of holding) consist of a very small portion of the
overall market. However, staying ahead of the pack is still very important.
> As someone said here, you must take a combination of
> what works for YOU, but be VERY flexible and willing to change with the
> markets. THAT is the exact opposite of optimization.
Absolutely. I've got to the point now where I don't even keep any more than
*15 days* of data on anything anymore. You still are "optimizing" in a sense,
but to current conditions, not so much to historical performance.
> Take a look at the super stars of the industry. If you can make just 15-20%
> per YEAR trading commodities, you are on a pedestal and will attract more
> money than position limits will allow. Do you think an optimized program
> (the easy way) is the answer?
Wow! I'm glad I don't trade commodities, if this is what the "super stars" are
doing. However, I suspect that there are many good traders who do much better
than this in these markets.
As far as the vendor thing goes, I'd be very wary here as well - first, if the
system was *that* good, they wouldn't be selling it. Second, you I'd be very
skeptical of such canned results.
Regards,
A.J.
|