|
PureBytes Links
Trading Reference Links
|
David,
I will try backtesting with 10 years worth of EOD bars, but I have a
couple of questions:
1) Let's say I backtest 50 systems with 10 years data so that I can
find the best system for a particular security. With 10 years data, I
will catch the unrepeatable bubble market (1997-2000). Will this not
corrupt the results?
2) Is another option to select (or build) a system that has performed
well over the last 250 bars, and then as the market conditions
change, to abandon that system for another better performing system
that fits the new market conditions (i.e. system rotation)?
Thanks.
--- In equismetastock@xxxxxxxxxxxxxxx, "David" <junk@xxxx> wrote:
>
>
> I'm sure everyone would more than emphatically agree with me that
the
> more historical bars the better to test on. When you use less bars
> in your system test, the more likely you are going to be curve
> fitting your system to the data. That, and what happens if/when
the
> market reverts to a state similar to previous years, ie. stops or
> begins trending.... Your system more than likely won't be able to
> cope with the changes. You want to make sure the system that you
> design is more than capable of performing well (or at least
> minimizing losses) in all types of market conditions thereby making
> it more robust. You'll thank yourself later. Two years is a
> ridiculously small amount in my opinion. That isn't enough to
cover
> bull/bear and trending/non-trending markets. For me, I would be
hard
> pressed to put trust in any sytem that hasn't been tested on at
least
> a decade of historical daily data. I'm skeptical of the system I'm
> designing now because esignal has only 5 months of 60 minute data
for
> forex. But even that is still well over 2000 bars. Plus I'm using
> the same system for 20 different pairs to make sure I can at least
> cover as many different market conditions as I can...
>
> Best Regards,
> David Piatek
>
> --- In equismetastock@xxxxxxxxxxxxxxx, "metastkuser"
> <andysmith_999@xxxx> wrote:
> >
> >
> > I usually do MS EOD System Testing with 500 bars (just under 2
> > years). I notice that the manual says I should use at least 1000
> bars
> > (preferably 2000) in order for the results/data to be
statistically
> > significant.
> >
> > My question: Is it really better to use 2000 bars instead of 500
or
> > 250 bars? The stock had a completely different "personality" 2000
> > bars ago, and the market itself may have been very different from
> > what it is today. Don't I want to focus on trading systems which
> have
> > been the most successful it the last few months?
> >
> > TIA
------------------------ Yahoo! Groups Sponsor --------------------~-->
$9.95 domain names from Yahoo!. Register anything.
http://us.click.yahoo.com/J8kdrA/y20IAA/yQLSAA/BefplB/TM
--------------------------------------------------------------------~->
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/equismetastock/
<*> To unsubscribe from this group, send an email to:
equismetastock-unsubscribe@xxxxxxxxxxxxxxx
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
|