|
PureBytes Links
Trading Reference Links
|
yes,
please share your work-around.
<FONT face=Tahoma
size=2>-----Original Message-----From: b519b
[mailto:b519b@xxxxxxxxx]Sent: Thursday, March 13, 2003 9:31
AMTo: amibroker@xxxxxxxxxxxxxxxSubject: [amibroker] Re:
Optimization - "Issues" testing portfolios with
AB.Chuck,At present, using AB to simulate
trading a portfolio raises a number of "issues". When Tomasz adds
portfolio level testing to AB (hopefully soon), all of these issues will
be history. Until then, users need to be very careful when using AB to
test such systems. As I said there are several issues, but this post
will focus on the one that is most likely causing your strange
results.> I am building the composite equity by saying:>
AddToComposite(Equity()/1000, "~Composite", "X");Dividing by 1000 is
not a problem and it has the advantage you state. But your
composite does not take into account how many stocks have contributed to
the equity curve on a given bar. As it stands, the composite justs sums
the equity curves of all the stocks tested. Since the number of stocks in
existance (even when including inactives) has increased over time, the
composite is NOT measuring return, but total capital invested (plus or
minus profit). Until AB adds portfolio level testing, there is no way to
tell AB to limit the total capital invested in any given year. So if 100
stocks met the criteria in year 1, the equity curve of the 100 stocks gets
added to the composite. If in year 2, 50 stocks met the criteria to
trade, then the composite will have the original 100 plus 50, and so
on. Here is another way to look at the same issue. When AB looks
at the first stock (it tests each independently), adds that stock's equity
curve to the Composite. Lets assume stock 1 started trading in 1991.
If the system looses money, the amount getting added to the composite
will decrease over time. AB then looks at the 2nd stock and it also
looses money. However, stock 2 has only been around since 1992 and so its
equity curve gets added to your composite beginning in 1992. So even if
both stocks 1 and 2 lost money, the composite equity curve would show a
jump up in 1992. Let's say stock 3 starts trading in 1993 and it also
looses money, but the composite will show another increase. Repeating this
for the 7,500 stocks you tested could make a loosing system show
"beautiful" equity. If this explaination is correct, it suggests that
you have yet to get that profitable MetaStock code into its Amibroker
equivalent.If any are interested I can share how I try to "work
around" this issue.b--- In amibroker@xxxxxxxxxxxxxxx,
"Chuck Rademacher" <chuck_rademacher@x> wrote:> I would
appreciate it if someone could help me to understand what I'm
seeing> in an optimization run.> > I am running about
1000 combinations of parameters over 7,500 stocks over 15> years of
daily data. Fortunately, my PC has plenty of grunt and the
run> actually finishes.> > Most columns in the results
grid show negative results. Net profit, RAR,> Average
Trade, etc. are all negative and the profit factor is less than
one.> > You might be tempted to say "try another
system".> > BUT... the composite equity curve is showing a very
positive and consistent> profit. Since this system was
converted from a profitable MetaStock> system, I expected to see a
profitable equity curve. I sorted the results> by
"Average Trade" and then did a backtest on the highest ranked set
of> parameters, even though the average trade was showing as a
negative value.> > I am building the composite equity by
saying:> > AddToComposite(Equity()/1000, "~Composite",
"X");> > The reason I divide by 1,000 is so that the profit
values are shown in an> acceptable format without overflowing into
scientific notation. The equity> curve is a very attractive
40% angle, gradually going from zero to $30> million.>
> Since the equity curve looks like what I expect, why don't I see
positve> values for the important metrics like RAR, Average Trade,
etc.?> > I want to believe the equity curve, of
course. But I'm uncomfortable> seeing all the negative
metrics.> > I'm also uncomfortable simply selecting the highest
ranked set of parameters> when the best value is negative.>
> Any assistance or comments would be appreciated.Send
BUG REPORTS to bugs@xxxxxxxxxxxxxSend SUGGESTIONS to
suggest@xxxxxxxxxxxxx-----------------------------------------Post
AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx (Web page: <A
href="">http://groups.yahoo.com/group/amiquote/messages/)--------------------------------------------Check
group FAQ at: <A
href="">http://groups.yahoo.com/group/amibroker/files/groupfaq.html
Your use of Yahoo! Groups is subject to the <A
href="">Yahoo! Terms of Service.
Yahoo! Groups Sponsor
ADVERTISEMENT
Send BUG REPORTS to bugs@xxxxxxxxxxxxx
Send SUGGESTIONS to suggest@xxxxxxxxxxxxx
-----------------------------------------
Post AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx
(Web page: http://groups.yahoo.com/group/amiquote/messages/)
--------------------------------------------
Check group FAQ at: http://groups.yahoo.com/group/amibroker/files/groupfaq.html
Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.
|