|
PureBytes Links
Trading Reference Links
|
Hi Mike,
That's sad you can't see the images. I am sure you would understand my questions. Ok, here is what I did step-by-step:
1) I used the "detrend' script with YHOO for 1998 to 2000. The ~YHOO file was created.
2) I opened the ~YHOO file and in the small formula window at the botton of the screen I see a bunch of pink, yellow and green lines around 1.30+ and at the down of that same window I see a bar "zero" with green arrows below and red arrows over. (I don't understand this; what are the lines for? and why the red/green arrows if no buy/sell signals have been issued yet). Those lines are all AFTER then end of the period that is in the data (after 2000), and between 1998 and 2000 I see a blue oscillator but no pink/yellow/green lines at all.
3) Then, I use the second script (but I don't know what to set as a foreign because I don't know what is the utility of this; I set COMP because I didn't know what else); And then after clicking backtest there was one (1) transactions only, beginning on the first day of the data and exiting on the next day. Many other raw signals, but no actual order issued. An another problem, as I said in my last message, is that if the rule is based on the log itself, I can't issue commands like (buy>2) as an example. I must stay with "general" rules.
This is complicated. I like it! ;-)
Thanks,
Louis
2008/3/3, Mike <sfclimbers@xxxxxxxxx>:
I don't receive these posts in the mail, and Yahoo strips out the
images, so I can't see your image.
However, it doesn't matter. The bias is not the number of long vs.
short trades, but rather the number of days long vs the number of
days short. Even then, you don't care. Since the second script
applies your strategy to detrended data, all bias is removed from the
results.
The only answer you want to know is whether your strategy performance
is better than zero, with statistical significance, once the
position/trend bias has been stripped out.
You get that answer by comparing the mean of the second script
results (i.e. your strategy) to the mean of the first script results
(i.e. the market).
The scripts have already stripped all position bias/trending. The
mean of the first results will always be approximately zero. The mean
of the second results will reflect your strategy's predictive ability.
A t-test will tell you if the difference is more than just chance.
Mike
--- In amibroker@xxxxxxxxxxxxxxx, "Louis Préfontaine"
<rockprog80@xxx> wrote:
>
> Hi Mike,
>
> I attached an image of what I got. I suspect that the green/red
arrows may
> be the buying/selling signals. So if there are more red than green
then the
> system has a negative bias, or if it stays with a red longer than
with a
> green it is negative, right? Could you please confirm this?
>
> I tried one of my own formula which has a strong positive bias
(like making
> zillions of money in the techno-bubble and losing from 2000 to
2003) but I
> got more red arrows then green. Maybe there is something that is
not right?
>
> Please tell me if you can't see the image.
>
> Thanks,
>
> Louis
>
> 2008/3/3, Louis Préfontaine <rockprog80@xxx>:
> >
> > Hi Mike,
> >
> > I am not sure I understand. If the first script did it all, why
did you
> > recommand me to add all the close and then divide by the number
of close to
> > get the average? What was the use of this operation? And why
the second
> > script, which creates yet another quote with the ELR alone... I
think I'd
> > need to subtract the ELR from the actual quote, but how to do
that... I am
> > not sure I understand how I can have a detrended data that can be
useful
> > with my rules to get a ratio of buying/selling (short/cover)
signals that
> > would allow me to know if a have a positive/negative bias.
> >
> > Could you please explain to me how you actually use this/those
script(s)
> > to see if your rules has a bias?
> >
> > Thanks!
> >
> > Louis
> >
> >
> >
> > 2008/3/3, Mike <sfclimbers@xxx>:
> > >
> > > I believe that SetBacktestMode, or at least some of its
argument
> > > values, were introduced in more recent versions of the product.
Since
> > > I'm setting it to the default, you can just leave it out.
> > >
> > > As for the ALR, my first script (Detrend.afl) does everything
for
> > > you. The values held in the Close field of the composite symbol
(e.g.
> > > ~SP-500) will *already* have the complete, detrended and
adjusted
> > > value. Nothing more for you to do.
> > >
> > > Mike
> > >
> > > --- In amibroker@xxxxxxxxxxxxxxx <amibroker%
40yahoogroups.com>, "Louis
> > > Préfontaine"
> > > <rockprog80@> wrote:
> > > >
> > > > Hi Mike,
> > > >
> > > > Thank you so much for your reply. I really appreciate your
step-by-
> > > step
> > > > advices. This is really amazing to see the work you've done
with
> > > Aronson's
> > > > ideas!
> > > >
> > > > I tried to follow your step-by-step guide but unfortunately it
> > > seems there
> > > > may be a mistake in the following code:
> > > >
> > > > SetTradeDelays(1, 1, 1, 1); // All trades on next Open
afterEOD
> > > Signal
> > > > SetBacktestMode(backtestRegular); // One symbol, no redundant
> > > signals
> > > > SetOption("InitialEquity", 100000);
> > > > SetOption("AccountMargin", 100);
> > > >
> > > > For some reason I get an error Line 2, Col.: 16 Error 30.
Syntax
> > > error.
> > > >
> > > > Did I do something wrong? Is it possible to do the same thing
in
> > > another
> > > > way?
> > > >
> > > > Anyway I now fully understand what we're trying to do here.
After
> > > I got the
> > > > ALR for each day, all that will be needed will be to subtract
it
> > > from day1
> > > > to day 2 log. Then we will need to set rules with +1 and -1
> > > value... Well,
> > > > I am not 100% sure I fully understand, but it's getting
close. If
> > > I can
> > > > only get past this error...
> > > >
> > > > Thanks a lot!
> > > >
> > > > Louis
> > >
> > >
> > >
> >
> >
>
__._,_.___
Please note that this group is for discussion between users only.
To get support from AmiBroker please send an e-mail directly to
SUPPORT {at} amibroker.com
For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG:
http://www.amibroker.com/devlog/
For other support material please check also:
http://www.amibroker.com/support.html
![]()
__,_._,___
|