|
PureBytes Links
Trading Reference Links
|
Hello,
I'm in the progress of doing just that. Optimizing in a time window
and than moving that window by some arbitrary amount of bars. This
would give me some feeling about the 'time stability of optimized
values' op a particular system. This is still work in progress ...
I've uploaded the VB code to the files section if you'd be
interested.
Regards
Leo
--- In amibroker@xxxxxxxxxxxxxxx, "willem1940" <w.j.a.struyck@xxxx>
wrote:
> Picking up your "invitation" under PS I wondered if the suggestion
> below is "do-able" in AB.
>
> I have studied the earlier exchange of messages on walk-forward
> methods but from the contenst I could not construct the way I
carry
> out the walk-forward exercise; trying to imitate as much as
possible
> the reality.
> The sequence I follow is:
> - establish the period for the optimization in nr of bars
> (period)
> - set the range to the point in time where the first
> optimization should end say one week from now: (endtime)
> - optimize
> - place parameters in formula
> - move the range an agreed number of bars (bars) nearer the
> last bar in the system (endtime - bars)
> - backtest for this period
> - copy the results to Excell
> - set the range to the point in time where the second
> optimization should being (endtime - bars)
> - optimize
> - place parameters in formula
> - move the range again an agreed number of bars nearer the
last
> bar in the system (endtime - bars - bars)
> - backtest for this period
> - copy the results to Excell
> - set the range to the point in time where the third
> optimization should being endtime - bars - bars)
> - optimize
> - etc.
>
> Subsequently in Excell I combine the entries in one trading track
> record by deleting the overlap in time leaving every time the last
> entries, so the true result of the day after the op[timization
> finished.
>
> As you can imagine this is tedious work and of course I wondered
if
> with the new functions in AB this can be done automatically.
> Unfortunately I am not a programmer otherwise I would give it a
try.
>
> Willem Jan
>
>
> --- In amibroker@xxxxxxxxxxxxxxx, "DIMITRIS TSOKAKIS"
<TSOKAKIS@xxxx>
> wrote:
> > "I don't know whether AA's optimization process works like the
> > Optimize function in IB but suspect it might."
> > Bill,
> > There is a [probable] confusion here:
> > From the definition in 4.40 users guide,
> > "With normal backtesting, scanning, exploration and comentary
modes
> > the optimize function returns default value, so the above
function
> > call returns default; "
> > In other words, a code
> > p=Optimize("p",40,10,50,10);
> > y=StochD(p);
> > Plot(y,"y",1,1);
> > would produce the graph of StochD(40) and nothing more.
> >
> > "However, the basic question is whether k was optimized for the
> > entire dataset or just 2002."
> > The described code [as the older ref] does exactly this : It
> > optimizes k for the restricted period.
> > "As far as I can tell it is for the entire dataset, reflecting
the
> > fact that the Optimize function does not have a time frame
argument
> > (s) and, therefore, no way to limit its operation to a specified
> time
> > window."
> > Since Optimize function does not have time frame arguments,
there
> is
> > *another* way to limit its operation : Apply Optimization to the
> > specified time window and you have the required result. Try any
AA
> > test and you will see.
> > "As a result, I wondered if anyone had a workaround that would
> force
> > the Optimize function to only use data within a specified time
> frame."
> > The described codes do this job. You may disagree with the
logic,
> but
> > the result is exactly the requested.
> > Some AFL functions/procedures are limited to a range of
properties,
> > clealy exposed in the definitions. When we need something out of
> > these limitations, we have to use another way.
> > There was a similar discussion with variable periods. Some AFL
> > functions do accept variable period, some others dont. If we
want a
> > variable per in EMA(C,per), we can not act directly, EMA does
not
> > accept variable period. We have to find another, indirect way,
to
> do
> > it.
> > You ask from Optimize function some results out of its [current]
> > definition . It is not possible. [Imagine the opposite !!!AFL
would
> > become an unreliable set of functions !!!]
> > IMO, the only way to get a result through AFL is to follow the
> > definitions "mot a mot". Any other attempt would cause mistakes.
> > Sometimes the mistake is prevented by the AFL Error window, but
not
> > always.
> > There is no code mistake in the lines
> > p=Optimize("p",40,10,50,10);
> > y=StochD(p);
> > Plot(y,"y",1,1);
> > but, dont expect from these lines an optimization procedure
similar
> > to the OPTIMIZE button, it is N/A [yet]
> > I hope it is clear.
> > Thank you for the interesting dialogue, I have nothing more to
add.
> > Dimitris Tsokakis
> > PS For "similar" IB procedures we may use for loops and
practically
> > have the same results with optimization procedure. I have solved
> the
> > general problem, I think it may be applied to any case, but it
> would
> > be better to ask a specific question.
> >
> > --- In amibroker@xxxxxxxxxxxxxxx, "wavemechanic" <wd78@xxxx>
wrote:
> > >
> > > ----- Original Message -----
> > > From: "DIMITRIS TSOKAKIS" <TSOKAKIS@xxxx>
> > > To: <amibroker@xxxxxxxxxxxxxxx>
> > > Sent: Friday, September 12, 2003 2:18 PM
> > > Subject: [amibroker] Re: Optimize function
> > >
> > >
> > > Bill,
> > >
> > > (Initial message)
> > > k=Optimize("k",40,30,50,10);
> > > > d2002=DateNum()>=1020101 AND DateNum()<1030101;
> > > > LastTradingBar=(d2002==0 AND Ref(d2002,-1))*DateNum();
> > > > Buy=d2002*Cross(StochD(),k);
> > > > Sell=d2002*Cross(StochD(),70) OR DateNum()==LastTradingBar;
> > > >
> > > > Hit now Optimize for
> > > > *current stock
> > > > *all quotations
> > > > AA will see only 2002 dates.
> > >
> > > (Second message)
> > > Optimization, by itself, has no restriction property.
> > > When we say Optimize, the AA will check the settings or
whatever
> > > overwrites the settings.
> > > In my example, the settings are "all quotations".
> > > But, the trading rules are restricted by definition:
> > > Buy=d2002*Cross(StochD(),k);
> > > Sell=d2002*Cross(StochD(),70) ;
> > > This d2002 is a binary quantity, equal to 1 or 0, as definrd
in
> the
> > > previous lines. The d2002*something will give something or 0.
> > > Consequently, when d2002 is true, the AA will read
> > > Buy=Cross(StochD(),k);
> > > Sell=Cross(StochD(),70);
> > > and when d2002 is false, the AA will read
> > > buy=0;
> > > sell=0;
> > > Make a scan for all quotations to see. The trades will be for
2002
> > > dates only.
> > > In other words, in order to optimize from datenum1 to
datenum2,
> just
> > > define trades for this period only and thatīs it !!
> > >
> > > As I mentioned to Leo a few minutes ago, I am not using AA but
> > rather
> > > generating charts with IB. I don't know whether AA's
> optimization
> > process
> > > works like the Optimize function in IB but suspect it might.
> With
> > that
> > > uncertainty as background, I think we are talking about apples
> and
> > oranges.
> > > Yes your code will produce trades only in 2002 with an
optimized
> > value for k
> > > (optimized in code from first message), and I use other code
at
> the
> > present
> > > time to achieve similar results. However, the basic question
is
> > whether k
> > > was optimized for the entire dataset or just 2002. As far as
I
> can
> > tell it
> > > is for the entire dataset, reflecting the fact that the
Optimize
> > function
> > > does not have a time frame argument(s) and, therefore, no way
to
> > limit its
> > > operation to a specified time window. As a result, I wondered
if
> > anyone had
> > > a workaround that would force the Optimize function to only
use
> > data within
> > > a specified time frame. One could, use brute force and
manually
> > change the
> > > number of bars and time period loaded, but that is too awkward
to
> be
> > > seriously considered.
> > >
> > >
> > > --- In amibroker@xxxxxxxxxxxxxxx, "wavemechanic" <wd78@xxxx>
> wrote:
> > > >
> > > > ----- Original Message -----
> > > > From: "DIMITRIS TSOKAKIS" <TSOKAKIS@xxxx>
> > > > To: <amibroker@xxxxxxxxxxxxxxx>
> > > > Sent: Friday, September 12, 2003 12:48 PM
> > > > Subject: [amibroker] Re: Optimize function
> > > >
> > > >
> > > > > Bill,
> > > > > I have posted in the past the walk-forward optimization.
> > > > > Here is the basic principle, for a restricted period
> > optimization
> > > > > [year2002 in the example]
> > > >
> > > > If I correctly understand, optimization of k is still being
> > > established on
> > > > the entire dataset, rather than only on 2002 data. Right?
If
> so,
> > > how would
> > > > you restrict Optimize to operate only in 2002?
> > > >
> > > > > of a Stochastic o/b, o/s system :
> > > > >
> > > > > k=Optimize("k",40,30,50,10);
> > > > > d2002=DateNum()>=1020101 AND DateNum()<1030101;
> > > > > LastTradingBar=(d2002==0 AND Ref(d2002,-1))*DateNum();
> > > > > Buy=d2002*Cross(StochD(),k);
> > > > > Sell=d2002*Cross(StochD(),70) OR DateNum()==LastTradingBar;
> > > > >
> > > > > Hit now Optimize for
> > > > > *current stock
> > > > > *all quotations
> > > > > AA will see only 2002 dates.
> > > > > It is important to exit any open trade by the end of the
test
> > > period,
> > > > > else the last probable Open trade will last for ever !!
> > > > > The solution is that{ OR DateNum()==LastTradingBar }
> statement.
> > > > > Dimitris Tsokakis
> > > > > --- In amibroker@xxxxxxxxxxxxxxx, "wavemechanic"
<wd78@xxxx>
> > > wrote:
> > > > > >
> > > > > > ----- Original Message -----
> > > > > > From: "Stephane Carrasset" <nenapacwanfr@xxxx>
> > > > > > To: <amibroker@xxxxxxxxxxxxxxx>
> > > > > > Sent: Friday, September 12, 2003 11:55 AM
> > > > > > Subject: [amibroker] Re: Optimize function
> > > > > >
> > > > > >
> > > > > > > Bill,
> > > > > > >
> > > > > > > do you mean if you Backtest for example NDX on the
last
> 500
> > > bars
> > > > > > > do you want optimize an indicator on the 500-400
bars??
> then
> > > keep
> > > > > the
> > > > > > > optimized value on the 400-1 bars ??
> > > > > >
> > > > > > I was not thinking of that, but your scenario is along
the
> > same
> > > > > lines and
> > > > > > would be a another important option. I was actually
> thinking
> > of
> > > > > the problem
> > > > > > from 180 degrees around. Specifically, for a dataset
of,
> for
> > > > > example, 1000
> > > > > > bars optimize only over the last 500 and display the
equity
> > > curve
> > > > > for that
> > > > > > range. It is possible to only consider (and display)
the
> > equity
> > > > > for a
> > > > > > specified number of bars (e.g., the last 500 bars), but
I
> > don't
> > > see
> > > > > how to
> > > > > > restrict optimization to those bars. Certainly not with
the
> > > > > available
> > > > > > arguments for Optimize.
> > > > > >
> > > > > > >
> > > > > > > Stephane
> > > > > > >
> > > > > > > > It is my understanding that the Optimize function
> operates
> > > on
> > > > > all
> > > > > > > data. So,
> > > > > > > > for example, if one is looking at equity for a
portion
> of
> > > the
> > > > > data
> > > > > > > (e.g.,
> > > > > > > > Equity(0, 1, 100) Optimize still operates on all
data.
> > Does
> > > > > anyone
> > > > > > > have a
> > > > > > > > work-around that will force Optimize to only
consider a
> > > > > specified
> > > > > > > range of
> > > > > > > > data?
> > > > > > > >
> > > > > > > > Thanks.
> > > > > > > >
> > > > > > > > Bill
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > Send BUG REPORTS to bugs@xxxx
> > > > > > > Send SUGGESTIONS to suggest@xxxx
> > > > > > > -----------------------------------------
> > > > > > > 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
> > > > > http://docs.yahoo.com/info/terms/
> > > > > > >
> > > > > > >
> > > > >
> > > > >
> > > > >
> > > > > Send BUG REPORTS to bugs@xxxx
> > > > > Send SUGGESTIONS to suggest@xxxx
> > > > > -----------------------------------------
> > > > > 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
> > > http://docs.yahoo.com/info/terms/
> > > > >
> > > > >
> > >
> > >
> > >
> > > Send BUG REPORTS to bugs@xxxx
> > > Send SUGGESTIONS to suggest@xxxx
> > > -----------------------------------------
> > > 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
> > http://docs.yahoo.com/info/terms/
------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for Your HP, Epson, Canon or Lexmark
Printer at Myinks.com. Free s/h on orders $50 or more to the US & Canada. http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/l.m7sD/LIdGAA/qnsNAA/GHeqlB/TM
---------------------------------------------------------------------~->
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 http://docs.yahoo.com/info/terms/
|