[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[amibroker] Re: Automated Backtesting Walkforward Validation



PureBytes Links

Trading Reference Links

Hello,

Another good book concerning the topic was recommended by S. Karnish 
some months ago. 
'Smart Momentum : The Future of Predictive Analysis in the Financial 
Markets' by Hugh Clark

Regards
Leo


--- In amibroker@xxxxxxxxxxxxxxx, "Herman van den Bergen" 
<psytek@xxxx> wrote:
> Yes ken/bill, I did misunderstand the procedure :-( actually all 
test
> windows are OOS. Thanks for explaining things for me.
> 
> I wonder how they would determine the Optimize/Test period 
ratio... the
> example uses 2:1. A critical factor that is not mentioned is trade 
duration.
> It would seem prudent to have a reasonable number of trades in 
both the and
> test periods. So this system would only work with short term 
trading
> systems...if you trades last more then 10% of the test period you 
wouldn't
> really have enough data to make results significant.
> 
> With spread of optimization values i mean the range of the 
parameters
> optimized. Suppose the system uses period optimization and the 
values, over
> 18 tests, would vary from 3 to 12... I would be highly skeptical 
this type
> of result.
> 
> Like was mentioned: there is no easy way...
> 
> herman
>   -----Original Message-----
>   From: Ken Close [mailto:closeks@x...]
>   Sent: Saturday, March 01, 2003 3:38 PM
>   To: amibroker@xxxxxxxxxxxxxxx
>   Subject: RE: [amibroker] Re: Automated Backtesting Walkforward 
Validation
> 
> 
>   Herman:
> 
> 
> 
>   I would tend to think that it does tell you something about the 
system
> under test.
> 
> 
> 
>   If your results in the "Trade Window" as it is called in this 
message are
> good, and they are "good" in say 15 out of 18 periods or cycles, 
then is it
> not reasonable to assume that the system under test is "robust" 
and is
> likely to perform ok (or at least as good as the average 
performance
> achieved in all of the Trade Windows).
> 
> 
> 
>   Yes, I interpreted the msg to say to reoptimize in each 
new "Test Window",
> the walk forward test in each Trade Window.  I am not sure what 
would happen
> if the beginning date of the first Test Window was anchored and 
the Test
> Window got progressively longer after each cycle.  It is also not 
clear what
> minimum number of trades must be achieved for the Test Window and 
Trade
> Window to be of adequate length.
> 
> 
> 
>   What did you mean by "spread of optimization values"???
> 
> 
> 
>   Ken
> 
> 
> 
>   -----Original Message-----
>   From: Herman vandenBergen [mailto:psytek@x...]
>   Sent: Saturday, March 01, 2003 9:14 PM
>   To: amibroker@xxxxxxxxxxxxxxx
>   Subject: RE: [amibroker] Re: Automated Backtesting Walkforward 
Validation
> 
> 
> 
>   The previous post read: "Start with a fraction of the total 
historical
> data, say 10%, call this the Test-Window, run the Optimizer on 
this data,
> apply the
>   trading system with these optimized parameters to the period of 
data
>   immediately following the Test-Window, say 5% of total historical
>   data, call this the Trade-Window, record only the Trade-Window
>   results, move the starting point of the Test-Window ahead by the
>   length of the Trade-Window, repeat (Does this mean re-
optimize?), in this
> case 18 times, until the walk-forward process reaches the end of 
the
> historical data, and then report the cumulative Trade-Window 
results."
> 
> 
> 
>   If i interpret this correctly than this really is an assessment 
of the
> optimization process itself. If so it has liitle to do with the 
trading
> system under test and has no value in predicting the effectivenes 
of the
> last (all previous values have no relevance) optimization values 
in the
> trading system on the next future period. There is not even talk 
about
> measuring the spread of optimization values...
> 
> 
> 
>   Hopefully I misunderstood...
> 
> 
> 
>   Herman.
> 
> 
> 
>     -----Original Message-----
>     From: bvandyke <bvandyke@xxxx> [mailto:bvandyke@x...]
>     Sent: March 1, 2003 11:48 AM
>     To: amibroker@xxxxxxxxxxxxxxx
>     Subject: [amibroker] Re: Automated Backtesting Walkforward 
Validation
> 
>     Hi d,
> 
>     The best definition of Walkforward Validation (WFV) i saw is as
>     follows and from a program called Inference Trader.  A google 
search
>     will show their website.
> 
>     "WalkforwardValidation (WFV) is a method of back-testing a 
trading
>     system using walk-forward optimization/trading.  It attempts to
>     simulate a more realistic trading system performance than the
>     Optimization only--whose trading results are overly 
optimistic.  The
>     method is as follows:
> 
>     Start with a fraction of the total historical data, say 10%, 
call
>     this the Test-Window, run the Optimizer on this data, apply the
>     trading system with these optimized parameters to the period 
of data
>     immediately following the Test-Window, say 5% of total 
historical
>     data, call this the Trade-Window, record only the Trade-Window
>     results, move the starting point of the Test-Window ahead by 
the
>     length of the Trade-Window, repeat, in this case 18 times, 
until
>     the walk-forward process reaches the end of the historical 
data, and
>     then report the cumulative Trade-Window results."
> 
>     "If the results as described above are positive,then one 
usually to
>     get the parameters to use for future trading, uses the 
Optimizer.
>     Since you back-tested this strategy of optimize/trade with
>     WFV optimizing the paramaters on a chunk of data "Test window" 
length
>     long, to attempt to recreate this performance in future
>     trading, you want to set the optimization length equal to "Test
>     window" length. Now run the optimizer on your last "test 
window"
>     period of data to get the parameters for future trading.  Now 
trade
>     for a period of time equal to the WFV "Trade Window Length", 
at the
>     end of this period, run the optimizer again then trade with 
the new
>     params, and so on."
> 
>     Hope this helps,
>     Bill
> 
>     --- In amibroker@xxxxxxxxxxxxxxx, "dingo" <dingo@xxxx> wrote:
>     > can you point me to some more info on "Walkforward 
Optimization"??
>     >
>     > d
>     >
>     > -----Original Message-----
>     > From: bvandyke <bvandyke@xxxx> [mailto:bvandyke@x...]
>     > Sent: Saturday, March 01, 2003 1:35 PM
>     > To: amibroker@xxxxxxxxxxxxxxx
>     > Subject: [amibroker] Re: Automated Backtesting
>     >
>     >
>     > Hi dingo,
>     >
>     > You wrote below about Optimization:
>     > >I should point out that according to most of the 
knowledgeable
>     > >sources that you should only optimize once for a range of 
dates
>     that
>     > will include bull and bear periods and then do a series of
>     > backtests "out of sample" to find out if your formula 
is "robust".
>     > That's the hard part!
>     > >
>     > > d
>     >
>     > I think Leo may be referring, in his Post,  to an 
Optimization
>     > technique called "Walkforward Validation", which is what some
>     people
>     > recommend and prefer.  It's another option to handle bull 
and bear
>     > years.  As you say, no matter what technique, it's not 
easy :)
>     >
>     > Bill
>     >
>     >
>     >
>     > >
>     > > -----Original Message-----
>     > > From: leo_amelc <leo.timmermans.lt@xxxx>
>     > > [mailto:leo.timmermans.lt@x...]
>     > > Sent: Saturday, March 01, 2003 10:09 AM
>     > > To: amibroker@xxxxxxxxxxxxxxx
>     > > Subject: [amibroker] Automated Backtesting
>     > >
>     > >
>     > > Hello,
>     > >
>     > > I've a question regarding the AA Automation objects.
>     > >
>     > > I want to do the following:
>     > >
>     > > 1) take a stock (or group of stocks) and optimize an 
indicator
>     > > during a period ((RangeFromDate - RangeToDate) of let's 
say 6
>     > months.
>     > >
>     > > 2) use the 'best values' to backtest this stock during a 
certain
>     > > period; this period directly follows the optimisation 
period
>     > >
>     > > 3) repeat steps 1 and 2 in a 'rolling mode'.
>     > >
>     > > My question is : how can I get the optimised values in my
>     > backtest ??
>     > > I guess one way is through the export of the results list 
to CSV
>     > file
>     > > but maybe there is a smarter way ??
>     > >
>     > > Thanks
>     > > Leo
>     > >
>     > >
>     > >
>     > >
>     > > Yahoo! Groups Sponsor
>     > >
>     > > ADVERTISEMENT
>     > >
>     > >
>     >
>     
<http://rd.yahoo.com/M=243376.3035052.4359433.1927555/D=egroupweb/S=1
7
>     > 05
>     > >
>     >
>     
632198:HM/A=1414910/R=0/*http://ad.doubleclick.net/jump/N879.ameritra
d
>     > e.
>     > > yahoo/B1054521.29;sz=300x250;adc=zhs;ord=1046531329358607?>
>     > >
>     > > <http://us.adserver.yahoo.com/l?
>     > M=243376.3035052.4359433.1927555/D=egrou
>     > > pmail/S=:HM/A=1414910/rand=487282882>
>     > >
>     > > 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 the Yahoo! Terms of
>     Service
>     > > <http://docs.yahoo.com/info/terms/> .
>     >
>     >
>     >
>     > Yahoo! Groups Sponsor
>     >
>     > ADVERTISEMENT
>     >
>     >
>     
<http://rd.yahoo.com/M=243376.3035052.4359433.1927555/D=egroupweb/S=1
7
>     05
>     >
>     
632198:HM/A=1414910/R=0/*http://ad.doubleclick.net/jump/N879.ameritra
d
>     e.
>     > yahoo/B1054521.29;sz=300x250;adc=zhs;ord=1046543713844329?>
>     >
>     > <http://us.adserver.yahoo.com/l?
>     M=243376.3035052.4359433.1927555/D=egrou
>     > pmail/S=:HM/A=1414910/rand=941839877>
>     >
>     > 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 the Yahoo! Terms of 
Service
>     > <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 the Yahoo! Terms of 
Service.
> 
> 
> 
> 
> 
>   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 the Yahoo! Terms of 
Service.
> 
> 
> 
> 
>         Yahoo! Groups Sponsor
>               ADVERTISEMENT
> 
> 
> 
> 
>   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 the Yahoo! Terms of 
Service.


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Rent DVDs Online-No late fees! Try Netflix for FREE!
http://us.click.yahoo.com/bbvVKB/oEZFAA/46VHAA/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/