|
PureBytes Links
Trading Reference Links
|
hmmm,
I hope your doctor knows what you are up to, he might want to prescribe
a dose proportional to the number of stocks you want to
test...
<FONT face=Arial color=#0000ff
size=2>
But
thanks d, yes i am reading it :-) however my application is slightly different.
I am not really performing a broad based scan. I use Backtest Reports and
TradeLists from existing systems to try and use the data in a second Backtest on
the sasme population and/or use it to supplement my RT operations. I haven't
made much progress yet but imho it "has to work"; there is no difference in
using information from an indicator as compared to using information from a
Backtest Report. To get some of the BT-statistics into arrays i may need to run
sequential backtests however it is unlikely that I need to go back in time
too far - certainly not to the extend of 1500 bars (the equivalent task of you
scanning 1600 stocks). Also some of the stats do not change by the bar, it would
be adequate to step in periods.
<FONT face=Arial color=#0000ff
size=2>
The
ideal solution would be to have access to the Statistics BEFORE the report
file is created, for example at the end of the Backtest when all arrays are
still in memory. Then we could extract whatever we need and maintain just one
summary file and not save the standard reports.
<FONT face=Arial color=#0000ff
size=2>
One
question Dale, how did you intend on extracting/sorting through the backtest
reports?
You
might also have a closer look at how the stats of interest are calculated,
some stats can only meet your criteria under certain rudimentary price
conditions. If you can define these conditions you can set the filter to exclude
stocks that do not meet the more basic prerequisites.
<FONT face=Arial color=#0000ff
size=2>
<FONT face=Arial color=#0000ff
size=2>Another consideration ... i might be wrong here ... is that the
Backtester tests full histories. Correct me if i am wrong, or perhaps it is
dependent on the Quick afl setting? Anyway with your automation skills you could
export limited data histories, say only for the last 30 bars, and create a new
"limited data" Database. Running your tests on such a limited database might go
faster but give you all the information you need. I may be totally wrong here...
it just appears logical to me.
<FONT face=Arial color=#0000ff
size=2>
<FONT face=Arial color=#0000ff
size=2>herman
<FONT face=Arial color=#0000ff
size=2>
<FONT face=Arial color=#0000ff
size=2>
<FONT face=Tahoma
size=2>-----Original Message-----From: dingo
[mailto:dingo@xxxxxxxxxxxx]Sent: Wednesday, April 21, 2004 12:31
PMTo: _amibroker_yahooSubject: [amibroker] Data mining
And the Individual Backtester - WARNING!Here's a
warning to any of you who might be thinking of taking a largenumber of
tickers and running a number back tests (not optimizations,explorations
nor scans) to generate some of the wonderful new statistics(Sharpe Ratio,
etc). And especially if you're running the "Individual"
backtest!Under the above scenario the back tester will create a
subfolder in theAmibroker\Reports folder for each run and when using
the "Individual" backtester a separate folder for EACH ticker. You
need to be aware of thisespecially if you are not looking at your Reports
folder on a frequentbasis. The number of folders can become huge and the
effects on yourcomputer can become severe! This is especially true
if you use automationto crank out untold numbers of back tests (are you
reading this Herman?) togenerate statistics for each ticker. I had
this all figured out in advance and put in code to delete the foldersas my
automated tests went along but I introduced a bug in that part of
theprogram and left a test running overnight on 2 computers.
YIKES! When Isat down this am to see what juicy results I might have
gotten both machineswere locked up tight as a drum. One computer had
50,000 (yep 50k)subfolders and the other had over 20,000 in the reports
folder. Justgetting rid of them so I could get my machine back has
taken a couple ofhours of watching the stupid hourglass move sand back and
forth.So - let this be a backtester emptor to those data miners out
there!One other thing - the creation/deletion of all of those folders
when runningthe individual back test can really slow down the
computer. Based on onetest I dreamed up where I take 1,500 tickers
and run a very intense seriesof individual back tests over 10 years worth
of data should take around 80hours of compute time. And this is on a very
fast computer. (Please save thecriticism on how off base this might be and
how insane I am to do somethingso complicated - My doctor has some
new meds for me).dSend 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
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
Yahoo! Groups Sponsor
ADVERTISEMENT
Yahoo! Groups Links
To visit your group on the web, go to:http://groups.yahoo.com/group/amibroker/
To unsubscribe from this group, send an email to:amibroker-unsubscribe@xxxxxxxxxxxxxxx
Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.
|