|
PureBytes Links
Trading Reference Links
|
<FONT face=Arial color=#0000ff
size=2>Charles,
<FONT face=Arial color=#0000ff
size=2>
I
performed some test and using static arrays (not available in AB yet) and
at first glance it appears that I can compute the RS for a 3755 x
100 stock matrix (375500 comparisons compared to your 490000) in less
than a minute. You could probably duplicate the procedure using
the OSAKA DLL table functions but i don't have time for that. Perhaps somebody
else can provide afl code for the custom functions, else you may be able
to substitute OSAKA code for the StaticArraySet functions the code goes
something like that below, or you may email me privately <A
href="">psytek@xxxxxxxx for the code i
used.
<FONT face=Arial color=#0000ff
size=2>
Best
regards,
<FONT face=Arial color=#0000ff
size=2>herman
Filter
= Status("LastBarInTest");SetOption("NoDefaultColumns",True);WL1 = 0;
//WL2 is selected with Filter;
//
Calculate the RS1 for WL1 just once, when the first stock in the filter is
testedif(Status("StockNum") == 0 ) StaticVarSet("IsInitialized",
0);
if(
StaticVarGet("IsInitialized") == 0 ) { list1 =
CategoryGetSymbols( categoryWatchlist , WL1); for( i = 0; ( sym =
StrExtract( list1, i ) ) != ""; i++
) { SetForeign(sym); MyArray =
C; sRelStren = IIf((IsNull(ROC(myArray,
260)) == False), ROC(myArray, 260) + ROC(myArray,
195) + ROC(myArray, 130) + 2 * ROC(myArray, 65),
IIf ((IsNull(ROC(myArray, 195)) ==
False), ROC(myArray, 195) + 2 * ROC(myArray, 130) + 2
* ROC(myArray, 65), 2 * ROC(myArray, 130) + 3 *
ROC(myArray, 65) )); VarName =
"Sym"+NumToStr(i,1.0,False);<FONT
color=#ff0000>// StaticArraySet(sRelStren ,"RS-"+VarName); // Save
the RS in a Static Array (custom
function) } StaticVarSet("NumTickers",i); StaticVarSet("IsInitialized",
1); RestorePriceArrays(); }
Ticker
= Name();AddTextColumn(Ticker," ",1.0);
//
Calculate the RS2 for each stock in the Filter processed by the AA
ExplorereMyArray = C;sRelStren2 = IIf((IsNull(ROC(myArray,
260)) == False), ROC(myArray, 260) + ROC(myArray, 195)
+ ROC(myArray, 130) + 2 * ROC(myArray, 65), IIf
((IsNull(ROC(myArray, 195)) == False), ROC(myArray, 195) + 2 *
ROC(myArray, 130) + 2 * ROC(myArray, 65), 2 * ROC(myArray, 130)
+ 3 * ROC(myArray, 65) ));
// Do
whatever you want with the two RSs, I just multiply as a test.for( i = 0;
i<StaticVarGet("NumTickers"); i++ ) { VarName =
"Sym"+NumToStr(i,1.0,False);// sRelStren1 =
StaticArrayGet(VarName); // Recall RS1 from the Static Array (custom
function) AddColumn(sRelStren1 *sRelStren2 ,VarName
,1.2); }
<FONT face=Arial color=#0000ff
size=2>
<FONT face=Tahoma
size=2>-----Original Message-----From: firstview2000
[mailto:firstview2000_1999@xxxxxxxxx]Sent: Thursday, August 05,
2004 2:20 AMTo: amibroker@xxxxxxxxxxxxxxxSubject:
[amibroker] Re: any ideas to reduce runtime for the price RS ranking
code?Herman,Your afl code also consumed too
LONG time to finish: I ran it for 18 minutes, and the estimated time left
is 19 minutes. So, it will be in the range of 40 minutes to complete 700
stocks based on this speed.I also tested your AFL in a 517 stock
watch list, it took 22 minutes to complete.For a 53 stock watch
list, it took 13 seconds to complete.I thought about your
suggestion #1, but could not find any good way to reduce the loops. There
are a lot of un-necessary repeating calculations in my AFL: the RS of each
stock is repeated 700 times. To avoid this, I need to store RS values for
700 stocks. I think it is not good idea to create 700 ATC composites for
this purpose. what do you think?Other information:The
status line of the AA window stated "Running ..." all the time during the
exploration.In my AFL preference setting, the quickAFL for indicator
is checked.RegardsCharles--- In
amibroker@xxxxxxxxxxxxxxx, "Herman van den Bergen" <psytek@xxxx>
wrote:> Don't know if it can be improved but i'll think about it; I
like this kind> of challenges. Coding is more fun than trading,
ooops: i should have said> that in public :-)> > For
comparison you may look at the code below which performs a similar>
action, it calculates the correllation between each of a group of
stocks.> Run it on your 700 and see how long it takes. If it is
much shorter than the> time-consuming code is not the StrExtract()
and you should look at your> other lines of code. I would be
interested to know how that compares on your> computer. It take
less than 10 seconds to correlate 100x100 stocks for me.> >
SetOption("NoDefaultColumns",True);> Filter =
Status("LastBarInTest");> list = GetCategorySymbols( categoryWatchlist,
0);> AddTextColumn(Name(),"Correlation",1.0);> Ticker1=
Name();> for( Col=0; (Ticker2=StrExtract( List, Col))!= "";
Col++)> {> Var2 = Foreign(Ticker2,"C");> Corr =
Correlation( C, Var2, 8);> Color = IIf(Corr>0, colorBrightGreen,
IIf(Corr<0, colorRed,colorWhite));> Color =
IIf(Ticker1==Ticker2, 1, Color);> AddColumn( Corr, Ticker2, 1.3, 1,
Color);> }> > 1) You may have calculations inside your
loop that are repeated for each> comparison but that need only be
calculated one time for each stock, outside> the loop. You could
use an additional loop through the watchlist to create> these
variables using SetVar. That way they would be calculated 700x
instead> of 700^2 x... that would give a 700x improvement for those
lines of code.> Your selfRelStrength function comes to mind
...> > 2) some things that come to mind is that you are
retrieving 700 x 700 x 3000> (assuming 10 yrs data) bars from your
hard disk. You can observe this by> noting the status line at the
bottom of the AB screen. The AA doesn't use> QuickAFL so perhaps it
would work faster to run this indicator in an> indicator using two
loops.> > 3) Your code could also create a OSAKA-like table for
all stocks and work> from there, with the same objective.>
> interesting problem, let me know how you progress, I'll let you
know if> something else comes to mind.> >
herman> > > > -----Original
Message-----> From: firstview2000
[mailto:firstview2000_1999@xxxx]> Sent: Wednesday, August
04, 2004 1:32 AM> To:
amibroker@xxxxxxxxxxxxxxx> Subject: [amibroker] Re: any
ideas to reduce runtime for the price RS> ranking code?>
> > Herman,> > I need all
available values for the RS ranking for backtesting>
purpose. I could plot a chart showing the rs ranking if I need so.>
> My code performs the following loops:>
> the outer loop is for all 700 stocks in database to have
their> rankings;> the inner loop is for
comparing one stock to every other 700 stocks> >
So, there are 700x700=490,000 loops. I don't know if we can
reduce> the number of loops here.>
> In addition, if stock quotes are 1000 bars in length,
this ranking> formula will have 490K x 1000 = 490
million calculations to complete.> > It takes
over 40 minutes to complete, which suggests about 10>
million RS calculations per minute. I don't know if this is
a> reasonable rate, but I just think 40minutes is too
long.> > Thank for your help.>
Charles> Check AmiBroker web page
at:<A
href="">http://www.amibroker.com/Check
group FAQ at: <A
href="">http://groups.yahoo.com/group/amibroker/files/groupfaq.html
Check AmiBroker web page at:
http://www.amibroker.com/
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.
|