|
PureBytes Links
Trading Reference Links
|
Ace,
My method is really slow. But it has merit. If you replace Val
[Barcount-1] with Val (without referring to the last bar), and put
addtocomposite(), you can generate the historical ROC ranking for
the whole database. Of course, it is really slow. For a database
with two thousand stock, it took me 8 hours to finish one ROC
ranking. But anyway, it is one time job. After that you can do
whatever backtesting for ROC ranking concept using the generated
ranking data.
I think your method also needs a lot of calculation. The most CPU
intensive portion in both yours and mine is the loop
for(i=0, (sym = strExtract(list, i)!=""), i++)
It will load the whole database for every symbol.
So the calculation amount is nxn
The real problem is it is not practical for me to do daily ranking,
because the exploration for the last bar also needs similar amount
of time.
The only solution is AB only load one day data when it does one day
exploration. but unfornately this function is not availabe in AB.
Cheers,
Gary
--- In amibroker@xxxxxxxxxxxxxxx, "acesheet" <acesheet@xxxx> wrote:
> Gary,
>
> I see that you are basically placing the ROC of whatever stock you
> are analyzing in the particular order within the 'ROC stackup' for
> the day under consideration.
>
> What you are doing is pretty much the same thing I did, but in a
> different way. So it is a pure statistical stackup in that sense
and
> is the proper way of performing a ranking. It is very slow and
> memory consuming for a large watchlist though. I like the results,
> but it sure takes some time! It seems correct.
>
> In terms of Exploring for RS stocks you'll find the code I
presented
> is much faster, once the calibration indices are created (~xroc
and
> ~yrank). It does the same thing that you did, but it only needs to
> create the stackup once. My code creates a reference index that
> stores the stackup information for today, whereas yours creates it
> for each stock you are scanning, which is why it takes so long.
Mine
> takes a long time, but only once per day. The RS calculation then
> references the stored information, which makes it a lot quicker.
>
> This is good work though. I'll be able to refer back to it for
ideas
> I think.
>
> Take that for what its worth.
>
> -ace
>
>
> --- In amibroker@xxxxxxxxxxxxxxx, "gary_tiger2001"
> <gary_tiger2001@xxxx> wrote:
> > Ace,
> > Please check my code below, it looks much simpler. It can rank
from
> > 1 to 100. But the exploration is really slow for big database. I
am
> > trying to reduce the calculation amount before I can use it for
> > daily exploration.
> >
> > //replace your watchlist number with 15
> > list = CategoryGetSymbols( categoryWatchlist, 15 );
> > p = 20;
> > Count=0;
> > rank = 0;
> > symVal = 0;
> > relval = 0;
> > Ownval = ROC(C,p);
> > for( i = 0; ( sym = StrExtract( list, i ) ) != ""; i++ )
> > {
> > symVal = Nz(Foreign( sym, "Close" ));
> > relval = ROC(symval, p);
> > n=Nz(IIf(relVal!=0,1,0));
> > Count[BarCount-1]=Count[BarCount-1]+n[BarCount-1];
> > if(relval[BarCount-1] > Ownval[BarCount-1])
> > rank[BarCount-1] = rank[BarCount-1]+1;
> > }
> > AddColumn(Count,"count",1.0);
> > AddColumn(rank,"rank",1.0);
> > rank[BarCount-1] = int(100*(Count[BarCount-1]-rank[BarCount-
> 1])/Count
> > [BarCount-1]);
> >
> > Filter = 1;
> > AddColumn(Ownval,"ROC",1.2);
> > AddColumn(rank,"rank",1.0);
> >
> >
> > Gary,
> >
> >
> > --- In amibroker@xxxxxxxxxxxxxxx, "acesheet" <acesheet@xxxx>
wrote:
> > > JOE,
> > >
> > > I'm 99% certain the way I propose to calculate the IBD RS *is*
> the
> > > way that it is done in terms of comparing relative strength on
a
> > > purely statistical basis. If you look in Excel at the
> PERCENTRANK
> > > function you'll see that it does essentially the same thing as
> > what
> > > I designed in AFL for this code.
> > >
> > > Other formulations may try to approximate the IBD number -
I've
> > > designed a few of those myself - but they aren't really doing
> what
> > > the essence of the 1-99 comparison is. The code I showed does
> > > exactly what a RS Rank compared to all stocks in the database
> and
> > is
> > > not really an approxiamtion, except for the 11 point curve fit
> of
> > > the CPD curve.
> > >
> > > The only difference between the one I show and IBD's is that
> they
> > do
> > > some weighting of either more than one ROC going into the CPD
> > > comparison. I don't know what IBD's weightings are, but I'm
> pretty
> > > sure it doesn't matter all that much as long as the idea finds
> > high
> > > RS stocks to trade. In other words I really don't think it
> matters
> > > if IBD says a stock is an 88 and this calculation says its a
84.
> > >
> > > Have fun.
> > >
> > > -ace
> > >
> > > --- In amibroker@xxxxxxxxxxxxxxx, "Joseph Landry"
> <jelandry@xxxx>
> > > wrote:
> > > > Ace - I subscribe to the forum digest so I don't get any
> > > > attachments, and I'm wondering if you could forward the AFL
> code
> > > to
> > > > me directly?
> > > >
> > > > The folks over in the TC2000 area have a way of calculating
> what
> > > > they say is IBD relative strength number and as soon as I
find
> > it
> > > > and the underlying rationale I'll post it here.
> > > >
> > > > Thanks
> > > > JOE
> > > >
> > > > --- In amibroker@xxxxxxxxxxxxxxx, "gary_tiger2001"
> > > > <gary_tiger2001@xxxx> wrote:
> > > > > Thanks Ace, Harvey & Jason,
> > > > >
> > > > > I will study the code...
> > > > >
> > > > > Cheers,
> > > > > Gary
> > > > > --- In amibroker@xxxxxxxxxxxxxxx, "acesheet"
<acesheet@xxxx>
> > > wrote:
> > > > > > Gary,
> > > > > >
> > > > > > There's a couple of different ways of doing it. Some
more
> > time
> > > > > > consuming, some less. One way is to compare the returns
> > using
> > > > ROC
> > > > > to
> > > > > > a reference index (like SP500 or Value Line for US
> stocks).
> > If
> > > a
> > > > > > stock's performance significantly beats out the index,
> then
> > > its
> > > > a
> > > > > > safe bet that it has high relative strength. I've used
> this
> > > > > concept
> > > > > > for back testing.
> > > > > >
> > > > > > Another way is to use what's called the 'Cumulative
> > > Probability'
> > > > > of
> > > > > > the market returns. See post # 55179. If you have a
large
> > > enough
> > > > > > database you should be able to get these AFL scripts to
> work
> > > > very
> > > > > > well in terms of comparison with IBD's RS ranks.
Currently
> I
> > > > only
> > > > > > have a script that uses "today's" data however, so it
> won't
> > be
> > > > so
> > > > > > hot for backtesting. This method of ranking works very
> well.
> > > > > >
> > > > > > You could take the idea shown in post 55179 and adapt it
> to
> > > > > collect
> > > > > > the data over time and create several historical ROC
> indices
> > > for
> > > > > the
> > > > > > 10 point curve fit. Something like
> > > > > ~xroc1,~xroc10...~xroc90,~xroc99.
> > > > > > It will probably take AB a very long time to crunch the
> > > numbers,
> > > > > > however.
> > > > > >
> > > > > > Hope that gives you some ideas.
> > > > > >
> > > > > > -ace
> > > > > >
> > > > > > --- In amibroker@xxxxxxxxxxxxxxx, "gary_tiger2001"
> > > > > > <gary_tiger2001@xxxx> wrote:
> > > > > > > Happy new year to all of you.
> > > > > > >
> > > > > > > I wish to test the idea from IBD, like the relative
> > > strenghth
> > > > > > > ranking, etc. I cannot figure out how to do it in AB.
> > > > Basically,
> > > > > I
> > > > > > > have a universe of stocks. Is it possible to rank the
> > > strength
> > > > > of
> > > > > > > stocks compared to my stock universe, from 1 - 99?
> > > > > > >
> > > > > > > Cheers,
> > > > > > > Gary
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 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:
http://docs.yahoo.com/info/terms/
|