Hello All

Borrowing from the idea of a RSI indicator with extremes normalized for bull and bear phases (from June 2009 issue of TASC by Giorgos Siligardos, copyright 2009 Technical Analysis Inc, all rights reserved), I have attempted to code an RSI that would have it's extreme boundaries automatically adjusted according to four "phases" of an ATR-based indicator, as follows:

indic:=ATR(21)/((H+L)/2)*100

The general formula for this normalized RSI is (rsi(prd)-rsi(prd) low boundary)/(rsi(prd) high boundary-rsi(prd) low boundary).

Here is my attempt at the code:

hi1:=Highest(ValueWhen(1,indichi2:=Highest(ValueWhen(1,1hi3:=Highest(ValueWhen(1,2hi4:=Highest(ValueWhen(1,indic>4,RSI(14)));

lo1:=Lowest(ValueWhen(1,indiclo2:=Lowest(ValueWhen(1,1lo3:=Lowest(ValueWhen(1,2lo4:=Lowest(ValueWhen(1,indic>4,RSI(14)));

hilo1:=hi1-lo1;
hilo2:=hi2-lo2;
hilo3:=hi3-lo3;
hilo4:=hi4-lo4;

If(indic
The problem is that it only plots the latest "phase" of the ATR indicator.

Would anybody like to take a swing at it?

Many thanks.

ZigZag

Somehow when I copy-paste the code it is not showing right in Forum. I apologize.

The idea is to normalize RSI for each phase (when ATR(21)/(H+L)/2*100 is below 1, between 1 and 2, between 2 and 4 and above 4).

Any help is appreciated.

 Here is my stab at it. Might give you some ideas.

ti := 100;
p1:=ATR(21)/((H+L)/2)*100;
p1f := 100/(HHV(p1,ti)-LLV(p1,ti));
p1m := (p1-LLV(p1,ti))*p1f;
{----------------------------------------}
a1:=ValueWhen(1,p1m=0,RSI(14));
a2:=ValueWhen(1,Cross(p1m,25),RSI(14));
a3:=ValueWhen(1,Cross(p1m,50),RSI(14));
a4:=ValueWhen(1,Cross(p1m,75),RSI(14));
b1:=(RSI(14)-a1/(a2-a1));
b2:=(RSI(14)-a2/(a3-a2));
b3:=(RSI(14)-a3/(a4-a3));
b4:=(RSI(14)-a1/(a2-a1));
c1:=If((p1m<25),b1,
    If((p1m>25) AND (p1m<50),b2,
    If((p1m>50) AND (p1m<75),b3,
    b4)));
If((c1>100),100,If((c1<0),0,c1))

johnl

Thanks for your code - you are clearly a talented code-writer.  What I actually had in mind was to have the following indicator plotted (sorry for having been unclear):

(rsi(c,14)-lowest(rsi(c,14)))/(highest(rsi(c,14))-lowest(rsi(c,14)))

using as the "lowest" and "highest" reference data the highest/lowest value of RSI(C,14) according to the range where the value of the formula below falls (ranges: x<=1, 1<x<=2, 2<x<=4 and x>4).

ATR(21)/((H+L)/2)*100

Not sure your code is doing that (I don't mean "not sure" as in "I don't think so"...).

My original (falied) attempt below:

indic:=ATR(21)/((H+L)/2)*100;  

hi1:=Highest(ValueWhen(1,indic<=1,RSI(14)));

hi2:=Highest(ValueWhen(1,1<indic<=2,RSI(14)));

hi3:=Highest(ValueWhen(1,2<indic<=4,RSI(14)));

hi4:=Highest(ValueWhen(1,indic>4,RSI(14)));

 

lo1:=Lowest(ValueWhen(1,indic<=1,RSI(14)));

lo2:=Lowest(ValueWhen(1,1<indic<=2,RSI(14)));

lo3:=Lowest(ValueWhen(1,2<indic<=4,RSI(14)));

lo4:=Lowest(ValueWhen(1,indic>4,RSI(14)));

 

hilo1:=hi1-lo1;

hilo2:=hi2-lo2;

hilo3:=hi3-lo3;

hilo4:=hi4-lo4;

 

indic2:=If(indic<=1,(RSI(14)-lo1)/hilo1,If(1<indic<=2,(RSI(14)-lo2)/hilo2,If(2<indic<=4,(RSI(14)-lo3)/hilo3,(RSI(14)-lo4)/hilo4)));

 

 

Thanks again.  Any help from Forum members is appreciated.

 

ZigZag

Hi ZigZag,

This code works as an Indicator:-

{Wabbit's "Division by Zero" Error Trap}
zero:=Power(10,-20);

data:=ATR(21)/MP()*100;

hi1:=Highest(ValueWhen(1,data<=1,RSI(14)));
hi2:=Highest(ValueWhen(1,data>1 AND data<=2,RSI(14)));
hi3:=Highest(ValueWhen(1,data>2 AND data<=4,RSI(14)));
hi4:=Highest(ValueWhen(1,data>4,RSI(14)));

lo1:=Lowest(ValueWhen(1,data<=1,RSI(14)));
lo2:=Lowest(ValueWhen(1,data>1 AND data<=2,RSI(14)));
lo3:=Lowest(ValueWhen(1,data>2 AND data<=4,RSI(14)));
lo4:=Lowest(ValueWhen(1,data>4,RSI(14)));

hilo1:=hi1-lo1;
hilo2:=hi2-lo2;
hilo3:=hi3-lo3;
hilo4:=hi4-lo4;

If(data<=1,(RSI(14)-lo1)/MAX(hilo1,zero),
If(data>1 AND data<=2,(RSI(14)-lo2)/Max(hilo2,zero),
If(data>2 AND data<=4,(RSI(14)-lo3)/Max(hilo3,zero),
(RSI(14)-lo4)/Max(hilo4,zero))));

 

I have used the the in-built Median Price Indicator which is the same as (H-L)/2.

Cheers,

oz 

Oztrader

Many thanks for your post.  I actually have the same problem with your code that I have with mine (posted previously): the indicator only starts plotting after ATR(21)/MP()*100 crosses above 4 for the first time (on 10/6/08 on daily SPY, for instance).  I wonder if you would have the same problem.

I can't seem to be able to figure out why this is happening...

Thanks again for the help.

 

ZigZag

ZigZag,

Yes, although I hadn't noticed this until you raised the issue, but the indicator does not start plotting until all 4 of the range values for the "indic" have been established and I'm not sure why this happens.

Out of interest this code
"indic > 1 AND indic <= 2"
gives a different result than using this code
"1 < indic <=2"
when used in your formula!

This can be checked by displaying the following indicators on a chart:-

{Ind1}
data:=ATR(21)/((H+L)/2)*100;
If(data<=1,1,
If(data>1 AND data<=2,2,
If(data>2 AND data<=4,3,4)))

{Ind2}
indic:=ATR(21)/((H+L)/2)*100;
If(indic<=1,1,
If(1<indic<=2,2,
If(2<indic<=4,3,4)))

oz

Oz

thanks for the post.  i haven't tried plotting but "indic > 1 AND indic <= 2" is mathematically the same as "1 < indic <=2".  it is curious that they would plot differently...

i am really perplexed as to why the indicator is plotting the way it is.

would anybody else like to take a swing at it?

ZigZag

zigzag:

i haven't tried plotting but "indic > 1 AND indic <= 2" is mathematically the same as "1 < indic <=2".  it is curious that they would plot differently...

Have a look in the MS Users Manual for the operator precedence.  In the expression:

Code:

1 < indic <=2

the LT expression will be evaluated first, resulting in a binary 0 or 1; then the LTE expression will be evaluated, so it will be evaluated:

Code:

(1 < indic) <= 2; {so, depending on user definition requirements, it's the same as saying something like} (0 <= 2) AND (1 <= 2);

so the expression will always be true, and, is not the same as:

Code:

1 < indic AND indic <=2

wabbit

---

"The question of whether a computer can think is no more interesting than the question of whether a submarine can swim."
Edsgar W. Dijkstra

 

MS: 6.52 EOD, 7.x EOD, 8.0 PRO, 9.2 PRO w/QC, 10 PRO w/QC C, 11 PRO w/QC & MDK
For custom MetaStock programming : http://www.wabbit.com.au
My SkyPE status :
My SkyPE account : wabbit.com.au

Hi Wabbit

Thanks for your post - educational, as usual.

Would you care to take a look at the code for the indicator?  It is not working as I expect and I cannot figure out why...

Thanks.

ZigZag

Hi zz,

You will need to deal with the periods when ValueWhen() returns N/A.  You could use a whole pile of calls to the forum.sum function but this would be rather unwieldy and probably quite slow; maybe even slow enough to warrant reverting to using a PREV based latch?? (Although I haven't looked too closely as to the specifics of how you'd employ it).

zigzag:

It is not working as I expect and I cannot figure out why..

Can you please post an annotated chart and more details?  Comments like this aren't much use in bug-hunting!



wabbit

---

"The question of whether a computer can think is no more interesting than the question of whether a submarine can swim."
Edsgar W. Dijkstra

 

MS: 6.52 EOD, 7.x EOD, 8.0 PRO, 9.2 PRO w/QC, 10 PRO w/QC C, 11 PRO w/QC & MDK
For custom MetaStock programming : http://www.wabbit.com.au
My SkyPE status :
My SkyPE account : wabbit.com.au

thanks Wabbit
the problem I am having is that the indicator doesn't start plotting until ATR(21)/MP()*100 first crosses above four.
how would one post a chart in Forum?
ZigZag

http://forum.equis.com/forums/thread/19928.aspx


wabbit

---

"The question of whether a computer can think is no more interesting than the question of whether a submarine can swim."
Edsgar W. Dijkstra

 

MS: 6.52 EOD, 7.x EOD, 8.0 PRO, 9.2 PRO w/QC, 10 PRO w/QC C, 11 PRO w/QC & MDK
For custom MetaStock programming : http://www.wabbit.com.au
My SkyPE status :
My SkyPE account : wabbit.com.au

ZZ,

Using the free forum.dll to eliminate the N/A from the ValueWhen() function calls:

Code:

data:=ATR(21)/MP()*100; i:=RSI(14); hi1:=Highest(ExtFml("forum.sum",ValueWhen(1,data<=1,i),1)); hi2:=Highest(ExtFml("forum.sum",ValueWhen(1,data>1 AND data<=2,i),1)); hi3:=Highest(ExtFml("forum.sum",ValueWhen(1,data>2 AND data<=4,i),1)); hi4:=Highest(ExtFml("forum.sum",ValueWhen(1,data>4,i),1)); lo1:=Lowest(ExtFml("forum.sum",ValueWhen(1,data<=1,i),1)); lo2:=Lowest(ExtFml("forum.sum",ValueWhen(1,data>1 AND data<=2,i),1)); lo3:=Lowest(ExtFml("forum.sum",ValueWhen(1,data>2 AND data<=4,i),1)); lo4:=Lowest(ExtFml("forum.sum",ValueWhen(1,data>4,i),1)); hilo1:=hi1-lo1; hilo2:=hi2-lo2; hilo3:=hi3-lo3; hilo4:=hi4-lo4; zero:=Power(10,-8); 100* If(data<=1,(i-lo1)/Max(hilo1,zero), If(data>1 AND data<=2,(i-lo2)/Max(hilo2,zero), If(data>2 AND data<=4,(i-lo3)/Max(hilo3,zero), (i-lo4)/Max(hilo4,zero))));

Hope this helps.


wabbit

---

"The question of whether a computer can think is no more interesting than the question of whether a submarine can swim."
Edsgar W. Dijkstra

 

MS: 6.52 EOD, 7.x EOD, 8.0 PRO, 9.2 PRO w/QC, 10 PRO w/QC C, 11 PRO w/QC & MDK
For custom MetaStock programming : http://www.wabbit.com.au
My SkyPE status :
My SkyPE account : wabbit.com.au

ZZ,

As much as yesterday's code showed how to eliminate the N/A by using the forum.sum functionality, I don't think the code achieved the aim?  I haven't spent much time thinking about what the resultant curve should look like and as no chart has been posted, I hope you'll find this version is faster whilst still eliminating the ValueWhen() N/A problem and produces a curve more inline with what I think the objective should be?

Code:

ind:=RSI(14); data:=200*ATR(21)/(H+L); zero:=Power(10,-8); i:=Cum(IsDefined(data))=1; x:=data<=1; hi:=HighestSince(1,i OR x=0,ind); lo:=LowestSince(1,i OR x=0,ind); y1:=x*(ind-lo)/Max(hi-lo,zero); x:=data>1 AND data<=2; hi:=HighestSince(1,i OR x=0,ind); lo:=LowestSince(1,i OR x=0,ind); y2:=x*(ind-lo)/Max(hi-lo,zero); x:=data>2 AND data<=4; hi:=HighestSince(1,i OR x=0,ind); lo:=LowestSince(1,i OR x=0,ind); y3:=x*(ind-lo)/Max(hi-lo,zero); x:=data>4; hi:=HighestSince(1,i OR x=0,ind); lo:=LowestSince(1,i OR x=0,ind); y4:=x*(ind-lo)/Max(hi-lo,zero); 100*(y1+y2+y3+y4);

I have been putting in 14-16 hour days on a contract this week so I haven't been able to spend much time looking too closely at anything else but my own stuff; some comments about what you think is working and what isn't will allow me to dedicate more time to this or other members, otherwise I will consider this thread resolved?


Hope this helps.

wabbit

---

"The question of whether a computer can think is no more interesting than the question of whether a submarine can swim."
Edsgar W. Dijkstra

 

MS: 6.52 EOD, 7.x EOD, 8.0 PRO, 9.2 PRO w/QC, 10 PRO w/QC C, 11 PRO w/QC & MDK
For custom MetaStock programming : http://www.wabbit.com.au
My SkyPE status :
My SkyPE account : wabbit.com.au