Stocks & Commodities V.

28:10 (38-44): A Smoothed RSI Inverse Fisher Transform by Sylvain Vervoort

Clarity In Decision-Making

A Smoothed Rsi
Inverse Fisher Transform
This indicator gives clear entry and exit signals, so your applied an inverse Fisher transform to the relative strength
ROY WEIMANN

trading decisions are easier. index (Rsi), changing what was a good turning-point indicator
to one that fine-tuned the timing of possible entry and exit
by Sylvain Vervoort points. Here is the inverse Fisher transform formula:

T he Fisher transform has been a familiar concept to

Stocks & Commodities readers in recent years
following its introduction in 2002 by frequent S&C
y = (Exp(2 * x) - 1)/(Exp(2 * x) + 1)

where x is the input value and y is the transformed value.

contributor John Ehlers. Its creator, Ronald Fisher, Figure 1 shows a plot of the inverse Fisher transform. The
was one of the leading scientists of the 20th century, transform creates boundaries to keep the output value in the
making major contributions to statistics, evolution- range between 1 to +1. Input values larger than 2 generate
ary biology, and genetics. Ehlers 2004 followup a result close to 1, while input values less than 2 generate a
Copyright Technical Analysis Inc.
Stocks & Commodities V. 28:10 (38-44): A Smoothed RSI Inverse Fisher Transform by Sylvain Vervoort
RSI value X-input Y-output Normalized
1 100 5.0 1.00 100
Y
90 4.0 1.00 100
0.5 80 3.0 1.00 100
70 2.0 0.96 98

0 65 1.5 0.91 95
60 1.0 0.76 88
55 0.5 0.46 73
-0.5
50 0.0 0.00 50
45 -0.5 -0.46 27
-1 40 -1.0 -0.76 12
X
35 -1.5 -0.91 5
-5 -4 -3 -2 -1 0 1 2 3 4 5
30 -2.0 -0.96 2
Figure 1: inverse fisher transform. The transform creates boundaries to
keep the output value in the range of 1 to +1. Input values larger than 2 generate 20 -3.0 -1.00 0
a result close to 1, and input values less than 2 generate a result close to 1. This
boundary characteristic is useful with any indicator moving between two fixed values,
10 -4.0 -1.00 0
like the RSI. 0 -5.0 -1.00 0
Figure 2: conversion of rsi data. Here you see the conversion of RSI data
to an inverse Fisher transform output and back to a normalized RSI value output.
result close to 1. This boundary characteristic is useful with Comparing the first and last columns, you can see how original RSI values will be
plotted after applying the inverse Fisher transform.
any indicator moving between two fixed values, like the rela-
tive strength index (Rsi).
But the input data for the inverse Fisher transform formula
needs to be in the range of 5 to +5. The normal Rsi is in the 14
range of zero to 100, but we can convert this to a range of 5 13
to 5, using the following formula: 12
11
x = 0.1 * (Rsi value 50) 10
9
If required, you can convert the output of the inverse 8
Fisher transform back to the range of zero to 100 using this
7
formula:
6
Rsi normalized = 50 * (y + 1) 90
70
This makes it easier to compare the normalized Rsi data 50
with the original Rsi input data when overplayed in the same 30
window. 10

RSI to inverse Fisher transform

90
70
Figure 2 illustrates the conversion of Rsi data to an inverse 50
Fisher transform output and back to a normalized Rsi value 30
output. Comparing the first and last columns, you can see 10
how original Rsi values will be plotted after applying the 90
inverse transform. 70
Rsi values above 60 will be squeezed into the top 12% of 50
30
the range, and Rsi values below 40 will be squeezed into the 10
bottom 12% of the range. The transform causes the Rsi transi-
tion from 40 to 60, to be plotted as a very sharp swing. 90
70
In Figure 3 you see a chart starting with a flat price move, 50
followed by an up and down price move. You can compare 30
10
the original Rsi in blue with its inverse Fisher transform in
METASTOCK

red (first and second graphs below the price plot). Sept. Oct. Nov. Dec. 2004 Feb.
In his 2004 article, John Ehlers used a nine-day weighted FIGURE 3: RSI TO SVE INVERSE FISHER RSI. Here you see the RSI, inverse Fisher
moving average on the inverse Fisher transform with a five- transform, John Ehlers inverse Fisher transform, and the SVE inverse Fisher RSI.
day Rsi to create clearer buy and sell points. This is the green Which one gives you faster buy and sell signals with clear turning points?

Copyright Technical Analysis Inc.

 Stocks & Commodities V. 28:10 (38-44): A Smoothed RSI Inverse Fisher Transform by Sylvain Vervoort
INDICATORS

CSX
38 {SVE_Rainbow_Weighted}
37 (5*Mov(C,2,W) +
36
35 4*Mov(Mov(C,2,W),2,W) +
34 3*Mov(Mov(Mov(C,2,W),2,W),2,W) +
33 2*Mov(Mov(Mov(Mov(C,2,W),2,W),2,W),2,W) +
32 Mov(Mov(Mov(Mov(Mov(C,2,W),2,W),2,W),2,W),2
31
30 ,W)+
29 Mov(Mov(Mov(Mov(Mov(Mov(C,2,W),2,W),2,W),2,
28 W),2,W),2,W)+
27 Mov(Mov(Mov(Mov(Mov(Mov(Mov(C,2,W),2,W),2,W
26
25 ),2,W),2,W),2,W),2,W)+
24 Mov(Mov(Mov(Mov(Mov(Mov(Mov(Mov(C,2,W),2,W
23 ),2,W),2,W),2,W),2,W),2,W),2,W)+
22 Mov(Mov(Mov(Mov(Mov(Mov(Mov(Mov(Mov(C,2,W),
21
2,W),2,W),2,W),2,W),2,W),2,W),2,W),2,W)+
100 Mov(Mov(Mov(Mov(Mov(Mov(Mov(Mov(Mov(Mov(
80 C,2,W),2,W),2,W),2,W),2,W),2,W),2,W),2,W),2,W
60 ),2,W)) / 20
40
20 I will make the Rsi period a free input choice with a
default value of 4:
100
80
60 RSIper := Input(RSI Period?,2,30,4);
40
20 Next, I call the Sve_Rainbow_Weighted function to
METASTOCK

get the smoothed closing price data:

Oct. Nov. Dec. 2006 Feb. Mar. Apr. May
FIGURE 4: SVE INVERSE FISHER RSI IN ACTION. Note how the SVE inverse Fisher RSI RbW:=Fml(Sve_Rainbow_weighted);
transform gives very clear buy and sell signals. This makes it easier for medium- to longer-
term traders to find additional entry points in a medium-term up move.
Now, I create the Rsi of the Sve_Rainbow_Weighted
data and convert the normal Rsi swing between zero
and 100 to a value that makes a limited swing between 5
curve in Figure 3 (the third graph below the price plot). When and +5 only:
the curve crosses above 27, it is a buying point, and when it
falls below 73, it is a selling point. x:= 0.1*(RSI(RbW,RSIper)-50);
What I want to achieve is the last black curve that you can
see on the bottom pane. If you compare this curve with the An extra measure you can take to limit the lagging even
green one, you will see that you generally get faster buy and more is to make use of the zero-lagging principle introduced
sell signals with very clear turning points reaching the zero by Joe Sharp in his January 2000 S&C article. I will make
and 100 levels. the value of the exponential moving average (Ema) also a
To create this black curve, I first tried using smoother input free input choice and calculate the zero-lagging value for the
values and not just the closing price for calculating the Rsi inverse Fisher transform with the following formula:
value. I tried a number of different smoothing techniques us-
EMAper := Input(Exponential Moving Average pe-
ing the median price, the typical price, and even the average
riod?,1,100,4);
heikin-ashi closing price. I got the best result with a smoothed EMA1:= Mov(x,EMAper,Exponential);
price, based on the rainbow chart. EMA2:= Mov(EMA1,EMAper,Exponential);
Difference:= EMA1 - EMA2;
Deriving the inverse Fisher RSI ZeroLagEma:= EMA1 + Difference;
with rainbow charts
In the July 1997 S&C, Mel Widner presented rainbow charts, The only thing left now is to calculate the inverse Fisher
and the Traders Tips in that issue provided the formulas for transform itself and let it swing between the standard Rsi
them. The MetaStock formula language shown is the basic values of zero and 100:
averaging technique I use to smooth the closing price. Because
I want to avoid the lagging effect as a result of this smoothing invfish:=((Exp(2*ZlEma)-1)/(Exp(2*ZlEma)+1)+1)*50;
as much as possible, I use a weighted moving average and invfish
give extra weight to the less smoothing part of the formula.
I call this formula the Sve_Rainbow_Weighted. Lets put it all together and shorten the rainbow average to
Copyright Technical Analysis Inc.
 Stocks & Commodities V. 28:10 (38-44): A Smoothed RSI Inverse Fisher Transform by Sylvain Vervoort

obtain the Sve_InvFisher_Rsi formula:

SVE inverse Fisher RSI
{SVE_InvFisher_RSI}
transform in action
RSIper := Input(RSI Period?,2,30,4); On the chart of Csx Corp. in Figure 4, it is clear that a swing
EMAper := Input(Exponential Moving Average Pe- trader looking at price and the four-period Rsi below the price
riod?,1,100,4); is not going to have an easy time finding entry and exit points
ma1:= Mov(C,2,W); in this up move. But after treating the Rsi with my inverse
ma2:= Mov(ma1,2,W); Fisher transform, you get very clear buy and sell signals. That
ma3:= Mov(ma2,2,W); also makes it easier for medium- to longer-term traders to find
ma4:= Mov(ma3,2,W); additional entry points in the medium-term up move.
ma5:= Mov(ma4,2,W);
Note that the inverse Fisher transform Rsi is not meant to
ma6:= Mov(ma5,2,W);
ma7:= Mov(ma6,2,W);
be used as an automatic trading system. The idea is to offer
ma8:= Mov(ma7,2,W); you clear short-term turning points as much as possible. We
ma9:= Mov(ma8,2,W); are not living in an ideal world, so the Sve_InvFish_Rsi will
ma10:= Mov(ma9,2,W); not always move perfectly between the zero and 100 levels.
RainbW:= (5*ma1+4*ma2+3*ma3+2*ma4+ma5+ma6+ma
7+ma8+ma9+ma10)/20; Combining it with other indicators
x:= .1*(RSI(RainbW,RSIper)-50); I suggest using two additional indicators with the inverse
EMA1:= Mov(x,EMAper,Exponential); Fisher Rsi. One is my fast short-term Arsi (asymmetrical
EMA2:= Mov(EMA1,EMAper,Exponential);
Rsi), described in the S&C October 2008, to find short-term
Difference:= EMA1 - EMA2;
ZlEma:= EMA1 + Difference;
reversal points based on divergences. The other indicator is a
invfish:=((Exp(2*ZlEma)-1)/(Exp(2*ZlEma)+1)+1)*50; medium-term stochastic oscillator to track the medium-term
invfish price moves.
Here is the MetaStock formula for the Arsi:
I am using a default value of four periods for the Rsi and the
zero-lagging moving average. These values give clear turning {ARSI}
points, which are often clearer and faster than the five-period Period:=Input(ARSI Time Period ->,1,50,8);
Rsi original formula. UpCount:=Sum(If(ROC(C,1,$)>=0,1,0),Period);
DnCount:=Period-UpCount;
UpMove:=ExtFml(Forum.MOV,If(ROC(C,1,$)>=0,
54 HOT
53 ROC(C,1,$),0),UpCount*2-1,E);
52 DnMove:=ExtFml(Forum.MOV,If(ROC(C,1,$)<0,A
51 bs(ROC(C,1,$)),0),DnCount*2-1,E);
50
49 RS:=UpMove/(DnMove+0.0001);
48 100-(100/(1+RS));
47
46
45 Remember to install the external forum.dll file that has
44 a function for a dynamic moving average. You can find
43 this file on the Equis forum (http://forum.equis.com/).
100 To explain the combined use of these indicators I will
use Figure 5, which displays a chart of the Starwood
Hotels and Resorts World (Hot) stock from May to
50 July 2005.
The blue curve below the price plot is an eight-period
Arsi. The red curve in the same window is a slow standard
100 50-period stochastic oscillator with three-period slow-
ing. The blue curve at the bottom is my inverse Fisher
50 Rsi transform with a four-period Rsi and a four-period
zero-lagging Ema.
0 At the first vertical dash-dotted line, the Sve_InvFish_
METASTOCK

9 16 23 31 6 13 20 27 5 11 18 25 1 8 Rsi crosses above 12. This is basically a buy signal. The

May June July August
stochastic moving up from a bottom level confirms that
FIGURE 5: COMBINING SVE INVERSE FISHER RSI TRANSFORM WITH OTHER INDI- this is probably the start of a medium-term up move.
CATORS. Here you see the eight-period asymmetrical RSI and a slow standard 50-period
stochastic oscillator with a three-period slowing. The blue curve at the bottom is the SVE As a swing trader, you would probably take a profit
inverse RSI Fisher transform with a four-period RSI and a four-period zero-lagging exponential when the Sve_InvFish_Rsi turns down, crossing the 88
moving average. Combining these indicators gives very clear buy and sell signals. level. Because of the limited correction, Sve_InvFish_Rsi
Copyright Technical Analysis Inc.
Stocks & Commodities V. 28:10 (38-44): A Smoothed RSI Inverse Fisher Transform by Sylvain Vervoort
INDICATORS

does not reach the zero level and turns back up only a couple the medium-term stochastic oscillator turning in oversold or
of days later (second vertical red dash-dotted line). Should overbought area, can make your decision to buy or sell a lot
you enter here for another long position? easier. I would encourage you to try this method on several
Yes, you should. First, the turning-up of Sve_InvFish_Rsi is charts. I am confident you will see an improvement in your
confirmed by a hidden divergence between price with higher trading results.
bottoms and Arsi with lower bottoms, pointing in the direc-
tion of a continuation of the previous uptrend. There is also a Sylvain Vervoort is a retired electronics engineer who has
small W-pattern indicating a short-term reversal. Second, the been studying and using technical analysis for more than 30
slow stochastic is not yet in its overbought area and is turning years. His book, Capturing Profit With Technical Analysis,
up to continue the uptrend. published by Marketplace Books, focuses on applied technical
About two weeks later, there is another down move of analysis and introduces the Lockit trading method featuring
Sve_InvFish_Rsi. You probably would not close an open his own techniques and proprietary indicators. Vervoort may
long position here yet because of the nearby support from be reached at [email protected] or via his website at
the previous up turning point (marked by the second arrow). http://stocata.org/.
Again, just a couple of days later, Sve_InvFish_Rsi turns up
once more (third vertical red dash-dotted line). So should you Suggested reading
enter here for another long position or hold on to an open Ehlers, John [2002]. Using The Fisher Transform, Techni-
position? cal Analysis of Stocks & Commodities, Volume 20:
Absolutely! First of all, the turning-up of Sve_InvFish_Rsi November.
is once again confirmed by a hidden divergence between price _____ [2004]. The Inverse Fisher Transform, Technical
with higher bottoms and Arsi with lower bottoms, pointing Analysis of Stocks & Commodities, Volume 22: May.
in the direction of a continuation of the previous uptrend. Sharp, Joe [2000]. Limiting The Lag Of An Average,
Second, the slow stochastic is not yet overbought and still Technical Analysis of Stocks & Commodities, Volume
moving up. 18: January.
Finally, the Sve_InvFish_Rsi turns down again, breaking Vervoort, Sylvain [2008]. Arsi, The Asymmetrical Rsi,
the 88 level (the blue vertical dash-dotted line). Should you Technical Analysis of Stocks & Commodities, Volume
interpret this as a medium-term selling signal? Yes! You can 26: October.
now see a negative divergence with higher tops in price and Widner, Mel [1997]. Rainbow Charts, Technical Analysis
lower tops in Arsi. The slow stochastic has been moving in of Stocks & Commodities, Volume 15: July.
its overbought area for some time and is turning down. Now
is a good time to take a profit.

Clear buy and sell decisions

The clear entry and exit signals given by Sve_InvFish_Rsi, S&C
in combination with the short-term Arsi divergences and

Copyright Technical Analysis Inc.