Contents

ChapterOne: Old Mythsand New Concepts

PartI:GettingStarted.................................................., ........ ..•. ... 19
.
Part IT:l11e RSI Half-pipe................................................................... JJ
Part ffi:ASmall Doseof RSI Mathematics....,............................21
PartJV:TheOldMyths.....................................................................Z3
PartV:NewIdeas............................................................- 29
Part VI:TheCardwellianSignals.....................................................34

Chapter"J\vo: T heSlanted RSI Universe

Part I: TheRSI Distortion...............................................................ifl
Partll: RSJZoom-in........................................................................51
PartIO: RSJ Segments andChannels...............................................53
Panrv:RSIChaimel Logic..............................................................5'/
PartV:The Nature of Price/RSIDivergences...................................o1
Part VI:RSI Signals Revisited.......................................................72
Part VIl:The RST Lighthouse........................................................81
Part VU!: (DD)-Based Support and Resistance Lines...................83
Part IX:Validity of(DD)Signals. .. ....... ... . .. . .. ...... .. ........ .,. 84
Part X:FAST(DD) Signals....................................................
Part XI:Patternsin theRSI.. ... .. ...... ... .......... ... .................. ..
PartXII:UseofRSI(3)........................................................................, CE
Part Xlll: Practice.............................................................................l04

Chapter Three: RSITimeFrame Correlation

Part l:lnrroduction.........................................................................111
Partn: Basisof TimeFrame Correlation........................................J13
Part ill: r ailureof (DD)Signals.......................................................115
PartJV:TFC- Zooming-in, Zooming-out.......................................121
Part V:Identical Signals in DifferentTime Frames........................125
Part VI: Conflicting SignalsintheSameTime Frame....................129
PartVU: Conflicting Signals in Difforent Time Frames...............134
PartVITI: Projected PriceTargets Revisited..................................161

Ch apter Four:Additional Thoughts andTools

Part 1: RSI Combined wit.h Other Technical AnalysisTools........187
Pa.rt. rJ: Practical Guidelines- Plotting RSI Channels.....................1

Chapter Five: Real-TimeApplication

PartI: D1ug Index($DRGX)..........................................................215
Partil:CrudeOiJFutures(CL)..........................................................lfJ7
PartIll:S&P500Index.....................................................................302
PartIV NVIDlA(NVDA)...............................................................325

Appendix I: Referencesand Recommended Reading.. .. .. ,........................342

Appendix 2: RSI MovingAverages Formula forl'radeStation.....................343

Index............................................................................................................344

AJI charts created onTradeStationl\l), flagship product of"TradeStation Teclmologies,Inc.

 PREFACE

How could I.ever forget October. l 9 1987? Jt was the day l fell in love with
technical analysis! I had started to study the financial markets a few weeks earlier,
looking for an oppo rtuniry to make some extra money. Back then, the internet
did not exist, so most of the information l could gather came out of The Wall
Street .Journal. In the week leading up to the 1987 stock market crash, I decided
to buy some Put Options.

My decision was based on what l had learned about the interpretation of the Relative
Strength Index, or RSI, which looked prettystraightforward lo me. Then, Black
Monday was upon us. On that day, making money never looked easier. l could not
believe my eyes as I watched the live pictures from Wall Street. I was making money
by the mjnute. having a hard time keeping score as my Put Options' value literaHy ex
ploded. Jus t watching the markets go my way and being on the right side gave me a
powerful feeling. I was addicted!

l had never expected it to be so easy and, more importantly, it seemed that l was
holding the key to further gains. My key was the magnificent RSI that had served me
so well.

Looking back, these are sweet memories! I must conclude tbat this small victory
over themarkets involved more luck than skill. Let's just call it "beginners' luck.''
All things considered. when I say that l based my trading decision on RSl
analysis,1 should add that I was merely following the RST guidelines published
in some popular books and articles that were available at that time. Did I really
believe that the path to easy success was paved with piece-of-cake RSI analysis?
Unfortunately, I did.

You have probably already guessed what happened next. In the months and years that
followed my injtial "brillian1 move" on Wall Street, I lost most of the easy money.
The losses left me puzzled and disappointed, because I had been stubbornly acting on
the RSI signals that I believed had correctly triggered my bearish profits back. in
October 1987. Still, while feeling duped by theobviously misleading pub!ications I
had studied, I remained convinced that the RSI had much to offer, if l only knew how
to read it con-ectly. And so, I started my own search for alternative views on the RSI
and new methods of interpretation. Still I remained convinced that the RSIhad a lot to
offer, if1 only knew how to read it correctly. And so I started my own search for
alternative views on the RSI and for new methods of interpretation.

TheRelative Strength Index was introduced by J. Welles Wilder, Jr. in 1978 in his
book, New Concepts in Technical Trading Systems. Wilder had been looking for a way
to detect oversold and overbought conditions in the mark.et. Great idea! It would be
fantastic to have a tool that tells you when the market reaches absolute oversold or
overbought conditions. Such a tool could accurately pinpoint trend reversals; the only
requirement would be to take alternating positions in the market as it went through
consecutive overbought and oversold conditions and rake in the profits.
Uofo1tunately, Wilder did not create an Absolute Strength Indicator. He developed
the Relative Strength Index (RSI), which does not signal absolute overbought or
oversold conditions. The RSI merely quantifies actual market conditions relalive to
earlier conditions within a given time period. Over the years, the RST has become
immensely popular with professional and private investors and traders. But could I
find a way to use it profitably? Fortunately, as the immense data resources became
available on the internet, it was not long beforel discovered some interesting
alternatives for the Wilder interpretation of the RSI.

The method taught by Andrew Cardwell caught my attention. It complements

Wilder·s method and sheds a new light on how to put the RSI to work in a profitable
way. Based on Cardwell's inputs, J found a way to advance and develop a new logic
for the interpretation of the RSI by redefining the signals, as well as the limits of their
validity. Most importantly, I was able to introduce the concept of Time F ramc
Correlation, adding an extra dimension lo the RSI toolbox.

Today, technical analysis has become an industry i11 its own right When surfing the
internet, thousands of websites areavailable offeringsoftware, books, recordings, courses,
consulting services and conferences on the subject of technical analysis and trading
strategies. To my amazement 1 find that when discussing the use of the RSI, none of
the publications refer to Cardwell's mies of interpretation, which is a pity.

In this book I have condensed most of the relevant information that took me so long
to gather from the early Wilder rules to the useful Cardwell rules and beyond, to my
own RSI Channel interpretation with Time Frame Correlation insights.

ls technical analysis worth the effort at all? lf this question means "can market
moves be predicted; ' the answer is: yes and no! When l say that startjng
tomorrow, the markets will move up, down and sideways, I make a valid
prediction tbal will surely t:ome true. But that does not say much, does it?

Some people wilJ say that the markets are moving " randomly.'' It is not always clear
what this means either. ln bis book Fooled by Randomness (Thompson, NY), Nassim
N. Taleb tries to convince us that market analysis is of no use because price action is
random. However, it has been documented that patterns appear in statistical data
describing selected, obviously random events. For math lovers , I refer to the "Arcsine
Law" theory in respect of the random-walk properties of the financial markets.

Contrary to Taleb's, my conclusion is: If patterns emerge, the study of the events
under examination, whether you label them "random" or not, should include the study
of those pattems. The kind of randomness that never generates any patterns at all is
very hard tocome by, as any cryptology expert will confinn.

The "random vs. predictable" issue reminds me of the " detern,inism vs.
probability'· discussjon in Physics and Philosophy. One may lookat the issue in
this way: lfyou start with enough dice players, you will end up with a dicer who
rolls double-sixes ten times in a row! That is deterministic; it is bound to happen
as detennined by the laws of statistics. That particular person who is the winner of
the contest may rightly consider
herself just extraordinarily lucky because any of her co-players could have hit that
mark, which defines randomness.

So it looks to me like randomness and predictability are complementary, rather Lhan

mutually exclusive depending on your definition of those concepts. If randomness,
chaos, theefficient market, game theo,ry prisoner dilemmas. or however you choose to
qualify market actio n generates patterns. we may profit from studying them.

Some will argue that studying an indicator such as the RSI is useless because it
merely repeats price information and expresses it in a different form.

IL is ce11ainly true that the "u ltimate reality'' in technical analysis is price. Sure, we
start with price data and study it in order to draw conclusions about probable future
price behavior scenarios. But if price is the ultimate reality to be studied, why not look
at it from various angles? When scientists attempt to define the properties of an
object. they put it through all kinds of tests, observing its behavior under various
unusual conditions. Each oneof those tests reveals a certain aspect of theobject's
"reality'' and this knowledge enables the scientists to anticipate the object's behavior
under varying conditions in the future. ln my opinion, putting price data through a
mathematical equation, such as the RSI formula, allows us to examine a relevant (but
usually hidden) aspect of price reality that would otherwise remain undetected.

So yes. I do indeed believe that technical analysis beats t1ipping coins when it comes
to profiting in the financial markets. 1n the following pages, after having reviewed
the old rules and insights. I aim to demonstrate that tbe best way to analyze the RSI is
to examine its behavior in relation to its channels in various time frames and to
correlate these pictures.

I hope that my book contributes to a better understanding of the RSI and that it will
offer an extra tool to analysts, traders and investors who are willing to put in the effort
to assemble the RSI puzzle.
 CHAPTERONE

OLD MYTHS AND NEW CONCEPTS

,
 PARTI

G ETTING STARTED

We need a tool that will provide a workable hypothesis about market behavior in the
near future. This hypothesis would include market direction, timing, price targct(s) and
the probability of thescenar io becoming a reality.The tool shouldalso provide an early
warning when it turns out to be wrong, so losses can be limited. The Relative St rength
Index is such a tool. lt allows us to enter the market with confidence, detennine the
risk/reward ratio, exit positions early if the hypothesis is wrong (taking a small loss)
or ride the market to a target price and beyond when the hypothesis is right

J. WellesWilder, Jr. developed the Relative Strength Indexas a tool to

detectoverbought and oversold conditions in the market. To do this, he compared
the average up and average downprice movements within a given period of time in
the past.The look-back time window favored by Wilder for the calculatio n of the
RSI values was 14 periods. Consequently, themethoddescribedin this book pertains to
14-period RSI data calculation. using closing prices. This means that noconclusio n
should bedrawn from an RSJ s ig nal until the closing price is in. Althoug h this
seems to be common sense, in the heat of trading, one is often temptedto act
prematurely.

The RSI charts in this book feature the same 14-periodsetting, except where
mentioned (some charts only cover 3 periods). Each market bas its own
peculiarities but,generally speaking, my RSI analysis method is applicable to all
kinds of markets, including stocks. futures, currencies, treasury bonds, interest rates
and indices in all time frames.

A word of caution: ln very short time frames, the RSI is volatile, frequently hitting
extreme high s and lows and generating contradicting signa ls in rapid succession.
Also, if the overnight sessions in futures markets are not plotted, the RSI readings at
the opening oft.he regular sessio n may look irrational in the short time frames
because of the overnight gap in RSI data.

In flat markets, the RSI will generate signals while pricesgo nowhere. It is imp
ortant to avoid getting caught during lackluster lunchtime trading periods or
overnight sessions. For RSl signals to work in terms of anticipated price action,
there should be some degree of price action momentwn and volatility in the
market.
 RSI: Logic. Signals & Time Frame Correlutiun

17".Jl t03o ,s-21 2Ct2• 2'21 Zl.11 IOl'IB I.CD IXI ltll 424 S4S 711 .av 914 10"l1 HI! Tl"lt l!l tS 1.fU lS.OJ 18t'Jli 11CD u,oo 1et,
C >
C WII\T,.tde.Sldlllor,

Figure 1.1 S&P Futures Overnight (3-Mi nute)

After 14 periods, the RSI starts fluctua ting between 30 and 70, while prices go nowhere.

PARTU

THE RSI HALF-PIPE

Watching RSI fluctuations on a chart is like observing a ska ter going up and down in
a half- pipe. Imagine look ing at theskater from a bird·s viewpoint, high above. From
this perspective, the half-pipe looks like a flat, rectangular metal surface and the skater
goesback and forth between the top and bottom edges. The half-pipe is the RSI chart,
and its extreme valuesof 0 ru1d 100are theedges where theskater seems to besuspe
nded before accelerating down the slope in the oppo site direction . The skater's speed
is highest in the middle of the half-pipe. As he zooms upward on either slope, his
speed slows dramatically, then drops to zero as his course reverses.

Due to the mathematical constraint of being squeezed into a range of 0 to I 00. th e

d istance traveled by theRSI in its chart is not directly proportional to thechange in
price that caused the move. as it is a logarit hmic funct ion. In fact, just like the skater
in the half-pipe , the RSI has its grea test ve locity when crossing the middle of tbe
RSI chart at the RSl SO-line. It will rapidly and increasingly slowdown when
approaching theupper and low er edges of the chart.

20
In practice, a relatively small change in price can cause a big move in the RSI value
when it travels around the SO-level. However, large price changes are required to
move the RSI valuea little, once it approaches theextremes, as if an ever-
increasingmomentum were required to push it up an increasingly steep slope.

The values where the Relative Strength Index runs into rapidly increasing resistance
are the 66.6 and 33.3 levels. Ln fact, this behavior is a consequence of the nature of
the RSI equation itself, designed to limit the course oftbe RSI within its Oto 100
range.

In other words, a small move in the RSI va lue, when it is nearextreme levels, con-esponds
10 arelatively big move in price!

PART III

A SMALL DOSE OF RSI MATHEMATICS

When studying the RSI( 14) - meaning the RSI takes into accowlt a l4 period look-
back window ( 14 days in the daily charts or 14 minutes in the I -minute cha1t) - we
are looking at the Relative Strength (RS). expressed by the price ratio:

Average UP (on close) over last 14 periods

Average DOWN (on close) over last 14 periods

This ratio is pressed into a 0 Lo 100 scale by the following formula:

RSI = I 00 - (I 00 / 1-r-RS)

ff all of theclosing prices had been UP over the given period, the RSI would rapidly
rise to 100.sin<.:e the[·RS' ] would tend to be infinite causing the [(100 / 1-t-RS)]to
lend10 be zero.

As such, this makes the equation unsuitable for use because the resulting RSI values
wiII be higher for a slow (but uninterrupted) price rise than for a strong price rise
tJ1at suffers the occasional down retracement.

ln order to avoid this effect, the Welles Wilder approximation system was adopted.
Lnstead of taking each of the 14 UP or DOWN values of the look-back period into
account, Wilder resolved to take the average UP or DOWN of the last 13 periods.
This value was multiplied by 13 and added to the value UP or DOWN of the most
recent period number 14. Then, tbis swn was divided by 14. This calculation also
made the daily manual RSI calculations much easier to manage as today's value was
added to the latest average and the new value was divided by 14.
21
 RSI: l ugic. Signals & Ttme Frame
Correlation

For an RSl(l4) calculation using dailydata, the RS equation shown above is changed
and the values for the Average UP and Average DOWN are calculated as follows;

UP = [(Average UP last 13 days) x 13] + Value UP today

14

DOWN = [(Average DOWN last 13 days) x u·1 + Value DOWN today

14

The fact that the value of the first day of the RSI( I 4) analysis is based on the
average oftJ1e 13-day period preceding that day means that today 's RSI calculation,
to a certain extent, incorporates daily data older than 14days. This method of
calculation smoothes out the RSI and eliminates the unwanted tendency of the RSl to
move to extremes after 14 periods of uninterrupted price increases or declines.

Co nseq uen tly. a time span covering a mini mum of 15 times the look-back period
needs to be observed in order lo obtain reliable RSI values. This means that for an
RSI(14) dui v study to be reliable, at least 210 days of data need to be available.

The RSI Ratio Table

For a given look-back period. when the UPAverage equals the DOWN Average in the
RS ca lcula tion, the RSI will display a value of 50 as the ups and downs balance.

When the balance shifts in favor of the UPor DOWN Average. the RS! value will rise
or fall, but not in a directly proportional or Jjnear manner.

The following tableshows the UP/ DO WN ratio required to move the RSI to the
displayed values:

RSI = 90 UP / DO WN = J O / J
RSI = 80 UP/ DO W N = 4 I I
RS I= 75
UP/ DO WN = 3 I I
RSI = 66.6
UP / DOWN = 2 I 1
RSI = 50
UP/DOWN = 1 / 1 NEUTRAL BALANCE
RSI = 33.3
UP/DOWN = I I 2
RSJ = 25 UP/ DOWN = l I 3
RS1 = 20
UP/ DOWN = 1 / 4
RSl = IO UP/DOWN= I / JO

This table clearly illustrates that the RSI value meets increasing resistance when
approaching the extreme values of O and I 00. A rise of IO points in the RSI from 75
to 85 must correspond to a price move UP that is much more significant than a price
move UP corresponding to a IO rmint RSI move from 50 to 60. The half-pipe effect
in theRSI becomes clearly visible above the 66.6 level and below the 33.3 le vel,
where the RS! must travel ever-steeper slopes.
 PARTIV

OLD MYTHS

No w, for the sake of completeness we are ready to look at the initial RSI rules. I am
sure that you are already familiar with some of them, as they keep appearing in many
of tbe present-day technical analysis manuals and websites.

Here is what I learned back then. The main RSI signals, we were told, were:

The RSl hitsor exceeds values of70 or30,signaling overbought ::md ove rsold
conditions respectively.

Divergence between price and the RSl signals imminent trend reversals .

The RSI crosses the 50-line up or down, meaning the market turns bullish
or bearish.

The swing failures, meaning the RSI reverse s course.

Lefs have a closer look at each one of these points.

Ove rboug ht and Oversold Levels

As demonstrated earlier, the mathematical resistance of the RSI is a built-in feature
that takes effect around the 66 and 33 levels. This means that more often than not. the
Relative Strength Index value will have a bard time moving past these levels. But
does this justify the conclusion that these levels signal the market trend is exhausted?

We know chat a tiny movement in the RSI value, when it is over 66 or under 33,
corresponds toa relatively large move in price. This means that at the RSI 70-level,
the best may be yet to come in a stro ng uptrend. This is not in tenns of further
important jumps in the RSI value, but in tenns of prices moving up strongly.
Conversely, in a downtrenc..l there is a hugedown potential for prices while the RSI
value is below 33!

Also, in my view, any claim relative to RSI signals should spec ify from which time
frame the signal is taken. Imagine that we have been in a strong uptrend for the last
five trading days. Would it make sense to determine that this rally is coming to an end
just because the RSI hits the 70-level in a 15-minute chart? In reality, the 15-minute
RSI will hit the 70-level several times a week in a rally of some magnitude .

When the 15-minute RSI reaches the 70-level, the RSI value on thedaily chart may be
at 60, with plenty of UP poten tial. We may be tempted to assume the daily time
frame was the main object of Wilder's research, but it is clear that we will need to find
a way to correlate the various RSI pictures. Tt is not hard to imagine the following
scenario: The RSI hits 30 in thedaily timeframe, allegedly signaling oversold
conditions, while the weekly RSl may be at the 40-level, suggesting there is still a
distance for the market to go on the downside.
 RSI: Logic, Sig11als & Time Frame
Correlation
From this illustration, it becomesclear that RSI leve ls sig nal different things when
taken from different time frames. Hence my convictio n that any conclusion drawn
from a given RSI pictme will be incomple te unless it is corre lated to otber time
frames.

What should be made of Price/RSI d ive rgence'! Imagine that the daily RSI hits 75.
According to Wilder, this would indicate that the rally is about to end, while prices
continue to h.it newhighs with the RSI meandering to tJ1e downside. (This is the
definition of Price/RSI negative divergence). In thjs example, so me will say the RSI
worked off its overbought condition. However,si nce prices kept rising, there really
never was an overbought condition.

There is no such thing as an overbought or oversold RSI and there are nosuch
things as absolute overbought/oversoldprice conditions. In fact, there are plenty
of examples of up-trending markets with major price jumps taking the RSI well
above 70, even in the weekly time frame. In bear markets, you will often see
prices accelerate their slide a the RSI value appears to be locked down below 30
for an extended period of time.

My advice: Do not let the 70 and 30 RSI levels fool you into initiat ing countertref\u
positions; it needs to be put in the correct perspective.

ilJO
O
10
9!6 00

Figure 1.2 IBM Weekly

This weekly chart shows a strong rally in IBM from 1996 to 2000. I have drawn a
horizontal line at the 70-leve l. If we had blindly applied the Wilder mies for RSI
in terpr etatio n, we would have so ld IBM in Quarter 1 of 1996, when the RS] reached
the 70-level. But look what happened after that. After a mild drop into mid-1996 , the
RSI: Logic, Sig11als & Time Frame
Correlation
market rallied and the RSI reached the 80-level by theend of 1996. Prices continued
to rise, while the RSI peaked well above 70 several limes in 1997 and 1999. The fact
that the RSI never dropped below 40 until Quarter 4 of 1999 is typical of such a
strong rally. This chart is not exceptional. There are plenty of examples that show the
RSI exceeding 70, after which the rally continues and strengthens. I believe it is closer
to the tmth to say that the rise of the RSI above 60 is typical of the start ofa rally than
it is to say that the RSI reaching 70 signals the end of the rally.

RSI/ Price Divergence

Price/RSI divergence provides another source of confusion. Many mainstream
publications about the RSI telJ you that this kind of divergence signals the end of the
cw·rent trend. Some even list divergence as buy and sell signals !

Unfortunately, as simple and as good as that may sound, divergence can bea
misleading interpretation anti (as far as I know) no one has offered any satisfactory
alternative views.

Missing here is the specification of the time frame reference when trying to anticipate
what Price/RSI divergences will cause in the longer run. All of this talk aboul
divergence makes them look special, but they are not. They occur frequently because
the y are inevitable; thus we should not search for any magical properties here.

After the initial trend pushes the RS I to its limit, given the significant relative
increase in momentum int.he recent past, the RSI flattens out or even declines,
simply because the rate of increase in momentum is decreasing. while the trend
keeps going. More than anything, the existence of Price/RS! divergence confirms
the present trend. Look at the evidence.

A negative divergence occurs when price reaches a new high, while the RSI only
makes it to a lower high. By definition, when prices hit new highs. we are looking at
an uptrend. A positive divergence occtirs when prices hit.lower lows, while the RSI
values already rise, hi1ting higher lows. By definition, when prices reach lower lows,
we arc looking at a downtrend. So, an interesting point about Price/ RS I divergences
is that negativedivergence occursonly in an uptren<l and positive divergence occurs
only in a downtrend. Obviously. this statement is an oversimplification, but should be
included because this is how Andrew Cardwell treats divergence in his Trend
Detennination Checklist. He sees divergence as one of several symptoms
accompanying a trend. It says tJlat the market is in an uptrend as long as there are
higher highs in price. To that I would add: That is correct, whether or not there is a
negative divergence. In other words, do not worry about negativedivergence until you
see lower highs in price.

Divergence occurs Jreq11e 11t l y and ii 1/oes uot have a11y pedal signijica11ce.

I like to think of Price/RSI divergences as follows: ln a strong downtrend driven by

longer-tenn traders that continues for several weeks, the RSI value in the hourly charts
will be propelled down below the 30-level in a matter of days. But the price decline is
far from over. What happens next? The RSI value is pushed even lower , let's say 1o the
15
 RSI: Logic, Signal s & Time Frame
Correlution

15- level. But the price drop is still not over. The daiJy RSI value drops below the 38-
leve l and appears to have further to go. Then what happens? The hourly RSI, at some
point, hitsits minimum value, meaning the maximum momentum of the decline is
unable to push the RSI down any furtJ1er. Remember the extraordina ry down/up
ratio that is requir ed (in the last 14 hours) to push the RSI below the 20-level and the
I 0- leve l.

The RSI is being compressed at the bottom of the hourly chart, but the price drop
continues (let's say that the daily RST reaches 32). The hourly RSI has nowhere to go
but up, as down momentum in relative terms is not increasing sufficie ntly lo depres s
it further. Consequently. the RSI starts to curl up in the hourly chart, while prices co
ntinue lheirdrop.

We will treat RSI divergence in accordance with the general rules that will be defined
later, meaning they need to be understood as a particular RSI behavior withjn RSI
channels.

11000

)10(10

:moo

"""'
""'00

"""°
260UU

lOOOO

24. 0W

.. ,.... 1)),1..

"'"'
00

Figure 1.3 Daily Oil Ind ex

In this Daily OIX (Oil Index) Chart, there are two positive divergences. The first one
sta rts at Point 1 in the RSl chart. While prices reach a new low in October 2002, the
RSI has been rising since August. 1t is clear that this positive divergence does not lead
to a rally, as prices resume their decline in January 2003 , w ithout exceeding the last
lop of October 2002. What follows is a lower lo w in price in January 2003.

The second positive divergence starts at Point 2. This time it persists lo ng enough to
lead the RSI witbjn its ascending channel into bull ten-itory and the appearance of a
buy
signal by the end of May, signaling the start of a rally. The bullish signal in the RSl is
marked by the acronym (DD+) within the ascending RSI channel.

1000

•
ci.. .d" hl u,dtSI-

Figure l.4 Daily Oil Index

In this chart, the point of interest is the negative divergence between price and RSI in
the period between Points A and B. Here we see prices reach consecutive higher
highs, while the RSI from Points A' to B· declines. Note the rapid pricerise into
Point A, corresponding to the initial price thrust and steep increase in momentum,
causing the extreme RSI reading at Point A' . This reflects the high rate of change in
momentum in the recent 14-day period. Past Point A, any momentum change will be
compared to the high earlier values in the recent past, hence the decline in the RSI
value despite the continuation of the rally.

Still, the negative divergence does not lead to a trend reversal to the downside.
Quite theopposite happens, as prices continue their rise past Point B. Again, it is
not hard to find many more examples where a negativedivergence does not lead to
the beginning of a bear market phase.

Price/RS{ divergence is not an exceptional signal. It occurs frequently, as it is

inevitable, and does not watTant any special treatment. lt is just one way for the RSl to
escape its mathematical constraints.
 RSI: Logic, Signals & Ttme Frame
Correlation

The RSI SO-Line

The SO-l eve l in the RSI value signals a balance between price moves up and down
within the L4-period look-back window.This leadssome people to think when the RSJ
crosses the 50-line to the upside, the market must be turning bullish, and when the
RS[ drops below the 50-level, the market is turning bearish.

From our earlier discussion. you already know that this rule is too simplistic. Again.
there is no specification about the timeframe from which this RSI reading is taken. It
is possible to see the hourly RSI drop below the50-level while the daily RSI value is
at 60.

The RSI around theSO-level sometimes shows evidence of the indecision in the
market. Such is the case when the RSI is oscillating in smaller moves above and
below the 50- level, building an RSI triangle that is often centered on the 50-level,
The corresponding price action features either a narrow trading range or tbe formation
of a price triangle. Tius RSI behavior illustrates the indecisive fight between equally
strong bulls and bears in that particular time frame. At some point, the RSI breaks out
of the triangle one way or theother. The RSI triangle is an interesting pattern to watch
for, as it indicates a swift move in both the RSI and price in the near future.

Failure Swing in the RSI

Failure swings occur when the RSI drops below the level of the trough in an M-
fom1at10n or whe n the RSI rises above the level of the peak in a W-fom1ation.
These signals are more significant when they occur within the oversold(< 30)
oroverbought(> 70) zones of the RSI chart. The general consensus is: these
conditions are met, a price correction is imminent. There are problems with this
theory.

First, when comparing different points on the RS I graph, the relevant reference lines
are not horizontal lines (such as the 70-Jevel or 30-level lines) but slanted lines,
defined by the upper and lower boundaries or their parnllels of RSI channels.

Second, the failure swing, especially near the extremes of the RSI chart, often leads to
the start of a Price/RSl divergence. This means that after a failure swing down has
occurred, the RSI often continues downwards and rebounds off of a lower low, while
prices continue going up. After a failure swing up, especially near RSI extremes, the
RSI may continue upwards, while prices just keep falling, defining aposilive
Price/RSI divergence.

In later chapters we will learn that these failure swings may be relevant if and when
they coincide with particular RSI reversals, namely those that cause the formation of
bullish or bearishsignals, labeled (DD+) and (DD-}.
 PARTV

NEW IDEAS

New RSI Realities

Let's leave the RSI world as described by Welles Wilder and examine some new
ideas hrought forth by Andrew Cardwell. Cardwell, who acknowledged the
inherent weaknesses of theold rules, developed some interesting insights. By
advancing the art of RSI analysis, his work enables us to use the Relative Strength
Index to:

I. Establish trend in a given time frame

2. Anticipate trend reversals
3. Detect buy and sellsignals
4. Calculate price targets
5. Calculate risk-reward
6. Determine trade entry and ex.it points and stop-loss levels
7. Establish signjficant trend lines and suppor t/resistance lines

In this chap ter, we will examine some of the items in thislist, while somewill be
reviewed when the signals and their validity are redefined.

These insights take the interpretation of the RSl to a new le vel. What made the
difference? In my opinion, the major breakthrough was the discovery of spec ific
buy and sell signals in the RSI. The method for Price Target Calcu lation,
associated with these signals, is taken from the Fibonacci method. Also. Wilder's
70/30 rule was redefined and a more realistic and useful way of looking at
specific RSI levels was introduced.

Let's take a closer look.

Significant RSI Ranges

It does not pay to look for a specific level where a market can definitely and
absolutely be called overbought or oversold. As we saw in our earlier disc ussion, it is
possible for the RSl in an uptrend to rise to a value of75 and retrace to 65, while
prices keep going up at the same time. This is how negative divergence fonns. From
65, the RSI may reverse up, rising back to 75, while prices rally to new highs.

In practice, it appears to be more relevant to look at the range where the RST is
moving. Obviously, the RSI travels mainly in the area between 30 and 70 becauseof
its mathematical constraints. However, in bull markets, positive market sentiment
shifts the RSI range upward by about IO po ints, where the range rests between 40 and
80. Thisallows an uptrend in a particular time fi-ame to be defined by simply looking
at the value range where the RSI is moving.
l9
 RSI: Logic. Signals & Time Frame
Correlution
Conversely, in downtrends, negative sentiment cause s the RSI range toshift
downward about IO points. This leaves an RST value traveling between 20 and
60, which is symptomaticof a downtrend in a specific time frame.

While the figures for the uptrend and downtrend RSI ranges do not constitute
make-or break limits, these important guidelines can be fonnulated. For a given
time frame, the RSI moving within the range from (roughly) 40 to 80 is
symptomatic ofan uptrend. The RSI moving within the range from (roughly) 20 to
60 is symptomatic of a market downtrend.

00

00

4100

4000

:lSOO

lOOO

,.00

lOOO

1000
0
IIOO
O
all.I()
Him

00
100
0
lOO
.,. O
T>]
000
u,4<d""",,_._

Figure 1.5 Weekly GM

1n this weekly chart of General Motors, prices drop from about $55 to $20 during
2005, before rebounding in 2006. While the RSl value reaches well over 80 in January
of 2004, it never exceeds the 60-level after the pricedrop begins inApril 2004. Also
note that the 30-level in the RSI is broken. The RSI va lue reaches the 20-level in
April and December 2005. Obviously, this is not a case where it would have been a
good idea to have purchased General Motors when the RSI reached the JO-level.
 RSI: Logic. Signals & Time Frame
Correlution

.WSE L.'4300 •lll!I ,1 B:'.ft.tlo65'20 .A.-860$ (),,8145 """'86:21 L 0'164 (),.8608 V••7781!BYJ

:Dm
2000
!000

.,.
,

Figure 1.6 Weekly IBM

This weekly chart shows the strong rally in IBM from 1996 to 2000, It is characteriz
ed by the RSIstaying well within the 40 to 80 range, thus identifying a bullish market.
The violation oftbe 40-level in the RSI(Quarter 4 of 1999) marks the end of the strong
bull market and the start of a consolidation phase, with the RSI unable to exceed the
70- level in 2000 and 200I.

A Special Kind of Divergence

We know that the shift in the value of the RSI is notdirectly propo11ional to the
change in the underlying price. As the RSI measures the relative rate-of-change of
prices, (when tbe rate of the price increase begins to lose steam in a rally lasting over
14 periods), the RSI value may actually start to decline, while prices are still going up
at a more moderate pace. lt is possible, in an uptrend, to see RSI values decline for a
while in that particular time frame. In a downtrend , rising RSI values often appear.

Jf we combine this characteristic with the fact that the RSI tends to speed up when
approaching the middle of the RSI range in tbe half-pipe effect, it becomes clear that
someseemingly very unusual RSTchart configurationsarepossible.
Theseconfigurations could include rising prices combined with rapidly declining RSI
values, or falling prices combined with rapidly rising RSl values. These are the
properties that make this indicator interesting, and it is important to know bow to
interpretthem.

31
Walter.l. Baeyens

We have a lready noticed that when the RSI va lue hjts the 80-level or higher. this
does notmean that the uptrend is over; conversely, a downtrend is not simply ending
because the RSI has reached the 20-level or lower.

So, how do prices manage to move higher after they have propelled the RSI lo 80 or
90?

In simple terms. we could describe this phenome non by saying that the RSI takes a
few steps back when hitting its limit, while prices barely move before charging
upward again. Each time the RSI reverses up, prices rally to new highs. lf the RSI fa
ils to attain new highs in this process, we are witnessing the formation of a negative
Price/ RS[ divergence.

After hitting its mathematical resistance in a downtrend, the RS I value may pop up
to a less extreme reading, before resuming its plunge with renewed vigor, taking
prices to new lows.

In some ins tances, the rapid changes in the RSI value are markedly out-of-
proportion with the change in price. These changes are the basis of Cardwell's buy
and sell signals.

Thjs is the underlying logic: If you notice RSI moves thatareobvious)y out-of-
proportion with the correspo nding price moves, expect pricesto resume their trend
with a vengeance when the RSI reverses. This provides the basis for an impo rtant
guideline: In an out-of proportion RSI move, you are almost always on the wrong
side of the marke t if you follow the RSI. In an uptrend, if the RSI fal ls dramatically
while prices hardly move downward, do not be fooled into taking a short position;
wben the RSI bounces back, prices will likely move to new highs .

Jn a downtrend. when witnessing an RSI j um, p while prices hardly move, do not be
fooled into a long position. Once the RSI reverses back down, prices will likely fall lo
new lower lows.

J'Z
 RSI: Logic. Signals & Time Frame Correlation

ro

S500

00

00

G'

>JO
O
8'

Figure 1.7 LRCX ( IS-m inute)

This chart illustrates the '"peculiar divergences " between price and the RSI. Note the
diEtere nces in variation in price versus the RSI. The price drop from Level A to
Level B corresponds to a modest decline in the RSI from A' to B'. Point B' in the RSI
is near the 20-level, where the RSI is being compressed. In contrast, the minor price
rebound from Level B to Level C propels the RSI to Level C'. The RSI move is
markedly out of-proportion with the rise in price.

The same thing happens from Points F' to G', where the RSI has a significant jump
after a minor price rebound (compared to D to E) from Level F to G. These out-of
proportionRSI moves, which I will relabel Disproportional Displacements (DD), form
the basis of the buy and sell signals in the RSI.

.33
Walter J. Baeyens

,,,

"'"
"

,111 ,.oe. 19 f6 ,1""' '7 -t Jll)t5 t,1'10 "'".s ,,,1 HU!. 1n4 11,t .,,.s 1815 MIJ,,'5 ,o,s '111 HS4! ,no
• J I)
1, --,;1-.

F igure 1.8 LRCX (15-minute)

This chartillustrates the DisproportionalDisplacements (DD) intheRSI inbullish territory.

PART VJ

T HE CA RDWELLIAN SIGNALS

The RSI Range-Shift

To quickly determine the trend in a particular time frame, look at the RSI chart and
check the range in which the RSl -is moving. lfthe RSI is seen meandering roughly in
the range from 40 to 80, the trend must be up. The range from 20 to 60 is typical for a
downtrend. Consequently, we can expect to see a " range-shift" whenever the trend
changes in the examined time frame. Although a range-shHlis likely to remain
undetected until after the fact, it is still a useful signal.

A range-shift down occurs whenever the RSI fails to rise above 60 in a rally after ru1
unusually large pricedrop. Imagine the RSI moving in bullish territory, while prices
are steadily rising. At a certain point, a price correction down takes the RSI lower to
the 25-level (in bear territory). If the rally in the market is to continue, the subsequent
rebound in price should see the corresponding RSf value exceed the 60-leve l, back
into bullish territory. If the RSI fails to break above the 60- to 65-level and reverses
downward from there, the price correction may have signaled the start of a trend
reversal.
 RSI: Logic. Signals & lime Frame
Correlatio11

We can now conclude that we should use two interesting RSI levelsas main
references to detennine the market's mood The RSI values of 40 and 60. In the smaller
time frames, RSI range-shifts will be observed more often than in the larger time
frames, simply because the average day-to-day moves in the market are large enough
to send the RSl whipsawing up and down near its extremes. Still, a market that is
unable to push the RSI value above the 60- to 65-level in a small time frame (e.g. 10
minutes) is signaling extreme weakness.

In a very strong market , price corrections to the downside may be so shallow that
they do not even cause the RSI to dip below 40 in the smaller time frames.

Figure 1.9 shows a range-shift using the skaters' half pipe concept. lfwe return to the
idea that an RSI chart is similar to a skaters' half-pipe, we can imagine this half-pipe
balancing on a pivot point, with the bulls on one sideandthe beru·s on the other. At
the side corresponding with the I 00-level in the RSf, the bulls are trying to tilt
thebalance their way. If they succeed, the market turns bullish and the half-pipe is
tilted in such a way that a skater (the RST) will have no problem reaching bigher-
than-usual RSI levels near the 100-edge. To do this. a swing-back to the 40-line on the
opposite side will suffice to build up the required momentum. Conversely, in a bearish
market, we can imagine the half-pipe being tilted the other way with the skater (the
RSI) going back and forth between the near-zero level and the 60-level.

An RSI range-shift could bevisualized by imagining the half-pipe tilting either

way as time passes, forcing the skater out of lhe bear part of the half-pipe, then
into the bull part and back.

.. J J ..

Figure 1.9 Daily Oil Index

35
Th is da ily chart of the OlL INDEX offers a good example of RSI range-shifts. In
May 2002, the RSI is unable to exceed the 60- to 65-level before theprice drop.
During Lbe subsequent bear phase of the market, the RSI is unable to exceed the 60-
level, while dropping as low as 20 in July 2002. In 2003, the RSI drifts upward and is
finally able to break above the 60-leve] in mid-2003. The RS I drop following this
range-shift in June and JuJy of2003 is markedly out-of-proportionwith the
corresponding dec]jne in prices, butstill remains within bull territory. This illustrates
the rnle that you should notbe short wben the drop in the RSI is out-of-proportion
with the decline in price.

Figure
1.10

8
0

BEAR
S
Fig ur e 1.10 The RSI Half-Pipe

Buy and Sell Signals

We have already briefly discussed the emergence of peculiar divergences in the RSI.
where the fluctuations in theRSI are markedly out-of-proportion with
thecorresponding moves in price . A buy signal is formed whenever thjg type of RSl
behavior results in the following situation: The RSI reverses up from a lower low.
while price rebounds from a higher low. lo other words, a shallow c01Teciton down in
price to a higher low has caused an out-of-proportion drop in the RSI to a lowe r low.
As the RSI reverses back up, we may expect prices to resume their rise, reaching new
hlghs.

A sell signal is formed when a shallow price retracement up to a lower high

causes the RSI to move up in a disproportional way to a new high. As the RSr
reverses down. we may expect prices to resume their decline with a vengeance,
reaching new lows.

This may sound complicated. However, in practice, these particular divergences are
easy to find, once you know they exist and what they look like.
 RSI: logic, Signals& 7ime Frame
Corrclatio11

Andrew Cardwell called these signals positive and negative reversals. It was his view
that the emergence of a positive reversal in a downtrend meant the trend was
changing to an uptrend. In an uptrend, the emergence of a negative reversal marked
the point where the trend changed to a downtrend.

ln my opinion, the designation positive or negative" reversal" is not appropriate, as the

subject signal often merely confirms the continuation of an existing trend. Only when
these s ignals appear for the first time, could they be seen as "reversal" signals. The se
signals were added to his Trend Determination Checklist, as he claimed that positive
reversals or buy signals will only show in an uptrend , while negative reversals or sell
signals only show up in a downtrend. If a market is in a downtrend and suddenly a buy
signal is detected. it could be concluded that the trend had changed to an uptrend. In
this case. lhe use of the tenn "reversal'' isjustified, but in a strong trend, a sequence of
signals of this type can be found, each one confirming thestrength of theexisting trend.

We must specify the time frame from which these signals are supposed to be taken. I
would like to add that these signals are indicative of the trend in the time frame where
they are detected.

The Trend Determination Checklist, including the RSI buyand sell signals, suggests that
there is an easy way of determining the existing trend in an absolute way. ln reality.
applying thechecklist to different time frames leads to different conclusions. Trend is
obviously time frame dependent. For instance, it is important to point out that signals
taken from the hourly RSI should not be used for long-term trend determination. Ir1
other words, when discussing theactual trend, always clarify the referenced time frame.
Lt is not unusual lo detect a buy signal in the dai y time frame, while a sellsignalshows
up in the hourly time frame.

J7
 RSI: logic.Signals & lime Frame
Correlmion

00
r,oOI)

..
SS
W

..,

,.,. AJ

w
...
.
F igu re 1.11 Weekly INTEL

This weekly chart of lNTEL offers some relevant examples of buy and sell signals in
the RSI:

Points A'- B'

The RSI is lower at Point B' than it was at Point A', while the corresponding price at
Point Bis higher than at Point A. This "peculiar divergence" is a buy signal. Also
note that the RS I signal occurred in bulJish territory, well above the 40-level. The
buy signal could be called a ''reversal" signal because it is the first buy signal in this
time frame leading to a rally.

There are other buy signals at Points C'-D', D'-E' aod f'-G'. These are val id buy
signals thatconfinn the strength of the ongoing rally.

PointsX'-Y'
The RSI is higher at Point Y' than it is at Point X', while Price Y is lower than Price
X. This is a "strong"sell signal because the signal is formed within a very tight M-
shaped pattern. A second sell signal is defined at Points Y' and Z'. Past Point Z', a
positive Price / RSI divergence leads to the end of the decline.

RSI Moving Averages

Cardwell pointed out that it is useful to add moving averages to the RSI chart. In
particular, the moving average crossovers mark significant points in the RSI chrui..
The recommended settings for these moving averages are the 9-period simple moving
average
and the 45-period exponential moving average. Add itional,ly it has been suggested that
the same moving averages be applied to price, which may allow a correlation
between both charts.

When a buy signal is witnessed, conservative traders are advised to wait until the
crossing up of the moving averages in price confirms the RSI signal in that particular
timeframe. Asell signal in the RSI should be con-finned by the moving averages in
price crossing down.

In my opinion, this degrades the relevance of this type of RSI signal. The crossing or
the moving averages in price is a signal, and one could just as well state that
whenever these moving averages cross, the existence of a buy or sell signal in the RSI
confirms the pricesignal. In other words, an RSI sell signal could be disregarded as
long as the 9-period simple moving average remains above the45-period exponential
moving average in price, as it is indicative ofa bull market. An RSI buy signal could
to bedisregarded as long as the 9-period simple moving average remains below the 45-
period exponential moving average, as it is indicative of a bear market.

Most analysts, traders and investors keep an eye on price moving averages because
they allow the state of the market to be assessed quickly. The question is: which
moving averages are significant in which time frame? Shorl-term traders may loo k
for signa ls in the 5-period simple moving average and 20-period simple moving
average in the hourly price charts, while conservative long-term investors may be
interested in the SO-period simple moving average and 200-period simple moving
average in the daily or even weekly price charts. Perhaps Cardwell's rationale for
selecting these particular moving average settings was due to the fact that they are on
the conservative. long term side of thespectrum.

In so me of the RSI charts in this book, the 9-period simple moving average and 45-
period exponent ia l moving average will bedisplayed. T find them useful to determine
the preferred RSI channel anchor point in cases where several such reference points
appear to beplausible.

39
 RSI: Logic, Signals & Time Frame
Correlatio11

12000

JA ;JA
II

figure 1.12 Weekly fBM

In this weekly IBM chait, moving averages have been added to the RSI. The arrow
marks the crossing down of the short-term and long-term moving average, which
coincides with lhe start of a major price decline. Points X' and Y' define a sellsignal.

Price Target Calculation

The Cardwell method for Price Target Calculation is similar to the one used in the
Fibonacci-logic. When price resumes its trend after a shallow retracement, we may
expect the extent of the next price move to be 80% to 100% of the previous move. In
an uptrend, if a price correction down is (logically) assumed to be shallow when it
causes a buy signa l in the RSI, we can say that the price move following the RSI buy
signal will take prices to a new high. The price target can be calculated by adding the
extent of the previous price advance leading up to the retracement down to the price
level where the uptrend resumes.

40
 RSI: Logic, Signals & Time Frame
Correlatio11

lOO
O
&10
0
6110
0
&loo
00
00

ll)Jl)

1000

0 l!l'll! A I A

liJ

Figure 1.13 Weekly lNTE L

Let's return to the weekly lNTEL chart and take a closer look at the buy signal that
was formed between Points A and B in the RSI. The price target for this buy signal is
calculated as follows: Determine the change in price between Points A and B and add
this value to the price high in between lhe signal reference points.

The difference in price between A and B is 26.5 - 21 = 5.5.

The price target for this buy signal is 35 .2 + 5.5 = 40.7

41
 RSI: Logrc, Signals & Ttme Frame
Correlatio11

fl" w

O<ll
,sac,

,..
..
..
..
.
"'°

'

" 11)

"" ..

Figure l.1 4 Weekly General Motors

'
This weekly chatt of GM was used earlier to illustrate the 20 to 60 RS1 range inm
bear markets. There is a sell signal at Points X' and Y'. The price target for this
JOW
,000
sell signal is calculated by taking thedifference in price between Points X' and 10.00
Y'
and subtracting this value from the low price between the signal reference points.

The price difference is 43 - 36.8 = 6.2.

The price target is 25.6 - 6.2 = 19.4.

RSI Signal Strength

Which RSl signals will prove to be valuable? What characterizes "s trong" RSI
signals (those which offer the best chance of reaching the price target)?

One opinion supported by Andrew Cardwell is that the fewer periods separating the
RSI signal reference points, the stronger the signal, with a higher probability that price
will reach the calculated target Using this logic, a 3-period RSI signal with only one
period between reference points shol.lld be considered the strongest possible signa l.
This rule holds up; it implies that the price retracement must have been extremely
shallow. But then again, a 3-period signal in the daily RSI chart could co1Tespond to
a 13 -pe riodsignal in the hourly RSI chart.

I remain skeptical about this statement regarding tbe strength of RSI signals as it
implie s that modest price targets barely exceeding the referenced high or low are
more likely to be reached, which is not surprising. Jn my view it is more important lo
 RSI: Logrc, Signals & Ttme Frame
Correlatio11

check whether or not there were adverse RS£ signals present in the smaller and
larger time frames.

Cardwell's Target Price Calculation method starts from the assumption that tbe
price move following a shallow retracement wiU coJTeSpond to I 00% of the price
move prior to the retracement. John Hayden links the probability of an ascertained
price target to the percentage of retracernent that caused the signal. This
Fibonacci-related method results in a matrix of probabilities, ranging from "will
easily exceed price 1argeC' in cases where the retracement is less than 23.7% to
"probably will not exceed 80% of calculated price change'' in cases where the
retracement is over 76.3%.

Studying Cardwell is certainly valuable, as it is an a lternative to the Wilder-style

rules with some practical interpretation guidelines and usefuJ signals. However, some
questions remain. ln particular. the interpretation of positive and negative
Price/RSI divergence is left unanswered. Hayde n believes that simple divergences
are just temporary deviations from the main trend. In this respect, he confirms
Cardwell 's view that negative divergence is merely symptomatic of an uptrend
and positive divergences are symptomatic of a downtrend.

According to Hayden, trend changes are preceded by multiple, long-term divergences

(as opposed to simple divergences). Unfortunately, this description is far too vague to
beof practil:al use when studying RSI charts and a definition of "multiple
divergences" is not given. It is only logical that an RS1 range•shift may be preceded
or caused by repeated, persistent divergences that eventually succeed in pushing the
RS I over the edge. Of course, we still have the RSI range-shift warning of potential
trend changes. lfthe RSIshows signs that point to the half-pipe tilting the opposite
way, the upcoming
move may be a major one.

Recap
Let us review:

There is no way to determ ine, in absolute tenns, when a market is oversold or

overbought. Io most cases, the occasional divergences between the RSI and price do
not lead to trend changes. Also. tTends are time frame-speci fie.

Some interesting concepts were introduced by Andrew Cardwell. He discovered thu1

some peculiar Price/RSI divergence signals can be used as buy and sell signals. Such
a buy signal divergence occurs whenever a lower RSI reading corresponds to a
higher low in price or when a shallow price retracement down in a bull market causes
a lower low in the RSI. A se ll s ign al occurs when a shallow price retracement up
in a bear market causes ahigher high in the RSI.

Price trend is reflected in the RSI in particular time frames in its range of travel. In bull
marke,ts RSJ values will range from (ro ughly) 40 to 80, whjle in bear markets. lhe RSI
range is from (roughly) 20 to 60. Atrend change in price will be reflected in the RSC by
a range-shift.
RSI: Logrc, Signals & Ttme Frame
Correlatio11
43
Cardwen also proposed a method to calculate the projected price targets for these buy
and sellsignals. This method is ta ken from the Fibonacci analys is method, which
says that after a shallow price retracement, the next pricemove with the trend will
be(nearly) equal to tbe one leading up to the retracement. Cardwell's preferred RSI
buy and sell signals are those with reference points that are very close, sometimes
separated by a single price bar. These were called " strong" signals, as he apparently
found that these often propelled prices to their projected targets.

Unfonunately, the reference time frame is not so specific that one can take these RSI
signals from it.Wilder supposedly made reference to the dai{v RSI; Cardwell seems to
indicate that his method is best suited for the weekly time frame.

Both interpretations lack the depth of Time Frame Correlation. At any one point in time,
the price reality can be described bya large number of valid RSI pictures. There is not
jus t one RSf chart that correctly describes a specific price action because each time
frame produces a djfferent RSl picture. There are many RSI pictures describing one
reality. Each picture tells a different (but true) story. so why restrict the RST analysis to
just one time frame?

In the methods we have discussed, some questions remain unanswered, but will be
add ressed in the forthcoming chapters. How does one interpret RSI signals in the
smaller time frames? When do RSI signals become invalid? When exactly is a signal
considered to have failed? What if we find contradict ing RSI signals in a particular
time frame? Also. what is the nature of Price/RSI divergences and what is their
significance? Whal is the significance of the signals ' price targets?
 CHAPTER Two
THE SLANTED RSI UNIVERSE
•
 PARTI

THE RSI DISTORTTON

In this section, I will demonstrate that the best way to sn1dy RSI charts is to
useslanted reference lines rather than horizontal reference lines as we would in
price charts. This means that when prices rally in a staircase-likeadvance, the
resulting RSI chart wiU be in the formation of a series of\-slanted segments or
channels. In a price decline, we will find /-slanted channels in the RSI. Under
standing this RSI logic is essential for further analysis.

When analyzing RSI charts, one of the first problems encountered by the user is that
the RSI chart looks so much different than the price chart. Look at a bull market, for
instance. Prices on the chart rise steadily, while the RSI values on the chart meander
up and down in a nearly horizontal range, roughly between 40 and 80 in RSI bull
teni tory.

Figure2.
1

120

115

110

10
0

Let's try to make the RSI chart resemble the bull market price action more closely.
If the RSI charts are printed on paper, then cut and pasted in a strip, this strip
could be overlaid in such a way that it mimics the price movement in a climbing
range. This results in someth ing that looks like Figure 2.1. which is an RSI
staircase going up in stepssimilar to the underlyingprice moves while the RSI
values are still confined to the range between 40 and 80.

In a bear market, the RSI chart cascades down and oscillates roughly between the
values of 20 and 60 in the RSI bear range. In Figure 2.1, the RSI strip bas been
tilted to
 RSI: Logic, Signals & Time Frame
Correlation

align with the price uptrend. The corresponding rising prices are symbolized by tbe
horizontal lines in the right comer of the chart and are labeled I 00 to 120.

Points of equal price, (e.g. Price 120). in the RSI define a declining Line A-B,
where Point B is at a lower RST level than Point A.

At Points C and D, the situation is more extreme: Point D is at an RSl level of 45

that corresponds to a price level (that is not equal 10 , but higher than) Point C,
which is at an RSI level of 60. This immediately brings us to an important point.
In bull markets , this specific type of Price/RSI correlation appears where a higher
price low is reflected in the RSI by a lower low. This corresponds to the RSI buy
signal introduced by Andrew Cardwell. It is fanned when the RSI rebounds off of a
lower low, while the corresponding price reverses up after a mild correction down.

When th.is appears in the RSI chart, it indicates that the market is a bull market in
1hat time frame. In a bear market, the horizontal line in the price chrutconnecting
points of equal price corresponds to a rising line in the RS I chart. Let me repeat that I
see no reason tocall these" reversal"signals, as theiroccurrence in most instances only
confinns the strength of the trend in force.

A down-up sequence in price trend results in a down-up sequence of the RSI strips, as
depicted in Figure 2.2. In the buJlish part of this chart, Line a-b connects points of
equal price; in the bearish part, Line x-y connects points of equal price.

Figure2
.2 RSI RANGE-SHIFT :
UPDOWN

49
 Figure
2.3
80

If an up-down trend change is represented in one horizontal RS I strip, as if presented

on a PC screen, the chart looks like Figure 2.3. This is what an RSI range-shift looks
like, possibly preceded by what could bea negativedivergence, pushing the RSI below
the 40-level.
 51)

1100
0

Y' 1000
154$ 1.!100 1'20 t.tOO 1121 1g_oo "" t9;00 l!l4 11m l(U5 H28 184115 Hl9 19'5 HOO ta◄: s l.Ot Ht.•s 11.«M HUS 12.A)S 1$4S '2D
I I >j
CloM6d WIilT1adeSlallar'I

Figure 2.4 LRCX (15- minute)

This chart was used earlier to illust rate the " peculiar divergences" that exist
between price and the RSI. To this example, the RSL hits resistance at Point Y'.
If this drop in the RSl is compared to the drop in price from Point X to Y, we can
see the effect of their logarithmic con-elation.

The modest price rebound from Point Y to Z causes a spectacular jump in the RSI
from Point Y' to z·. T his RSI distortion of price reality leads to the following
conclusion: In
a bear market , points of equal price (Line W - Z) will define a rising reference line
(Line W'- Z') in the RSI. So, if we study price changes in relation to their horizontal
references or price levels, we must then study thecorresponding RSI changes in
reference to their slanted reference lines.
 RS!: l ogic, Sig,wls & Time Frame Correlation

tU5" 11}00 1107 1900 IIAlB 1900 ''°' 19'0C 1,no 19:tll 111, 1100 tltAI 1i"OO v,s 18:UJ 1HHi 16100 nn, 18tl'.l '1ml 1900 ,m HHO

F igure 2.5 LRCX (15-minute)

Ln bull markets, due to RSI distortion, the line that should be used as a reference
to study changes in the RSI versus changes in price (Line D-E), is a declining line
(Linc D'-E').

PARTII

RSI ZOOM-IN

Another particularity oftbe Relative Streng1J1 Index isthat its picture changes
significan tly when switching from one time frameto another. Price levels remain
unchanged whether they are examined in a daily time frame or in a I 5-minute chart.
This is not tJ1e case in the RSI. The outcome of the RSI calculation in the hourly
timeframe, typically covering l4 hoursof data is totally different from the RSI picture
in the daily chart covering data of the last 14 trading days.

This time frame speci fie behavior makes the RSl study interesting because it allows
us to analyze and correlate various RS[ pictures describing one price action within a
given timeperiod. This is like looking at a particular pricebehavior from various
angles, which can be correlated because they describe different aspects of the same
reality.

51
Figure 2.6 PRICE-RSI/ TIMEFRAME RELATIONSHIP

I WEEKLy PRICE

·-• -• _ .. ,OM _,, , ,. MM o H ,,. ,., ,

I WEEKLY
RSI

I DAILY
RSI
S2
 RSI: Logic. Signals & Time Frame Correlation

Let's take a look at Figure 2.6. The chart shows a long-term buJl market in the weekly
chart. The course of t)1e weekJy RSI (more or less) resembles the course of the price
action in that it is heading up for most of the time, meandering within bull territory.

Looking at the same price action up for the same period of time, but examini ng
the correspondingRSI chart in the daily time frame, we notice that the general
aspect of the chart does not directly reflect the uptrend. The RSI values oscillate
up and down, roughly between the 40-and 80-level. This fact indirectly tells us
that the trend must be
up.

Zooming in even further on the hourly chart, we see that there are more frequent and
wider oscillations in the RSI that often alternate between bull and bear territories. The
reason for this behavior is that relatively shallow price retracements accompanying a
bull market of this magnitude are sufficient to push the hourly RSI values into its bear
range for a period of time. Strong moves upward may lift the RSI above its normal
values, occasionally reaching a value of 85 or more. This is often the case after a
shallow price correction down. when p1ices rally strongly. In the RSl calculation
period of 14 hours. momentum turning from mildly negative to strongly positive
results in the largest momentum jumps, relatively speaking.

In tbe hourly RSI chart, the uptrend will be indirectly visible because the RSl
spends moretime in the range between 40 and 90than it does in the60-to l 0-zone.
Additionally, within the 40- Lo 90-range, a se ries of buy signals provide further
indirect evidence of the uptrend. There might be the occasional sell signal, caused
by price declines that lead to the formation of buy signals in the daily time frame.
These price declines that are shallow in the daily picture. may tum out to bedeep
enough to cause sell signals in the hourly time frame. From this series of
comparative charts, the need to correlate different RSI lime frames in order to
assess the market conditions and anticipate high probability scenarios becomes
clear.

My prefen-ed set of time frames for RSI analysis is the weekly-daily-

hourlycombination. For shorter-tenn trades, I prefer to use the daily-hourly-15-mintue
combination.

PART HI

RSI SEGMENTS AND CHANNELS

In a bull market. rising prices fin<l their expression in the RSl chart in the up-down
cycles in bull territory. This ischaracterized by the peculiarity of lower lo ws in the
RSI value corresponding to higher lows in price. [n a bear market, the RSI up-down
cycles in bear territory are characterized by higher highs in the RSI corresponding to
lower highs in price. These features were used by Andrew Cardwell to define his buy
and se ll signals.
53
Remember that these signals can be used to determine trend direction and in some
cases point to possible trend reversals. We also find the need to use slanted reference
lines in the RS I chart when studying price change. This leads us to study the RSI
price relationship by examining RSI behavior in relation to these slanted reference
lines. Jt makes sense to anchor these reference lines on RS I extremities with some
significance. For example, we could use the buyor sell signal reference points. Let's
see where this idea leads us. Figure 2.7 shows the 15-minute LRCX chart that was
studied earlier when discussing the peculiar Price/RSI divergence that produces the
buy and sell signals.

" -S&!ll
ssoo

1114 1900 11ns 19W 11116 HIOO tlt7 tll'OO 1/20 191)'.] 1 m 190, tt:J
j I)
c,.,.,d., hIIOOISt,1.,.

F igure 2.7 LRCX ( IS-minute)

In this chart, we find that there is a buy signal at PointsA'-C'. If Line A'- C' is used as
a reference, other significant parallel lines can be defined as well.

For instance. Line B ' -F' -G' runs parallel to Line A'- C' , starting from the RSI top at
Point B'. There is another buy signal at Points F' -G'. In fact, a buy signal is formed
when the RSI bounces off the top reference line of the previous RSI slice or segment.
This is what happened at Points A'- C' as well. We end up with three RSI segments in
thischart.

There is one segment below the Line A'- C; there is another RSI segment defined
by Lhe lines A'- C' at the bottom and B'-G'at thetop; and finally there is the one
building above the Line F'- G·. These RSI segments correspond to the different
steps of the price staircase. The arrows indicate where the RSI jumps from one
segment into the next higher one. Each time this happens, prices rally with
renewed vigor. Let's call these RS! segments "channels." Later, we will see that
these channels are being formed within a larger Channel2.
 RSI: Logic, Signals & Time Frame
Correlation

IL should be noted that large Channels2 may correspond to " norn1al"channels in the
larger time frames.

Descending channels are typical for bull markets; ascending channels are relevant in
bear markets. rn practice, we will look for significant descending or ascending
channels
whenever tbey can be anchored to RS£ signal reference points and clearly correspond
to the consecutive phases in the w1derlying price behavior. Important lessons about
probable future price moves can be learned by watching the RSI move within or
break out of established channels.

Traders who know the Elliott Wave theory will be familiar with the concept of
the "Russian Dolls" aspect of this analysis method. As you zoom in on an
impulsivewave, sludying parts ofit in smaller time frames, you wiU find
eversmaller identical impulsive waves as subdivisions within theoriginal, larger
impuls ive wave.

The same goes for the various RSI channels. What is labeled a "ChanneF" in one
time frame could very well correspond to a channel in a larger time frame. In this
respect, these channels provide a link between various time frames and can be
used as a tool for the correlation of the RSI pictures. Most relevant will be
thestudy of RSI value moving within thechannels, of the break-outs from these
channels, and of theeffect this has on its ChanneF or on its channels in the larger
time frame.

FIGURE
2-8

Chann
el2
- -- -

-- - -
CHANNEL3 ----
----
-
F igure 2.8 illustrates how channels of various sizes appear in the RSI chart of an

55
 RSI: Logic, Signals & Time Frame
uptrend. The initial primary RSI channel Correlation
leads up to Point X, where the RSI jumps
into the next RSI segment, or channel. Suppose a buy signal is formed at Points A- 8,
so Line A- B - D can be used as the lower boundary of a descending ptimary channel.
J\s the RSI reverses up at Point D, at around the 40-level, we could imagine that there
is

55
another buy signal that defines the lower boundary of Channel2. At Point C, the RSI
jumps into the nextRSl channel.

In short. we have primaiy up-down RST channels withfo a descending Channel2,

anchored on (DD +) reference points. All of this could be situated within an even
larger ascending ChanneP. All charu1els, no matter their size, have RSI signal
reference points on their lower or upper boundaries. To find these larger signals, we
switch to a larger time frame, where the ChanneP and its reference points will be
clearly visible.

1IXUl
O

10 00

Figure 2.9 Weekly Dow Utility Index

This chart illustrates thechannels-within-channelsconcept. The descending RSI

channel is the down leg within the larger descending ChanneF. This ChanneP is a
down-leg within the ascending ChanneP. Note that the RSI low in April of2006 is
near the 40- level in bull ten-itory.
 RSI: Logic, Signals & Time Frame
Correlation

3800!)

JEIUX>
3'l)OI)

:moo
JIJ!
OO
l\llJIO
MIA>
1'000

iOilO

r---1::------ ,
iT"""----;----.r======------...-.... 00

.....

101)1

Figure2.I0 Daily Dow Utility Index

This is a close-up picture of the weekly chart where a negativedivergence defines the
upper boundary of thedescending Channel2. Note that the low price in April 2006 does
not cor respond to a lower low in the RSI as it did in the weekly chart. The dashed line
in the RST defines the lower boundary of a potential new ascending channel. Tn the
weekly chart, this would mean that the recent RSI rise from the bottom of its ChanneP
is the start ofan up-leg within ChanneP.

PARTlV

RSI CHANNEL LOGJC

We have analyzed how the RSI formula distorts the underlying price picrure and
lransfonns it in a segmented RSI strip. These RSI segments, or channels, have
properties depending on whether theyare a product of bullish or bearish conditions.
Consequently it makes sense to anaJyze the RSI by checking which channels are
meaningful.

Up-trending markets are characterized by descending cha1mcls where buy signals

will form, indicating that the rally is continuing. In fact, we will look for RSI channels
that are anchored on the buy signal reference points. Downtrends are characterized by
ascending channels, where sell signals will fom,. The significant channels are those
anchored on the sell signal reference points in the RSI. When an RSI buy signal
appears in a downtrend or an RSI sell signal appears in an uptrend , these may be
early indications of a trend reversal in that time frame.

A A

Figure 2.11 Daily BG

Let's analyze the BG daily chart. Our first task is to find some relevant segments in
the RSI. LinesA, Band C connect some tops and troughs in tbe RSI chart. Are these
lines significant? Line A connects higher highs in the RSI, but these correspond to
higher highs in price, so there are no sell signals defined on Line A. Line B connects
some RSI lows at the left of the chart, then, to the right of the center of the chart,
Line B connects some RSI tops, but again, the RSl Lops correspond to higher highs in
price, so there are no sell signals here either.

No wonder there are no sell signa ls here. The price trend is clearly up in the right
half of the chart! Line C seems to havesome significance as it connects some RSI
extremes, but no buy signals are found on Line C.

But wait a minute. If prices are in an uptrend, we should be looking for

descending channels (where there should be some buy signals).
 RSI: l ogic. Signals & 1i,ne Frame
Currelaam,

Figure 2.12

This is much better! The descending channels make much more sense. They bad been
staring us in the faceall the time. The segments in the RSJ correspond to the
consecutive phases in IJ1e pr ice rally. The switch occurs when the RSI breaks out
from one segment and jumps into thenext. Where are the buy signals? Actually,
several were detected in thischart.

Line A connects lower lows in the RSl that correspond to marginally higher lows in
price, so line A defines a modest buy signal. Lines Band C define the bottom of the
second RSI segment. Here, we also have a buy signal. From the Line B - the RS[
,C rises and pops into the third segment.

Lines D and E define buy signals. Please note I.hat the RSl in thjschannel doesnot
rise above the 60- to 65-level. As we have learned, this indicates market
weakness.

At PointF. the RSI breaksout once more in a FASTbuy signa l with the RSI well
within bull tenitory. The price advance becomes steeper in this section and
consecutive shallow retracements in price cause buy signals in the RS I.

Let's move from this discussion and define the 2nd degree channel in this char t as
shown in Figure 2.13. When the extremities oftJ1e RSI channels are connecte d in the
previous chart, a Channel2, which ca n be used for further reference, is obtained.
 A

2.13 Dajly BG

The upper boundary of Channel2 reaches the 95-level at the extreme right side of the
chart. Does this mean that the RSI could keep going up to that level? It is certainly
possible to see the RSI reach the 90- level, as it did in January 2004, but that is not
required for ptices to continue their advance.

In November and December of 2004, tJ1e RSI struggles at the 80-level. Each time the
RSI reverses up after a setback, prices reach oew highs. So, we could see the
development of a negativedivergence with prices flat or still rising moderately while
the RSI drops to more reasonable levels, possibly to the lower boundary of its
Channel2, where another buy signal could form.

There is a similar scenario between January and May of 2004, from Line B - C.
After the RSI hits the 90-Ievel, a negative divergence pushes it down toward the
lower bow1dary of ChanneF, while prices keep rising. The pricedecline that
follows in April of 2004,as the RSI is finally pushed below 60, results in a shallow
retracement. The reversal upof the RSI in May 2004 triggers the buy signal. Also
note the steeply out-of-proportion drop in the RSI in February of 2004, which
leads to the formation of a subsequent buy signal.

Let's discuss how the end of the rally will become visible in the RSI chart.
lmagine the RSI meandering in one of its consecutive bullish segments, like the
ones we have just
60
 RSI: logic, Signals & Time Frame Correlatinn

discussed. However, this time the RSI does not break out into the next up segment, but
rather remains within its descending channel.

As the RSI value drops lower within its descending channel. it will reach bearish
te1Titory at some point in time; no more buy signals can bedetected on the lower
boundary of the channel. If the RSl remains unable to break out, sell signals will
appear within the descending RSl channel. In such a case, this initial sellsignal could
becalled a "negative reversal," as it suggests the end of a series of b11y signals and
may lead to a trend change.

[n this trend-reversal process, it takes quite some time before the fust sell signal is
detected in the daily RSI chart. After all, a sell signal forms after a shallow price
retracement up within a price decline. In the daily time frame, such a pricedecline
can be quite substantial, meaning thatsome serious losses could be suffered before
the selI s ignal materializes.

In order to limit the time and the price swings required to produce a sell signaL
switch the analysis to thehourly time frame, There willbe other signs in the RSI
warning usof potential trouble. In the case we have just discussed, the first warning is
the RSI-drop below its Chan.11el2.

Figure 2.14 shows a potential scenario for the BG Chart previously discussed.
We left the BG Chart with the RSI moving within its ascending Channel\ which
correspo nds lo the channel at the lop left of the hypothetical RSI chart.

At Points A- B, there is one morebuy signa l within the Channel2·. At PointAl, the
RSI fails to break out and jump into the next-up RSI segment. Prices reach a new high
because we are around the 80-level in the RSI chart. At that level, each reversal up in
the RSI takes price to a new high. As a result, there is negativedivergence at PointA l
.

At Point X, the RSI valuedrops below Channel2, but is still well within bullish territory.
As the RSI reverses lip from Line I , it prod uces a buy s igna l and the RSI rebound to
Point Y takes price to a new high again. In other words, the negative divergence
persists. After this buy signal, if the RSI manages to stay within its new ascendfog
channel or above Line 1, the RS[ will be rising to new highs in bull territory and it wi ll
put an end to the negative divergence. However, the RSI does not rebound off Line l; it
breaks below it.

No problem, the RSI is still in bullish territory (although getting closer to the bearish
zone) and the reversal up from Line 2 produces yet another buy signal. This time the
RS I rise,following this buysignal, results in a lower high in price. The rally is
weakening.

By now the RSI is caught in adescending channel. This is typical for an uptrend as
long as buy signals fonn at the lower boundary. So far, this has been the case. At this
point, if the RS I manages to stay aboveLine 2, a new ascending cha111e l will
develop, parallel to the initial cha nne l, but at a lower RSI level.

61
 Figure
2.14
Channe
l2

- -
-
Buy
1
CHA

------ -
NN EL3

-
The RSI fails to stay above Line 2 and fails to break the negative divergence. As the
RSl breaks below Line 2, it drops to Line 3. However. the reversal up does
notproduce a buy signal. When dropping to Line 3, the RSI enters bearish territory,
breaking well below the 40-level. This means that small moves down in the RSI
correspond to larger price Jrops.

The rise in the RSI, following the rebound off of Line3, ends with a reversal
down after reaching Line 2, this time forming a sell signal in the process (reversal
signal), while failing to break out of the descending channel once again. The RSI
breaks below Line 3 at a value well below 40 and the price decline accelerates.
Eacb time the RSI turns
lower, prices get hurt. rn a final decline, the RSI reaches the 20-\evel just shy of Line 4.
as prices continue their drop. At last the RSI hits its mathematical resistance and starts
to curl up past Point C.

This does not mean that the pricedecline is over. It is the start of a positive
divergence into Point E. In thjs zone of the RSI chart, each rise in the RSI means
little change in price, while each reversal down in the RSI causes prices to move
lower quickly. Past Point E, the positive divergence persists as the RSI stays
above- Line 4 and the RSI finally escapes from its descending channel.

Again, this doesnot mean that price resumes its uptrend, as we haveyet to see any buy
signals. Past Point E, the RSJ pops up and reverses down from Point F, forming a
sell signal. Following this sell signal, if the RSr should drop below Line4, prices
would get hit hard again.
RSI. Logic, Signals & Time Frame Correlation