Академический Документы
Профессиональный Документы
Культура Документы
echnical
Tr~der .
,'1 '
I'
Martm S. Fridson
OYNAMIC ASSET ALLOCATION
David A. Hammer
INTERMARKET TECHNICAL ANALYSIS
John J. Murphy
INVESTING IN INTANGIBLE ASSETS
Russell L. Parr
FORECASTING FIN ANCIAL MARKETS
Tony Plummer
PORTFOUO MANAGEMENT FORMULAS
Ralph Vince
TRAOlr-.G A'IO 1 VESTING IN BONO OPTIONS
M . Anthony Wong
THE COMPLETE GUIDE TO CO'IV[RTIBLE SECURITIES
W O RLDWIOE
Laura A. Zubalake
MANAG ED FUTURES IN THE INSTITUTIONAL PORTFOLIO
Charles B. Epstein, Editor
ANALYZING ANO FORECASTING FUTURES PRICES
Anthony F. Herbst
CHAOS ANO ORDER IN THE CAPITAL MARKETS
Edg.1r E. Peters
INSIDE THE FINANCIAL FUTURCS MARKETS, 3 RD EDITION
M ark J. Powers and Mark G. Castelino
RELATIVE D IVI DE O YIELD
Anthony E. Spare
SELLlr-.G SHORT
Joseph A Walker
TREASURY OPERATIONS ANO THE FOREIGN EXCHAt-.GE
CHALLENGE
Dimitris N. Chorafas
iv
THE NEW
TECHNICAL TRADER
Boost Your Profit by Plugging
into the Latest lndicators
J.
Baird
UNDERSTANDING SWAPS
John F. Marshall and Kenneth R. Kapner
J.
Bauer, Jr.
@
JOHN WILEY & SONS, INC.
New York Chichester Brisbane Toronto Singapore
To my other cooulhors
Vid)'a, Raw, and Aroon
llus \Cl1 ts pnrued on acad--free p1ptt
Copyn&ht
c.
TC
Chande, Tuhar s.. l 9S8The New Technical Trndcr : lloosc Your PTOfic by Plugging
n10 the La1e.1a Jnd1ca1ors / T ushar S. Chlndc aod Stan.ley
Kroll.
p.
cm. - (Wlley ftnancc edil101u)
Jncludes bibtioniphical rcfcrcnccs and l ude.A.
ISBN 0-471-S9780.S
l. lnvcsuncn1 analy11i.-Da1a proctS$1ng. 2. Futurcs
markcc-Da11 proee$$1n. l. Kroll, S11nlcy. JI, Ti~<.
111. Senes.
H04S29 C'4S 1994
332.6'028S--Oc20
10 9 8 7 6
4 3 2 1
93-3663
C'IP
Preface
PREFACE
Acknowledgments
xi
Contents
1 An Abundance of lndicators
19
49
Contents
xiv
73
93
143
161
181
199
195
1
An Abundance of
lndicators
An Abundnce of lndicators
'5.llD
'A~~
411.llD
'
3SOl)
1,
'.DOl)
e
:1500)
'1
tSOCO
10.CXD
ftb
M"
a
Mo
You can fmd brief descriptions of different indicators in various books or software roanuals, but you wi ll rarely see th e
derivation of these indicators or an analysis that describes
wbat is new or different about an indica1or. We think i1's
reasonablc to ask: What do the Lndicators mean? Hcre we will
examine several popular Lodicators based oo price and show
how similar they are.
Tbese popular indicators measure price momentum in one
form or an other: directional movcment system, momenturo,
relati ve strcngth index (RSI), stochastic oscillator and commodity channel index (CCI). The William's %Risa complement of thc stochastic oscillator; hencc, it has the same Loformation as 1hc stochastic osciUator. The price-rate-of-
An Abundance of lndicators
~~~~~~~~~~~~~.:..;;;..:.;;:.:::.:::.:.:..:::.::....::::...:..:::::.::::~
Figure 1.3 shows the similarity bctwecn the 14-day RSI and
the 14-day plus d.irectional index indicator for MO. Observe
how both indicators peak and bottom together. Figure 1.4
with
compares the 14-<lay commodity channel index
the 14-day stochastic oscillator. T hese two indicators are also
similar in appearance, and their turning points occur at about
the same time.
The CCI is compared with a price oscillator in Figure 1.5.
Tbe price oscillator is !he ditTereoce bctween today's close
aod a 14-day simple moving average of tbe close. The CCI
behaves like this oscillator, lhough it bas ditTerent scaling
factor; both track momentum (see Figure 1.6).
There are other examples ofusing the diJference in moving
averages to measure momentum. For example, notice how
the 14-<lay momentum to the popular moving average con-
cccn
.v...v.
~~,,........~./\.,.....
IDO
2.00
10 m
00
(.::::.;:.:.=-=:.=..::.:::.....-=-~__..:.~~~~~~~~~~~~1-00
00
1-~.:....-=-=:....:..~__:~~..x....:.....:....-l1.:-~-=-1..o<===L-~---l ~m
N
93
93
An Abundnce of lndicators
OOl
SOOl
PRIC~
OSCUAi;oA C.A~RAGE(C.!'l
OOl
OOl
l!il OO
moo
SOLIO~: ce~ OASHEO ~ PRICE OSCl.lATOR
o
FIGURE 1.4 A comparison of the 14-day commodity channel index
to the 14-day slow stochastic oscillator for the same data as in Figures
t.2 and t.3.
93
..
An Abundnce ol lndicators
ro aoo
93
FIGU RE 1.6 A comparison o lhe 14- day momen1um and the price
oscillator in Figure 1.5 far Philtp M orris stock.
FIGURE 1.7 A comparison of the 14-day price momentum to the moving average convergence-dovergence indicator or Philip M orrs s10<.k.
r'
0.93
0.78
0.93
0.77
0.78
0.82
10
An Abundance of lndicators
11
its " lcngth" automatically, based on price a.ction? Tbe v_ariable length dynamic index (VIDYA) does JUSt that. h is a
modified exponential moving average tbat adapts to market
volatility, incrcasiog its lengtb wh~n prices trad~ in a narrow
rangc, and shortening it wben pnces move rap1dly. ~YA
slows down when prices are quiet, and speeds up when pnces
make their move. You can adjust the resp0nsiveness, or dynamic range, to suit your trading stylc. Tbe dynamic range
is thc range of effective lengtbs tbat VlDYA can use, say, from
J to 30 days. VlOYA is therefore a lcxible moving avera.ge,
which is a significan! improvement on fixed-lengtb movmg
averages.
Momentum Oscillators
The Chande momentum oscillator (CMO) is a pure momentum oscillator. The stochRSl combines tbe p0werfuJ ideas of
relativc strength and a rangc location oscillator. They will
often show pricc extremes that the relative strengtb index
(RSI) will not, hclping to overcome its limitations.
Thc CMO can show oet momenturn at a glance, and can
be combined with VlDYA to forro a dynamic average keyed
12
An Abundance of lndicators
to market momcntum. The stochRSI quickJy shows price extremes and momentum swing failures because it reaches new
lows and higbs faster tban the RSI itself. You thus can combine the two popular ways to use RSJ into a single indicator.
We complete our oscillator group by extending the idea
bebind VIDYA to RSI, defining tbe dynamic momentum index. DM 1 also adjusts its own length using market volatility;
tbus, you do not have to specify lhe number of days in the
calculations. DMI often leads RSI into overbought or oversold territory by many days, a usefuJ featllJ'C most traders
couJd exploit.
Contingency Planning
Contingency planning sbould also be part of a risk control
strategy. We look at the practicaJ issue of developing a gamc
plan, suggcsting a way to projcct tbe possible price range for
the next day using absolute momentum. These numbers then
providc targets for a "what if?" simulation, a proactive approach 10 trade planning. Tbis will help you preplace your
orders and trade more mechanicaJJy. In our discussion we
also touch u pon practicaJ issues of trailing stops, trading tactics, portfolio selection, and asset allocation. Sorne of these
ideas may alrcady be familiar to you.
14
An Abundante ol lndicators
s. -
Thcre are many new ideas in this book that you can integrate
into your trading s tyle. The ideas are flexible and powerful,
so you can easily adapt them to your analytical approach and
planning prooess. The more ideas you can integrate, the more
you can boost your trading pro6tability.
Co - C,.
(1. 1}
ICo - c,1
(1.2)
(l. 3)
15
S4
(1.4)
( 1.5)
s..
.
(Co - LL,.)
s1ochas1tc - (HH,. _ LL,.)
(I.6)
The el ose tcnds to be near the higb or Jow of the day wben
markets makc new highs or lows. You can check many price
cbarts to vcrify this observation. Hence, highest bigh can be
replaccd by the bighest close (HC). This is usually a good
approxirriation over tbc calculation period. Similarly, the Iowest low is rcplaced by the lowest close (LC).
.
(Co - LC,.)
stochasuc - (HC,. _ LC,.)
( l. 7)
This now takes on thc appearance of a momentum calculation, where the number of days can vary between today
and x days. Assume that the lowest close was 14 days ago.
The numerator would then be (Co - C,.), a momenlum calcuJation. Thus, we expect to see a broad similarity between
16
An Abundance of lndiC<ltors
(1. 8)
CCI
D / (0.0l 5 DJ,.)
(1.9)
Note that the nume rator D determines the sigo aod change
in CCI. We can replace the mean p rice by the daily close C.
This provides an cxcellent approximation more thao 95 perceot of thc time. Hence, the deviation D now becomes a
momentum-like calculation, where we take the difference between today's close and its 14-day simple moving average
(C..):
( l.10)
17
>
rand(), 1, - 1)).
The built-in Excel function rand() rctums a random number between O and 1 inclusi ve. This rule generates two diffcrcnt random numbers; if the 6rst number is greater than
thc second, today's price is greater than yesterday's price
(+ 1). O therwise, today's price is lower than yesterday's (- 1).
To start the series, we assume that the close of tbe 6rst day
is 40.00. This eould be lhe price of a stock or the price of a
commodity in cents per pound or cents per gallon. We also
assume tbat today's high or low c.an be no more than 2 cents
above or below the close. We thus generate a random number
18
An Abundance of lndicators
between O and 2 and add it to the closc to get the higb. Similarly, we generate anotber random number between O and 2
and subtract it from the close to get the low. Tbc rules are as
follows:
today's higb - today's close + rand( ) * 2
today's low - today's close - rand() 2
t.oday's close - yesterday's close + rand( )*2*( + 1 or -1)
Tbe Excel function rand( )*2 generales a random number
berween O and 2 inclusive. For day 2, we generate a random
number betweeo O aod 2, muJply it by either + 1 or - 1 as
determined before, and add it to the previous close. Next, we
find two more random numbers to compute the bigb and low.
We continue the process for ten periods. You can add
rand( )*3 to increase tbe range of price action.
2
Linear Regression
Analysis
20
chapter, you'll discover how to trade with the slope and trend
strength.
Using !he equatioo of the best-fit line, you can estimate
values of the possible prices for the oext trading day. The idea
is not to predict the precise high or low for the next day,
although you could occasionalJy come quite close; it is instead
to have price targets that can be used to develop a game plan
for trading. You can tbeo trade objectively within the heat of
banle.
X+ C.
(2. 1)
Here x is the indepeodent variable, y is the dependent variable, m is the slope, aod C is a constant intercept. You can
imagine a plot along the x and y axes of the two variables.
This equation describes their relationship in a quantitative
form.
The output of the regression calculations gives values for
m and C. We also get !he coefficient of determina ti o o, denoted
by r2 Refer to the tutorial at the end of this chapter to sce
why r2 measures the relative trend strength.
We like to use five days of closing data for short-term trad
ing using regression aoaJysis. You may wish, however, to experiment with the number of days used in the calcuJations,
as well as try the daily high and low prices for mak:ing fore
casts.
Samp/e Calculations
Let's look at a sample calcuJation using the closing pric.es of
a recent gold futures contract Assuming time as the inde-
21
B
y
x-squared
D
y-squared
E
xy
1.00
2.00
3.00
4.00
5.00
378.10
376.10
379.50
379.20
388.20
1.00
4.00
9.00
16.00
25.00
142959.61
141451.21
144020.25
143792.64
150699.24
378.10
752.20
1138.50
1516.80
1941 .00
15.00
1901.10
55.00
722922.95
5726.60
A
X
Sum
n q, q, q, -
sn -
slope -
q, / q, 0. 1 (5726 6
22
culations, q,, Qz, and q,, which are computed using rhe totals
in the various columns and the number of data points. You
can find the slope and intercept of the 5-day regression from
the following equations:
slope(5 day) = 0.1 *(sum of producl - 3*sum of y values),
(2.2)
intercept = 0.2*sum of y values - 3*slope.
23
390
388
386
384
382
380
378
376
374
372
370
O"
... .. .
go
....
...~--"
.. .:.:--~-----"'
God futuras dallv close
o.,..
FIGURE 2.1 An example of linear regression calculations using the
daily dose of the August, 1993 Comex Gold futures contract. The
solid line j oins the daily closing prices.
....
390
Unea1 regrNsion
\...
385
380
....
.... ...
.. ~
'
375
3
Oaya
FI GURE 2.2 The day-6 closing price forecast using the " best-fit" linear regression line.
24
and negative wben tbey are fa!Jing. Tbe slope measures tbe
cxpected change in price per uni1 of time, wbich, wbeo coovcrted into dollars per cootract or share, iodicales iftbe m arkct is making big or small moves. The slope cooverted into
dollars is a useful filler for trading or ignoring markets.
Linear regression aoalysis also gj ves us 1he strenglh of tbe
linear relationsbip, denoted by r2, the coefficieol of determination. lt is a quick measure of 1he trendiness in the data,
ranging from O to l. Ifthere is no trend, that is, random price
ac tion , the r2 value will be close 10 O. A perfect linear trend
givcs a value of l. Figure 2.3 shows two simulated sets of
data: a random set and another with an exact linear trend.
Thc r2 val ue is close to O for th e random set, and is 1.0 for
the cxact linear treod.
Using r
Thc primary use of r2 is as a confinning indicator. It is a
lagging indicator that shows the s1rengtb of tbe trend. The
critica! val ue ofr2 depends on the number of days in tbe data:
whcn the r2 value is greater than the critica! value, a statis-
12
10
......
---
......
_...
'
',
' '
2
o ~~~~~~~~~~~~~~~~~~~~~--'
1>..,.
25
26
90.000 . . . . - - - - - - - - - - - - - - - - - - - - - - - - .
27
1 6-0t't' SIOl)4
1.5 '-----~-------------------'
08/03/92 0 8131/92 09129/92 10127192 11124/92 12/24/92 0 1/25/93 0 2123/93 0 3123/93
FIGURE 2.5 The 15-day slope and r' from a linear regression analyss
of the CSI data in Figure 2.4 for coffee.
2-r-----:-- - ----------.----.-----.
7-daJ slope
1.5
'{
0.5 .
80.000
0 .5
70 .000
60.000 .
50.000 .
40.000 ...__ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _.._
FIGURE 2.4 The daily close of the Commodity Systems, lnc. #39
Perpe1ual contract for the coffee futures prices.
.,
. , .s
1S-ctay slope
FIGURE 2.6 A comparison of the 15-day and 7-day slopes for the
coffee data in Figure 2.4.
28
29
, -/'J _
. ....
! ~ iT
cv.
(\14
>
,.
_J
: .
, (
,. ':
~ ~
1\ ..,.
- . \...) /JtJ . ~
- '
'\ . '.
+,. :
! ... .,;,,_,,:.,, ..;. . . ti ... ' . .
.l.. -
....
~
..
"'
,.
'J,.. .."'
....
'.
., ,
.. .':;
;::;
0
:.
:
;
i,,,r'
. ! - :;
' : . f
1.
'
:~." ~
1 1
~ r";~~.,.,~:"
,. .. , ,.
ri
'91
11
t '1~ ; ,._. ,
l'I
l'I
.i ;
. : :
"
~
;
= .. . . ~ ~ . :
,
'
l'I
"
l'I
,.
,,_
.."'
l'I
..,.
....
....
.)
..
....
..
'
'
-~
.,
E L
. - ~ ...
;)O
20
'
'
10
!OC
'
7 11
... '
+- "
_,.,
"
'
.1
'
,.
'
'
..
'
'
..
31
,,
'
..,... .
+ '!
............. '
.i
.,
~
r
~I\.'. . ' - . . ~ .
")!:!
'
'1
1'11'1"
:u"\1x- '
JI
1'1
SO
'
''
'
w0
' 92 f
..,_, '"'
nl\.ltJ
''
l\SO
NO
to
l ,.
eo
,..,
'
'
, r
....
.,... .,.
..
.,
.. .
.. .
f ~
:: .: : : .
..
.. :
40
. 'l{.
'
'
t 1
l'I
'
'
1/': ....
~~~'(!
...
"'
"
"'
- .. .
r;,..: >;,u:;
- '1~
~
,;.; .
,,
. ..
..
i-
-- . . . - .. -
. <
-~ ,,,,,
, ., . , i ;
:,. : :1:
... ~.
! ..
"V., ftl""
o
,.
o . ., r
' '
'
"
"'
"
'
'
.1
,.
''
"
FIGURE 2.8 The 14 -day simple rnoving average o f the 14-day vertical
hori zontal filter used as a measure o f trendiness in lntel stock.
30
.'
-......
o..,... u - ..,_,......,
~,..,,
"-
of
,_...., .-. .
I~..
,,,.,.....,~
,.
,,..
::'.
....
..
........
..
'
"'
'"
1:
.'''
: ,!
20
'
,.
..
u
.
1 '
...:
'
s1
t1
.s
"'
FIGURE 2.10
....
....,.
..,...
..
........
"'
l'I
o n ,. "
"
"
'
.J
"
li
lntel.
..
....,.
,.
....
......"
......
..
~
folid , . .,..
~--.ci...,
__ ............
O
- u,,.,
- '""'
I~ ...l ... lo -o"lf'>O
..
:
""
. '.
'.
,.:
'
'
'
r:
:
1
~; :
"
'
' ....
'
' J'
..
v
'
i(
"
r ;
:: '. : :
: '' i :
....'.
'
'
-t i
ti
11
ti
"'
.o!
,_.,
,.
11
"
ti
.J
/1
..'
.."'
....
..
20
'
'{
...."'
..
" ,."'
! '
~~
.:.~~~..
-f'~
...."?,
,'!
""'"'""",."e1
FIGURE 2.11
33
T-Bond
A Common Weakness
For~cuts
35
Strategies
We don't cxpcct to forecast precisely the actual bigh, low, or
close. The purpose of the forecasts is to develop a range of
cxpccted prices for the next trading day, and then to use the
range to chalk out a trading plan for specific contingencies.
You can plan a variety of actions with the forecast, such as
sctting stops, initiating new positions, or closing old ones.
You can estmate your risk and reward, and write orders to
takc profits, cut losses, or selecti vely add or reduce existing
positions.
A T-Bond Spreadsheet
36
104.5
104.969
105.094
104 .938
104.719
104.813
105.531
105.906
105 844
105.688
105.563
105.5
105.844
106.156
105.875
104.156
104.344
104.438
104.375
104.375
104.406
105. 188
105
105.563
105.219
104.563
104.563
105.031
105.219
105.344
104.313
104.5
104.719
104.938
104.469
104.781
105.344
105.875
105.563
105.438
104.625
105.344
105.5
105.375
105.844
HIGH
LOW
CLOSE
104.966
104.7
105.244
106.006
106.366
106.175
105.659
105.372
105.631
106.116
104.478
104.406
105.01 6
105.287
105.797
105.675
104.797
104.419
104.472
105.059
104.813
104.775
105.178
105.906
106.191
105.859
104.806
104.769
105.228
105.481
Tbis defines the first five values ofthe bigh in cells B2 through
B6 as the o oes to be used for the forecast. By not speci fyiog
the range of x values, we are using thc dcfault valucs 1, 2, 3,
4, 5 for the fivc days. This is exactly what we did in Table
2.1 earlier. The !6} givcs us tbe forecast for the oext tradiog
day. The last i1em, TRUE, says thal lhe inlercept should be
calculated as usual. This ensures that the best-fit line will not
go througb the origi.n.
Note that we always use the days 1, 2, 3, 4, 5 as tbe independent x variables for each successive forccast. Wc
forecast the high for day 6 usiog data from days 1 through 5;
for day 7, we use data from days 2 through 7, droppiog the
data from day l. Accordingly, we assume that the market's
" memory" cxtends five days back. We will show bow to extend the calculatioo to cell ES usiog the TRENO function:
forecast high in cell E8
- TREND(b3:b7,,{6},TRUE) (2.5)
N~xt
37
Troding Day
Forecasts for the low and tbc close use exactly the same
calculatioo me thod. lf you do not have a TREND function
in your spreadsheet, you can use the calculation scheme
showo in Table 2. 1. For completeness, we will show bow we
wrote tbe TRENO function for forecasting thc low and the
close. ln effect, you can use 1he COPY function in the spreadsheet 10 copy the formula from cell E7 into the neighboring
columns, cells F7 and G7. Similarly, you can then copy the
formula from these cells into ali subsequent rows to calculate
tbe rcmaioing forecas1s.
forecast low in cell F7 - TREND(c2:c6,,{6},TRUE}
forecast close in ccll G7 - TREND(d2:d6,,{6],TR UE}
(2.6)
103
102L-~~~~~~~~~~~~~~~~~~~--'
$
: ~ ~
~
~ b
X ~ o
~ ~
e
~ ~
~ ~
~ ~
~ ~ 9
~ ~ ;
~ ~ ~ ~ ~ ~ ~
~ ~ ~ ~ 9
~ ~
~ ~ ~
ACURE 2.12 The da1ly pnce range of the Seplember, 1993 Treasury
Bond futures contract and a S-day forecast of the daily close.
38
can see that a close above or below the forecast tipped off the
trend for the next severa! days. For example, a olose above
the forecast on February 11 Jed into a tradeable 8-<iay rally.
The market corrected briefly by going sideways for three days.
1t resumed anothcr smaU four-day swing, lcading to a shortterm top oear 111-16. The next series of swing moves were
also tradeable on a short-term basis.
Figure 2.13 shows the same T-Bond 09/93 contract in the
months of April, May, and June, 1993, indicating the envelopcs formed by the 5-day forecast of the daily bighs and lows.
T he envclopes tend to narrow beforc signiftcant moves. A
daily close outside the envelope a lso provided clues offuture
market directioo. For example, note the close above the higb
forecast in early April, May, and Junc that led to higber prices.
A close below tbe envelope also preceded down moves in
mid-April and mid-May. The envelope oflen acted as a point
of resistaoce and support as the markct searched for direction,
aod couJd be used to plan tradiog strategy.
The T-Bond market trended steadily upward in June 1993.
1t was oot making large swing moves. Figure 2.14 shows the
39
115
cr'
11
113
112
~-ol\
, 11
" J1
'
{'"' i'
j~
t~
f'f r
'
.)
1 ~~
-"'
\
Otitv rano
daily forecast close and price range. Here, too, a close above
the forecast helped gauge market directioo. Nevertheless, the
markct often traded in a narrow raoge in this month; therefore, oot every close below the forecast led to down moves.
115
11 4
01l1v rnoe
113
112
111
110
109
1oe
107
108
=
ACURE 2.13 The daily price range of the Seplember, 1993 Treasury
Bond futures contract and a 5day forecast envelope.
Thc daily forecast is usually a good gucss for the next tradi ng
day, and can be uscd as a tcmplate. When the market is particularly strong or weak, it doses above tlle forecast higb or
below the forecast low. The most important benefit of using
the forecast template is Lhat you will ba ve speciJie oumbers
for planniog actions if lhe market trades at or beyond a particular price.
The types of plans you can develop are varied. One simple
strategy is to buy or sen on the close. Let"s assume that the
el ose is goi ng to be above thc forecast for that day. You could
then buy on the close, expccting higher prices ahead.
Aoother approach is to use the daily forecast higb and low
40
l~(C -
Cr)/C
(2.7)
41
Table 2.3
Date
Close
Forecast
Close
930127
930128
930129
930201
930202
930203
930204
930205
930206
930209
930210
930211
930212
930216
930217
104.3125
104.5
104.7186
104.9375
104.4&68
104.7613
105.3436
105.675
105.5&25
105.4375
104.625
105.3438
105.5
105.375
105.6437
104.81255
104.7750&
105.17818
105.90&28
10&.19061
105.85935
104.60623
104 76879
105 22815
105.46126
Forecast
Oscillator of
Close
- 1
16
22
-12
-25
- 40
17
23
4
11
42
40
30
Summary
43
20
10
o
10
20
30
(2.8)
FIGURE 2.15 A forecast oscillator for the Septem ber. 1993 Treasury
Bond futures conlract.
The linear regression method provides tbe slope and r2, the
coefficient of determination. You can measure trendiness witb
r2 since the slope can be used for following price trends. Tbe
slope can also be used to identify overbougbt or oversold
condit.ions and to develop forecasts for the next trading day
tbat can be used to develop a game plan for your trading.
as:
e; = y - Y;
y - (MO
* X; + CO)
(2.9)
Here, the subscript (.) denotes di.fferent values of the independent variable x. The least squares method minimizes
thc sum of the squares of thc error terms. The error terrn is
thc vertical deviation from the line to the actual data at that
value ofx; the deviation can be positive or negative. By squaring the deviation, we get only positive numbers. The linear
regression finds the particular values of slope and intercept
tbat min.imize the sum of the squared deviations. Note that
tbe first useful piece of information from linear regression is
the slope. Once we know the slope and the intercept, we can
draw or calculate the best-fit line for different values of x.
We will now calculate two deviations. The fi.rst deviation
is the vertical distaoce between the fined line and a given
data point:
44
Uneor Regression Analysis
..;,_~~~~~~~~~~~~~..;;;;
e, - Y - Yr
Here we bave used the subscript r to show the error from the
best-fit line. lfwe squared cach dcviation e, and added them,
we wo uld bave the SSE or Sum ofSquares from fitting errors.
Thus:
SSE - sum (e,)2
(2.11)
Y. - (y - yJ
(2.12)
r2
(2.10)
(y~)2
(2.13)
(2. 14)
Oegrees o
Freedom
Regress1on
model
Mean Square
F-ratio
MSR SSR/ 1
MSR/ MSE
SSR
Error
(residual)
(n -
Total
(n - 1)
Sum of
Squares
2)
SSE
MSE - SSE/
(n -
SST
2)
46
MSR/ MSE
(n - 2) (SSR/ SSE)
(n - 2) (SST - SSE)/ SSE
(n - 2) r, ( 1/ (SSE/ SST))
(2.15)
- (n - 2) (r2/(I - r 2))
47
Summary
(2. 16)
Confidence Level
is
DOF =
n - 2
F-ratio
0.95, l , n - 2
G - F/
(n - 2)
r' Critica! =
G/(1 + G)
10
20
30
50
8
18
28
48
5.32
4.41
4.20
4.04
0.665
0.245
0.150
0.084
0.40
0.20
0.13
0.08
48
3
The Variable lndex
Dynamic Average
SO _ _ __ _ _ _ _ ___T::.:h:.;:e~V~~
ri:.:.:.b~
le~l~
n::.
d::.~D~y~
n~
m
~i~
c~
A~
v~~
g~
e
Close
100
125
110
90
108
11 9
135
110
100
125
140
20%of
Close
10% of
Close
5 day SMA
10-day SMA
20.00
25.00
22.00
18.00
21.60
23 .80
27.00
22.00
20.00
25.00
28.00
10.00
12.50
11.00
9.00
10.80
11 .90
13.50
11 .00
10.00
12.50
14.00
106.60
11 0.40
1 12.40
11 2.40
114.40
117.80
122.00
112.20
116.20
A Volatility lndex
(3. 1)
52
* k * ~ + (1-alpha * k) C,.
53
(3.3)
1)).
(3.2)
Table 3.2
No. Oays in
Vola11fity
(lndex. k)
Alpha (lndex
for 9-day EMA)
k'Alpha (1ndex
for VIDYA)
Average
(effective lenglh)
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
0.20
0.20
0.20
0.20
0.20
0.20
0.20
0.20
0.20
0.20
0.04
0.08
0.12
0.16
0.20
0.24
0.28
0.32
0.36
0.40
49.00
24.00
15.67
11.50
9.00
7.33
6.14
S.25
4.56
4.00
0.20.
54
k alpha vares tenfold from 0.04 to 0.4 due to rising volatilhy, the effective lcngth of VIDYA decreases from 49 to 4
days, a wide dynamic range. An exponential moving average
with a length of 49 days is much slower than one with a Jength
of 4 days. Therefore, VIDYA will move faster than the equivalent exponental moving average i.n response to rising market
volatili ty.
TRADING STRATEGIES
Ali your usual strategies with moving averages can be used
with VLDYA. For example, you can use a shon and a long
VIDYA and use a moving average crossover decision model.
Or, you can take the diffcrence between the two avcrages and
trade that witb a third moving average, whicb is the MACD
or moving average convergence-divergence approach discussed in Chapters 1 and 2. You can use moving averages
and trading bands or use VlDYA to set trailing stops. Since
VIDYA adjusts its length automatically in response to market
volatility, it will generally remain closer to prices than the
equivalen! exponential moving average. Hence, its responsivcness can give you more timely trades, which is the main
advantage ofusing VIDYA.
You can use volatility bands around VIDYA to make trading decisions and crcate a breakout system, trading prices in
thc direction of their brealwut outside the bands. Your next
signal to close the trade or reverse is when prices close inside
the bands for the first time. You also can use bands shifted
above and below VlDYA by a fixed percentage, sucb as 1
percenL Tbe upper and lower bands act as points of suppon
and rcsistance. Wben prices close near tbe uppcr or lower
band, you would look for markei reversals. As usual, when
the market is trending strongly, it can close outside the bands.
This is a signa! that the trend wilJ persist. You wou ld then
wait far the prices to close inside the bands befare taking
countertrend positions.
'1lt 11'.6
"
57
91>0[6
tlt()(6
lll'I~&
LOtU6
SOt0[6
Qt0[6
llf :'<&
9WK6
tZfOf 6
Uf!Y:~
<
o
>-
Q()[()(6
>
..
;orors
Z0.11(6
!!
>
....
.5
: Pll0C6
UZ()(6
a. '6
:j
91lOC6
'll0C6
60ll)(6
<;ol0f6
rozor6
10lOC6
Qll (6
9ZIOC6
Zll0f6
!Vl()(f
...i
j
(,)
>
o"'
..,u
~
~
~6
ti 10[6
ll 10[6
ij\1 ,{6
-llll Jl6
P010[6
Ol:lllt
.8
"'
;;)
56
-- -
"'o
...o
FIGURE 3.2
3.1 .
: 9ZZO6
>
-...
QlfO\ 6
91(0C6
Zlf0\6
O'I '(6
ll3
11 2
1 ll
"
109
108
FIGURE 3.3 The )une, 1993, T -Bond futurc>s contracl, smoothed wilh
a variable lndex dynamic averJge .
S8
bounced orr the lower band, with a close stop for proper risk
control. During the upward move in ApriJ, 1993, prces ctosed
outside the upper 1 percen1 band, showing a strong upward
move. As expected, prces continued tbeir climb for four more
days before moving sideways.
OlHlr6
91Hlr6
rl70f6
- Ztt0f6
LOr0(6
~1'0(6
l01(){6
(){f(){6
9lrof6
lf(){6
Zlf0f6
g1f(){6
9lrof6
e?
lf(){6
01r01:6
!IOfOf6
KK'Of6
,,..
-g
-..
9Zl0f6
~ZZOf6
lll()(6
Qll(){6
9ll(){6
1ll0f6
6Ql(){6
.9
:f
g)l(){6
i..
lndexi ng to Momentum or r2
Ano1her useful feature of VIDYA is that it can be indcxed to
any dimcnsionless market variable that vares from O to l.
Thus, you don't need the standard deviation to calculate
VIDYA. For example, you could index VIDYA to thc coefficient of determination, r', since r2 measures trendiness, is
dimcnsionless, and vares between O and l. You could also
indcx VIDYA to a momentum oscillator that varies bctween
O and 1, such as the rclative strength index (RSI). We like
using the absolute value of the 9-<!ay Chande momentum
oscillator (CMO) as the momentum index. The CMO, a variant of the RSI, is described in Chapter 5.
l0f()(6
1>
fOZOf6
10l0r6
9l1(){6
9ZUJl"6
lll0f6
Oll0f6
91l0f6
711(){6
ll l0f6
90l0f6
901(){6
r010f6
orz1z&
-3
..
.e
~
<
>o
l.!
>
...~
l
"-
-N
-o
.,o
o
S9
60
Figure 3.5 shows the June 1993 T-Bond contract in February-Marcb, 1993, as it becarne the actively traded coo tract.
The VJOYA, based on absolute CMO, was seositive to market
action during rapid moves. Simultaneously, it flattened out
during sideways periods.
In Figure 3.6, one VlDYA is calculated using standard deviation and another uses a 9-day absolute CMO. Tbe scaling
multiplier for tbe CM O was 0.50. T his meaos tbe 9-day CMO
varied between Oand 0.50 for these calculations. (Tbe details
of thc calculations are in tbe tutora!.) VIDYA based on a
scaled, absolute CMO was more sensitive for the Juoe, 1993
Treasury Bond contract. lt accelerated more quickly than the
VlDYA based on standard dcviation. Note how the CMObased VIDYA flattened out in December, 1992 and accelerated in January, 1993. The two metbods ofcalculation change
tbe way VJDYA responds. Note that you can also cbange Lhe
sensitivity by altering the scaling variables.
In Figure 3.7, we look at the same contract in tbe same
time period, but calcula te VlDYA using r2. Here too, we scaled
Lhe r2 with a multiplier of 0.50 so that its values range from
61
..---------------~-=
~~
108
Tl!o<Ml~doN
105
104
103
/'
102
l
101
l\2
111
110
109
112
101
110
ot 9-<NV hlll t~ of ~
,..,....., o.s r - " '
1oa t:~=~~~___,~~-=-=~::-;:~;-;;;:,
~
~~O-N~~-~~sa~~~~a~
109 .
~~!!~~!~~~~~~~~~~~~~~~~~.~~~
108
107
108
111
':"s.
mde:ed to
ng 0 5 times the coefficient of determ1nauon (r ) from
.
mark et act1on us1
62
112
SUMMARY
contttct
"1
110
109
108
107
.,....,,.._
Tutorial: SprNdshee!!
ts~fo::,:r_V:..:1:::
0.::
YA.:.__ _ _ _ _ _ _ __ _ _ _6
_3
__ .. _
- - 0 6 6 ..-
FIGURf 3.8 A VIOYA for the June 1993 T-Bond da1ly dose similar
to the one in Figure 3.7. W e used 0.65 times 1he r' value from lhe 9.
day linear regression to calculate VI OYA.
O to 0.50. You can compare Figures 3.6 and 3.7 to see how
changing the indexing variable changed lhe response of VJ.
DYA. Whcn the linear rcgression analysis showed a strong
trend, lhe VIDYA based on r2 accelcrated rapidly. H moved
more quickly Iban lhe VLDYA based on the absolute CMO
for example, co mpare thc lwo types of V!DYA in m id-to-lat~
February.
Not only can you changc lhe iodcxing variable, bul you
can also change the scali og variable. In Figure 3.8 you can
see the effect of using a new multiplicr of 0.65 for r2. The
effect of tbe cha ngo in multiplier from 0.50 to 0.65 did not
malee much difference when thc market was trendlcss. Once
lhe linear regression detectcd a trend on 02/ 19/93, lhe VJDYA
with the higher multipUer moved more quickly. Notice how
~e VIDYA llattcncd out as r2 decreascd during consolida
uons. You can use a VIOYA bascd on r2 to set aggressive
trailing stops that are unique and not easily gunned
=
=
0. 13/0.64 - 0.21
l.47/0.64 - 2.29
(3.4)
Alpha
k Alpha
0 .21
0.1
0.1
0.1
.021
0.1
0.229
2.29
No. of Days
(roundcd)
94
19
65
Date
Close
921127
921130
921201
921202
921203
921204
921207
921208
921209
921210
921211
921214
9212 15
921216
921217
102.00
101.97
102. 19
102.19
102.31
103.16
103.97
104.19
104.19
104.25
104.13
103.88
103.91
104.09
104. 13
10-day
StdDev
0.96
0.94
0.65
0.74
0.58
0.30
k-c/0.64
1.50
1.47
1.33
1.16
0.91
0.47
VIDYA
104.19
104.20
104.19
104.15
104.12
104.12
104.12
EXPMA
104.19
104.19
104.19
104.16
104.13
104.13
104.13
(3.5)
66
Close
930201
930202
930203
930204
930205
930208
930209
930210
930211
930212
930216
9302 17
930218
9302 19
930222
930223
930224
930225
930226
107.41
106.94
107.25
107.81
108.34
108.03
107.91
107.09
107.81
107.97
107.84
108.31
109.34
109.56
110.3 1
111.72
111 .09
110.97
110.88
10-day
StdDev
c/0.64
'k'
VIOYA
EXPMA
0.59
0.55
0.46
0.44
0.48
0.53
0.93
0.86
0.72
0.69
0.75
0.83
0.79
0.73
0.69
0.68
0.67
0.59
0.85
1.09
1.47
2.11
2.30
2.19
2.08
l 06.1 s
1OG.22
106.29
106.40
106.54
106.67
106 76
106.79
106.86
106.93
107.00
107.07
107.27
107.52
107.93
108.73
109.27
109.64
109.90
106.02
106.11
106.23
106.38
106.58
106.73
106.84
106.87
106.96
107.06
107.14
107.26
107.47
107.68
107.94
108.32
108.60
108.83
109.04
o.so
0.47
0.44
0.43
0.43
0.38
0.54
0.70
0.94
1.35
1.47
1.40
1.33
67
Date
921127
92 1130
921201
921202
921203
921204
921207
921208
921209
921210
921211
921214
921215
921216
921217
921218
921221
0.96
0.94
0.85
0.74
0.58
0.30
0.13
0.26
1.50
1.47
1.33
1.1
0.91
0.4 7
0.21
0.41
104. 19
104.20
104.19
104.15
104.12
104.12
104. 12
104.12
104.15
104.SO
104.48
104.41
104.35
104.30
104.21
104. 16
104.23
103.90
103.89
103.88
103.88
103.93
104.02
104.08
104.07
68
VIDYA + 2 * 0.1 * k
104.20 + 2*0.1*1.50 - 104.50,
VIDYA
2 0. 1 * k
104.20 - 2 .. 0.1 1.50 - 103.90.
(3.6)
CMO - (S.
S.)/(S. + S.),
absCMO - <:MOj.
(3.7)
(3.8a)
69
s. s. (S. - S.) -
+ S.)
CMO
absCMO
VlDYA,
k
VIDYA
VTDYA
(S0
70
Date
Close
930201
930202
930203
930204
930205
930208
930209
930210
930211
930212
930216
930217
930218
930219
930222
930223
930224
930225
930226
106.1563
105.6875
106
106.5625
107.0938
106.7813
106.6563
105.8437
106.5625
106.7188
106.5938
107.0625
108.0625
108.28 12
109.0625
110.4686
109.8437
109.7188
109.625
9-day Abs
CMO
71
Mtm Up
Mt m Dn
0.4688
0.3125
0.5625
0.5313
o
o
o
INDEX(LINEST(b3:b l l ,$c$2:$c$10,TRUE,TRUE),3)
(3.9a)
o
o
0.3125
0.125
0.8 126
r' -
0.7188
0.1563
o
o
0.4687
1
0.2 187
0.7813
1.4063
o
o
o
o
o
0.6251
0.1249
0.0938
o
o
VIDYA
0.125
0.1 40614
0.24786
0.278674
0.352925
0.301554
0.517702
0.670306
0.727246
0.643316
0.599983
106.5625
106.5735
106.576
106.6438
106.6941
107. 1033
107.6104
108.5684
109.0321
109.253
109.3646
(3.9b)
Close
930201
930202
930203
930204
930205
930208
930209
9302 10
930211
930212
930216
930217
930218
106.1563
105.6875
106.0
106.5625
107.0938
106.7813
106.6563
105.8437
106.5625
106.7188
106.5938
107.0625
108.0625
R-Squared
VIDYA
0.65
VIDYA
106.5625
106.5756
106.576
106.5767
106.7593
106.5625
106.5726
106.573
106.5735
106.7142
o.so
1
2
3
4
5
5
7
8
9
0.1292494
0.0336062
0.0022159
0.1889946
72
4
QSTICK: The
Quantitative Candlestick
74
lcts you combine candlcsticks with trend lncs, moving averages, trading bands and so fonb.
lo candlestick analysis, each day's pricc action appears lk:e
a candle. The body of thc candle is drawn as a rectangular
range betwcen the open and close, and it is black (or filled)
if the close is below the open. If the close is above the open,
then the body is white (or cmpcy). "Down days are dark" is
an easy way to remember this scheme. Thin Lincsjoining the
high and low are called the upper and lower shadows, or tails.
Figure 4. 1 shows a bar chan and the equivalen! candlestick
chan. The data are for thc July, 1993 Couon #2 futures contract. While the two chans use the same data, tbe candlestick
chan is more striking in revealing bullish aod bearish days.
Most analysts look at 1wo imponant factors in cvaluating
candlesticks. The first clcment is the appearance of thc new
75
00
00
00
We have noted that thc range between the open and the close
is the most imponant e lcment of candlestick analysis. Tbe
diffcrence close minus open, is a mcasurc of intrad ay (or
intraperio'd) momcntum. We have quantified this rangc to
devclop a trend-following indicator. This indicator, called
Qstick, can be dcfincd as the moving average of the intraday
momentum.
We will use eight days of data for calculating the simple
moving average. You can experiment witb o tber time periods,
from minutes to montbs.
00
m
Jun
Jul
FIGU RE 4.1 A bar chJrt and candleslick churt for the Cotton #2
futures conlract using the same mformation for ead1 day
76
77
Figure 4.2 shows Qslick data for the Standard & Poor's-500
Index during the three montbs prior to tbe crash of 1987. We
llOOOO
m .ooo
310000
300000
01t1tk -026 018 000
100
100
FIGURE 4.2 An 8-day Qstock plotted under the S&P-500 index in the
period pnor to the October, 1987 crash.
78
3'11.cm
335 cm
330.cm
325.cm
m cm
316.0::0
310 cm
co
305 cm
300 cm
Ostlck 1 51 -O 3l OCO
1.50
CO
-OCO
50
FIGURE 4.3 A 16-day Qslick and 1ts 8-day simple moving average,
shown with the Standard & POO(sSOO index m 1987.
.. oo
is a1 the hcart of candlcstick analysis and quantificd this difference using a sim ple moving average to get Qstick. Further,
~t ~ould be useful to have an oscillator 10 pinpoint extremes
to mtraday momeotum. We could develop such an oscillator
the ~traday momentum index {IM!), by adapting the com~
putauonal approach used in relative strength index (RSI) calculat10ns.
Follow!ng the ~SI approach, we separa te thc intraday momentum 1.nto bu llish days and bearish days, gctting positive
numbers in .both columns. Wc can then define an intraday
momeotum mdex (IMI) over, say, 14 days using the cquation
below:
ISu = Sum of Intraday Momentum UP (C > 0),
!So ~ Sum of ln1raday Momcntum Down (C < O),
lMI _
!Su
(4.2)
81
:::
90.00
8000
0.625
0 .620 .
Dailydose
70.00
6000
0 .615
::::
l
0.600
- 5000
::.:1
o595
0.590
0585
1--~-- -~--~
11 /> 3~2
,.,,
11'25192
12; 0Ml2
12111192
11
2000
'00
_ __,_ 000
12'31192 01 13i93
01'25~
FIGURE 4.5 The interna! momenium index for deutsche mark data
8
_ 2_
_ _ _ _ _ _ _ _ _...::>.
Q.STICK: The Quantitative Candlestick
83
higb
higb
open
close
open),
open),
open),
open).
(4.3)
85
CXll
CXll
(XI)
ollJ (XI)
350
251'.l
150
50
, Sil
Jun
JUI
l4JJ
St
Oct
"""
Dtc
92
F"tb
Mar
Juo
86
93, abead ofthe actual peak in prices two weeks later. Qstick
dropped below its own moving average tbe wcek of 1/ 17/ 92,
wben Amgen closed at 70. Thus, tbere was ad vanee notice of
a significant top in the sbarp peak in Qstick and its decisive
break below its own moving average. The sell signa! was in
force for three weeks before Amgen began its slide into ApriJ,
1993.
Figure 4.9 shows weekJy Amgen data from July, 1992
through Jul y, 1993. Note again the sharp peak in Qstick in
thc weck of 11/ 13/ 92, before tbe stock peaked at 78 the week
of 12/~~93. This w~ a double top in Amgen. Tbe Qstick
was dec1s1vely below lis own movingaverage in January 1993
before Amgen prices fell bclow the 10-wcek average
collapsed. Qstick bottomed witb prices m Marcb and accelerated above its moving average close to the act~al bonom in
and
go..w eeklr 0111sm C-35 lf,O 500 0-35 500 HooJ6 7SJ L"35 DDO
Analyzlng Thrce
Slocks_::
w:_::ll:h...::
:.: Q~s::.:
lic::.:k:..__ _ _ _ _ __
_ _ _-"-87
v.11sn
cm
otQ+
+o
DDO
oco
'll.DDO
O<i<k 000 -036 000
"l<k o35-12
o Jl'j oIXXJ)
200
050XI
O.DO
O.Dl'.l
2DD
.,..
Od
...
IODO
00
o.e
93
M A M J J A S 0"081!F M AM
FIGURE 4.10 A muh1year weekly candles11ck chart for General Electric along w11 h a long-term Qstick and its moving average.
88
"' k 0360 16 0 00
olll
...........
060
Summuy
89
Finally, Figure 4.12 is example o f intennediate tenn analysis usi ng a 20-day Qs ck aod its 20-day moving average. We
used Philip Morris (MO) daily data from its top in September,
1992 through May, 1993. Tbe Qstick fcll below its average in
mid-September, signalling tbe coming top. The 20-day moving a verage turned negati ve in mid-October, suggesting lower
prices to come. The 20-day a verage moved up to tbe zero line
following tbe rally in mid-December. It stayed in oegative
territory ali tbe way to late May, 1993. The Qsck also did
oot 1urn positive from the secood week of December, 1992
through the first week of May, 1993. The Qstick indicator
called the top in MO and stayed oo the short side of tbe
market throughout tbe decline. Using Qst.ick, you could have
avoided a big decline in Philip Morris, o r mo re tban doubled
your mo oey with a shon postioo.
lll
J
S ON
D91 F
MAMJ
SO N
0 92
MA
fl~URE 4.11 A weekly candlestic k chart for Gene ra l Elec tric a long
w11h
Ph1lMofflOMy 07113193 <>-49125 31!> 0-9 f'm H=S0.500 L"l 750 """29-'92
-~"'1--
a Jong-term Qstick.
90
SUMMARY
The purpose of quanti fy ing candlesticks is to improve identification of candlestick patJerns. We quantied the patterns
using tbe candle body, via Qstick (a movi ng average) and the
interna! momentum indcx (an oscillator). We also quantified
tbe sbadows by sepa rating upper and lowcr shadows. Tbcsc
indicators can be uscd by themsclves, or to clarify interpretation of candlestick pattcms. We showed bow thcsc ideas can
be used for shon, intermcdiate, and long-tenn anal ysis of
stocks or futures.
TUTORIAL ON QSTICK
Calculating Qstick
TutoriI on Qstick
Day
1
2
3
4
5
6
7
6
9
10
11
12
91
Open
(0 )
H1gh
6063
6t5 1
6152
6185
6155
6241
6295
6287
6256
6249
6166
607t
6162
6176
6191
6207
6053
6322
6338
6290
6314
6281
6t73
6t05
low
Close
(C)
Daily
(C-0)
Qstic k
5-day
Average
6057
6t29
6152
6132
6t51
6235
6274
6251
625t
6156
6066
6038
6153
6163
6185
6149
6241
6300
6288
6260
6266
6166
6071
6055
90
t2
33
-36
86
59
-7
-2 7
10
81
-95
-1 6
37.00
30.80
27.00
15.00
24.20
-9.20
- 40.00
- 4 1.80
Table 4.2 Sample Calculations for IMI Using Deutsche Mark Data
Open Close
(C)
(0)
6063
6t51
6t52
6185
6 155
6241
6295
6287
6256
6249
6166
6071
6153
6 163
6t85
6149
6241
6300
6288
6260
6266
6168
6071
6055
5-Day
Down Sum of
Up Oay Day Up Day
(C - 0) (C - 0)
(E)
90
t2
33
86
59
o
o
tO
o
o
o
5-0ay 5um of
Sum of Up&
Down Down
Oay (r) Day (G)
IMIE/G
100
o
o
o
36
o
o
7
27
81
95
16
221
190
178
145
t55
69
10
10
36
36
43
70
34
11 5
210
2t9
257
226
221
215
t89
184
220
229
85.99
64.07
80.54
67.44
82.01
37 50
4.55
4.37
92
QSTICK: The Quantitative Candlestick
Calculating IMI
Let's look at 1MJ cal ti
5
New Momentum
Oscillators
94
9S
S4 )/(S.
+ S.)
(5. 1)
101.0313
101.0313
101. 125
101.9687
102.7813
103
102.9687
103.0625
102.9375
102.7188
102.75
102.9063
102.9687
o
0.0937
0.843 7
0.8126
0.2187
o
0.0938
o
o
0.0312
0. 1563
0.0624
o
o
o
o
Up-Mtm Dn-Mtm
10-Day 10-Day
Total
Total
0.125
0.2187
0.375
0.375
0.3 75
69.62
71.43
71.08
can ~hen sm_ooth the CMO values using a simple or exponcnual moving average.
A be~er way to visuaJize the CMO is to rewrite it in terms
of rela~ve and absolute momcntum. Relati ve momcntum is
thfc dail~ s1gned momcntum, and is given by the difl'erence
o today s close and yestcrday's close:
CMO - 100- (Momentum)/ (jMomentump.
o
o
2.0937
2.25
2.2187
97
CMO
0.0313
o
o
o
{5.3)
Lct's illustrate the impact of the smoothing scheme on indicator values by showing 14-day RSI calculations, in Table
5.2, using daily closing data for thc S&P-500 indcx. Figure
5.1 also shows the daily close for the S&P-500 index and uses
a pon ion of these data in Table 5.2. Note the panicular pattem of closing prices for the index, espccially the sequcnce
of closes and their positions relative to one another, beca use
we'U refer to them later.
Table 5.2 shows the up-day and down-day momeotum. The
up-day momen tum is Oiftoday's close is less than yesterday' s;
otherwise, it is the absolutc difference belween thc two closes.
The down-day momentum is zero if today's el ose is greater
than yesterday's close; otherwisc, it is the absolute difference
between the two closes.
Thc oext two col umns ha ve the 14-day sum of the up-day
(SJ and down-day {S.) momentum. Figure 5.2 shows the unsmoothed 14-day sum of up-day and down-day momentum.
Notice that these lines also show price extremes. The two
lincs cross one another when tbe net momentum over 14 days
. When thc 14-day net momentum is postivc, the line representing the up-day sum {S.) is above thc down-day sum
{Sd). Thc reverse is true when the 14-day net momentum is
negatve. The CMO shows precisely this relationship, sincc
the numerator is the differcnce between S. - Sd.
The RSI calculations, described by J. Welles Wilder, J r. in
his book New Concepts in Technical Trading Systems, smooth
the s. and Sd data from day 2 forward. This smoothing adds
1/ 14 of the ncw vaJue to 13/ 14tbs of the previous value of
these variables, wbich amounts to roughly a 27-day exponential moviag average.
98
Ow>d~ Mo~tum
Otdlltor
Date
01/04/93
01 / 0S/ 93
Ot/06/93
0 1/07/93
01/08/93
01/11/93
01/12/93
01/13/93
01/14/93
01/15/93
01/18/93
01/19/93
0 1/20/93
01/21/93
01/22/93
0 1/25/93
01/26/93
01/27/93
0 1/28/93
01/29/93
02/01/93
02/02/93
02/03/93
02/04/93
02/05/93
UpS&P-500 Day
Close
Mtm
435.380
434 340
434.520
430.730
429.050
430.950
431.040
433 .030
435.940
437.150
4 36.840
435.130
433.370
43S.490
4 36.1 10
440.010
439.950
438.110
438.660
438.780
442.520
442.5SO
447.200
449.560
448.930
0 .00
0.18
0.00
0.00
1.90
0.09
1.99
2.91
1.21
0.00
0.00
0.00
2.12
0 .62
3.90
O.DO
o.oo
O.SS
0 . 12
3.74
0.03
4.65
2.36
0.00
DnDay
Mtm
14-Day
Up Down RSI
Sum
Sum Smth
RSI
UnSm
460-000 - - - - - - - - - - - - -455.000
S0.000
S.000
- - - - -- ,
0.000
435.000 ,.__ _
430.000
1.04
0.00
3.79
1.68
0.00
0.00
0.00
25.000
20.000
415.000
L..::==-------:-:--::-:::-:::-.:--::"'.::-:;-::;--;:;-::;-;:~
~ ~ iE. ~ ~ ~ ~ ~ ~ ~;;? p' ~ ~ ~ 7' ~ ~ ~ ~ !:: ~ 5 ~ ~ ~ ~
-~~,..I ~~~r
;::,.
~~-.._.se:
~~~
bo,o:"""
~~~V':; 8!:::<~
,-..~-S?.-..
~~~~~~~~a~~GGS
OC'
o.oo
0.00
0.31
1.71
1.76
0.00
0.00 0.79
0.00
1.01
0.06 0.94
1.84
0.87
0.00 0.85
0.00 0 .80
0.00
1.01
0.00 0.94
0.00
1.20
0.00
1.28
0.63
1. 19
s.
s.
3000
25.00
20.00
0 .74
0 .68
0.64
0.72
0.67
0.62
0.58
0.54
o.so
0.46
0 .48
s.
s.
15.00
51.71
59.66
S9.SO
54.60
55.77
56.04
63.44
63.SO
70.62
73.46
71 .48
51.71
61.73
6 1. 29
66.70
72.91
70.40
75.13
72.80
74.89
76.10
75.09
10.00
6.00
o.oo L-------~-:--_..:,~::-=:_::-.;:_"""..=-::_-:
~::-::~:---:-;:;"'.~--:::-::;:;-;:;:-;::i~~-::
r;;t~o' ..... .e... ""Q'>CI'~
f7'~~
. ;:.': .
~-~~ss~~>~~-.a~~-i~~
~~
~o""'
- - .t:::.,...._ ~......._......__,
--~I
.J;::::.;:=;.
:;;
S~ci
e.a
o ...
o ,.,..;;-~,..._.,
.......o ~ ""
....
_..
......
<"11
.....
~~s-~
,
-. i!:.~g"" i.s
FIGURE s.2 This shows plots of the 14-day sum of upday _mo'.11en tum and down-day rnomentum for the S&P-SOO index data in Figure
5.1.
s. s. -
(5.4)
~
59.56.
100
The 01/25/93 value of RSI in Table 5.2 is 59.66. Thc difference between our sample RSI value of 59.56 and tbe tabulated
val~e is caused by the rounding of tbe numbers.
Figure 5.3 shows the 14-day smootbcd RSI and thc 14-day
unsmoothcd RSI (RSI'). Notice how tbe RS I looks similar to
the price changes (price partero) of the S&P-500 index. The
RSI' dfers markedly from the RSI. It looks more likc mome~turn, as you will see below. Note how the RSI' was above
80 tn Mar~h! even though th e RSI barely reached 70. Your
tradmgdec1s1oos would be quitedilferent in the two instances
w~ch highligbts the importance of smootbing in indicato;
des1gn.
Figure 5.4 shows th e 14-day unsmootbed CMO and the 14day RSI. Note that there are significant dilferences in tbe
appearance ofthe two curves, even though CMO and the RSI'
are re~at~d, _sin~e CMO = 2 * RSI' - 100. You can clearl y
~ee this s1m1lan_ty by comparing thc appcarnnce of thc CMO
m Figure 5.4 with the RSI' in Figure 5.3.
. Figure 5.5 plots the 14-day momentum of tbe S&P-500
mdcx over the same time period as in Figures 5.3 and 5.4.
''
6000
40.00
<RI
~)
o 00 '>--"-----~:.. ;.---.;;,J"T---.,p:i....:.+-----1
~~ p'\p ~ \:t~r
~ooo '-----------------;:;.........:._ ___.
20.00
14 doy CMO
t!'tf~
ACURE S.4
day CMO.
25.000 - - - - - - - - - . - - - - - - - - - - - - - - ,
20.000
15 .000
10 000
5000
0.000
-5.000
10.000
1oooo
r-------------------~
9000
8000
70 00
00 00
50.00
4000
h~~
. \:
-.~--
....
..\.
...
f\..;
14day RSI
FIGURE S.S
Figure 5.1.
3000
2000
14--dav 1moothed RSl
10.00
000 , __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ j
Osdllator>
102
New
Mom~nlum
s. =
(5.5)
"1
~~~~~~~~~~~~~~~~~~~~~-~~~~~
~~~$~~s~~i~~~s~ss~ss~~~ss,~s
-~R~~~~~~
~~~~~-~C~~~N~*-
OO$SSSSSs
ga~aa~~~~~~~sa
FIGU RE 5.6 A chart of RSI values with the usual 27-day exponential
smoothing and a short 9day exponential smoothing.
01/04/93
01/05/93
01/06/93
01 /07/93
01/06/93
01/11/93
01/12/93
01/13/93
01/14/93
01/15/93
01/18/93
01/ 19/93
01/20/93
01/21/93
01/22/93
0 1/ 25/ 93
01/26/93
Ot/27/93
01/28/93
01 / 29/93
02/01/93
02/02/93
435.360
434.340
434.520
430. 730
429.050
430.950
431.040
433.030
435.940
437. 1so
436.840
435.130
433 .370
435.490
436.11 0
440.0 10
439.950
438110
438.660
438.780
442.520
442.SSO
0.00
0.18
0.00
0.00
1.90
0.09
1.99
2.91
1.21
0.00
0.00
0.00
2.12
0.62
3.90
0.00
0.00
O.SS
0.12
3.74
0.03
1.04
0.00
3.79
1.68
0.00
0.00
0.00
0.00
0.00
0.31
1.71
1.76
0.00
0.00
0.00
0.06
1.84
0.00
0.00
0.00
0.00
0.79
1.41
1.13
0.90
0.83
0.69
1.30
1.05
0.74
0.59
0.48
0.75
0.60
0.48
0.39
0.31
51.71
70.57
70.04
54.48
57.97
58.83
77.10
77.20
104
100
080
14-day CM O
080
O.O
0 . 20
o 00 'f...~~---+--
020
.()40
.() 80
.() 80
9-dayCMO
..
Chande Momentum
Osc~
il~
la::;
IO
~
r
60 000
1 00
080
0 .80
0 .40
r.
"'
0.20
o.oo
-0.20
-0.40
.() 60
.() 80
~r
100 ~----------------~'-----"'
55000
450 000
5.000
:!
:~= ~~
30.000
25 000
420000
FIGURE 5.8 The values of the 9-day CMO supenmposed upon the
S&P-500 daily close.
VHF - R/P.
...
_ __ _ _ _ _ _ __ _ __ 105
_
(5.6)
In calculatiog CMO and RSI, we separatcd the daily closetCH:lose changes into those for up days and down days. We
can take the sum of thosc changes over x. days to get (S. +
s.). You will recogniie that P. is essentially cqual to (S. +
Sd).
.
At key turoiog points, the close is not far from lhe h1ghe~t
h.igh or lowest low of the period. Thus., to a good apprmumation, we can replace the range, R, w1th the absolute dif-
106
value ofVHF. The price range can widen without rising momentum, because of small progress on a closing basis. The
VHF will show trendioess when absCMO does not. Thus, we
can say absCMO is a measure oftrendiness on a closing basis.
Tbe fact that absolute momentum shows trendiness is reinforced in Figure 5. 1O, which used weekJy data for GE. We
bave plotted Lhe absolute 10-week momentum and its 5-week
simple moving average. Note how the absoJute momentum
increases during a rapid move in priccs and falls when prices
enter a trading raoge. Also ootice how momeotum was flat
ali through the trading range in 1992. The momeotum peaks
at differeot values each time because its scale is not fixed.
In shon, absolute momeotum, and therefore absCMO, can
be used to measure treodiness in data just like VHF. Wheo
Lhcse indicators show that a trend exjsts, you can pul on trendfollowing positions with greater coofideoce of success.
.IXll
IXll
V A BS M T M 963 8 80
3 00
01122193 0 2'5193 02122193 03~3 03122193 04J05193 04120193 05'04193
ACURE S.9 A companson of the 14-day VHF and the 14-day absolute CMO for the S&P-500 index d ata.
108
109
The beauty of the CMO approach is that it converts momcnrum into an oscillator that you can plot on a fixed scale.
By using absCMO to plot the absolutc momentum on a
bounded scalc from O to 100, you can get a consistent scale
to compare different rnarke1 moves.
You can also use CMO 10 define overbough1 and oversold
levels a1 + 50 and - 50 where the ratio of s. to s. is 3: 1 or
1:3. By simple substitu1ion you can find that a ratio of 3: 1
corresponds to a value of +50. This means tha1 the up-day
momen1um is three times the down-day momentum, a rare
occurrence. The reverse is also true; ifthe up-day mornentum
is only one lhird of the down-day momentum, then this oscillator has a value of - 50.
Ano1her convenienl feature of CMO is thal we get positive
and negative values. Hence, wc can tell if the net market
action is falling or rising. Figure 5.11 shows a 14-day RSI and
14-day CMO for a direcl comparison using tbe July, 1993
Wheat contract. The reference levels for CMO (upper iodicator) are +60 and - 60, and they correspond 10 RSI values
of 80 and 20. Note how 1he CMO stayed in ncgative territory
during the downtrend in prices from mid-January througb
March. The divcrgence between the CMO and prices as tbe
markct made new lows in June was a harbinger of the rally
in July. The RSI did no1 show any significant market extremes
during this pcriod, nor did it show a significant divergence
witb thc price as the contract made new lows.
In addition to RSI, we can compare CMO to other iodicators and find new uses for it. For iostance, CMO resembles
both the true strength index {TSI) and the average directional
indcx (ADX). But, as you'U see, gjvcs us additional information. The TSI, described by William Blau, uses doubly
smoothed exponential averages for the relative and absolute
momentum. Hence, its vaJues diffcr significantly from tbe
CMO and will show grea1er sensitivity to price changes. On
the other hand, because of the exponcntial smoothing, the
3111'6
..,_7'>'.,,..J"'...____
'V'/"
93
F..
....
0-, .
..
~~
00
:IJ.00
...
the
TSI rarely approachcs extremes of - 100 and + 100. Conscqueotly, thc CMO can show tbe extremes more readily. Thc
choice of CMO or TSI will depcnd on your trading borizon
and your prcfcrcnce for indicator sensitivity.
In Figure 5.12, you can see 1hat thc average absolute value
of the smoothcd CMO closel y tracks the ADX of the directional movement system. The ADX is the thin line, and the
averaged absolute CMO is th c thick line. We used the Septernber, 1993 US Bond contract to show tbe similarity between the 14-period ADX and the 14-period averaged absolute CMO. These indicators rose when the market was
trending, and declined whcn the market was going sideways.
Both indicators measure treodiness, but the absCMO was
more responsive than the ADX io this example.
The ADX is a popular measure of trendiness io data. Many
110
..
... ...
~-----------------..,.-----,
112
"
108
......
.A
106
....
.. ....
FIGURE S.12 The absolute CM O and the 14-period ADX for lhe
September, 1993 T-Bond contract.
We calculated the CMO as usual, using 5, 10, and 20 daytime intervals. We cbose these lengths arbitrarily, using popular round numbcrs. You can see ali three CMO lines in Figure 5.14. The 5-day CMO is the most volatile, aod easily
reaches values of + 1 and - t. Ths happens when there are
five up days or down days in a row. The 20-day CMO is the
smoothest curve, showing the least volatility. The 10-day
CMO has a volatility bctween tbat ofthe 5-and 20-day CMO.
Different time pcriods are better suited to meet lhe needs
of certain types of traders. The 5-day CMO and its volatility
may suit a short-tcrm trader, while tbe JO- and 20-day CMO
volatility may suit an intermediate-term trader. The frequeney of sgn changes decreases as the time period increases,
which just meaos that there is more smootbing in the data
as tbe calculation period increases.
You can achieve a more seositive alternative to the 20-day
CMO by calculatiog the average CMOA. This is simply an
aritbmetic average ofthe tbree values and is plotted in Figure
5. 15. Tbe 5-, 1O- and 20-day CMO values are designated as
CMO,, CM0,0 and CMO As a result
112
CMOA - ((CMO,
o.e
08
O
02
-o 2
-O
-o 6
-o.
5-dovCMO
..
..
-'
10..W, CMO
"!'
...
20-dovCMO
::'!!
FIGURE 5. 14 Chande momentum oscillator shown using three different tome honzons for the T Bond data in Figure 5.13: 5 days, 10
days and 20 days.
+ CM0,0 + CM020)/3).
(5.7)
(5.8)
0.8
0.6
O.
0.2
o.a
0.8
O.
-0,2
02
-O.
-0.6
-0.2
ti
:;;:
:;
~
-O.
-o 8
'
\
Avtrt09 CM O
-08
5.14.
A GU RE 5.16 A companson of the average composite CMO shown
in Figure 5 1S wil h a vola11h1y-we1gh1ed composite CMO.
114
115
To gcnerate trading signals, you can use the CMO with its
simple ~ovi.ng average (SMA). (Figure 5.17 shows the 10 day
~. with l!S 10 day SMA.) This approach allows you to
IDJttate a trade before a zero crossing of the CMO. Be aware,
however, that it sometimes gives early signals. You could buy
when the CMO crossed above its average, and sell when it
feU below its moving average. Sucb an approach would have
a given a good cntry pcint in January, 1993 at the start of
the rally. Howevcr, the shon signal in early February was
premature; thc market went sideways and lost sorne upward
momentum. The loss ofmomentum pushed the CMO beJow
Ch if C0
- o
Mt m(up,f) - C0 -Ch
- 0
C, (unftcred),
(s.s>
116
08
CMO
06
04
02
i!\
.() 2
-O.
-O&
-08
I ' - - - - - - - -- - - - -- - - - - - -....--~
through 5.17. We calculated the filler using 100 days data and
the average plus one standard devfation of the daily momentum changes. The filter for up-day momentum was 20/
32 and 14/32 for down-day momentum. This means that, if
the daily change on up days was less than or equal to 20/32,
theo t.hat day's up momentum was set to O. Similarly, if the
down move was less thao or equal to 14/32, the down-day
momentum was set to O. The daily momentum equation is
as follows:
up-day mtm - f( (C-o- C,)
down-day mtm - if( (C, - C.,,)
Close
101.03
101.03
101. 13
101.97
102.711
103.00
102.97
103.06
102.94
102.72
102.75
102.91
102.97
10313
103.72
104.22
Up
Mtm
On
Mtm
o
o
o
o
0.0937
0 .8437
0 .8 126
0.2 187
0.0938
o
o
o
0 .1563
0 .0624
o.1563
0 .5938
0 .5
0.0313
0. 125
0 .2187
o
o
o
o
o
o
Filtered
Up Mtm
(5/32)
Filte red
Dn Mlm
(5/32)
Filtered
10-day
CMO
Regular
10-day
CMO
o
o
o
o
o
o
o
o
o
o
0.2187
o
o
o
o
o
o
0.79
0.81
0.81
0.72
0.67
0 .73
0.69
0.71
0 .71
0 .60
0 .55
0.61
0.8437
0.8 126
0.2187
o
o
o
o
o
0 .1563
o
o 1563
0 .5938
0 .5
118
119
~~~~~~~~~~~~~
rr.
CMO Summary
As you havc scen, the CMO is a flexible and versatile momentum oscillator tbat you can casily adapt to a variety of
trading stylcs. ln sum, you can:
Use C MO to identify overbought or ovcrsold conditions.
Use CMO as a trendiness indicator.
Combine CMO with a moving average to generate trading signals.
Use CMO with a different amount ofsmoothing for trading.
Combine CMO values calculated for differcnt time periods into a composite CMO.
Fil ter the CMO to ignore small changcs in market values.
(5.10)
(5.11)
120
We can conrinue this exercise in a modified format by separating 1.be daily momentum into values for days wben tbe
market closes up or closes down. In eacb instance, tbe momen1um values wiJI be posive. On an up day, tbe momentum is 1oday's close mfous yes1erday' s close. On down days,
momentum is yesterday's close minus today's close.
Table 5.6 gives us two key results that will ultimately be
used to define RSI. The relalive momentum over four days
Table 5.5
S&P-500
11/29/ 91
12/ 02/ 91
12/03/ 91
12/04/ 91
12/05/91
375.22
381.40
380.96
380.07
Jn.39
6.18
-0.44
-0.89
- 2.68
6. 18
0.44
0.89
2.68
2.17
2.17
10.19
Total
s. -
Date
11 /29/91
12/02/92
12/03/91
12/ 04/91
12/05/91
Total
Relati ve
Absolute
Up-Day Down-Day
S&P-500 Momentum Momentum Momentum Momentum
377.39375.22 2.17
0.5*(momen1um
+ lmomentuml).
(5.13)
Calculating RSI
375.22
381.40
380.96
380.07
377.39
(5. 12)
You can vcrify this result from the data in Table 5.6 (S. O. 5 ( 10. 19 + 2.17} = 6.18). We will use 1hese resulls to define
Wilder's RSI (see Bibliograpby).
Table 5.6
sd,
s.
Wc can solve tbese two equalions for s. using algebra Simply add thc two equations and gather terms 10 find:
Relalive
Absolute
Momenlum M omentum
Date
121
6.18
- 0.44
-0.89
- 2.68
618
0.44
0.89
2.68
6.18
0.00
0.00
0.00
0.00
0.44
0.89
2.68
2.17
10.19
6.18
4.01
(S.)
(S.J
Thc rcla1ive strengtb index measures the proportion of momcn1um change over a given period caused by momentum
on up days. Tbis will become clear from the following definition:
RSI - 1OO*(RS/ (1 + RS)).
(5.1 4)
(5.15)
122
= 100*(0.5*(momentum + momeotumJ))/(lmomentum),
= 50*(momentum
+ lmomentuml)/(lmomcntum).
(5.18)
We have found an important result bere 1hat directly ties
RST inlo rclative and absolute momentum. Over x days, the
RSI is tbe proportion of absolute momentum beca use of up
days. Hence, the x-day RSI is equivalent to the x-day momeotum.
In Table 5.7, we did tbe RSI calculations using the sum of
up-day and down-day momentum. Next, we will redo the 4day RSI calculations using tbe relative and absolute momentum to show that we gel the same value using eitber method
(see Table 5.8). Now it is easier to understand what the RSI
is trying to capture by using the ideas ofrelative and absolute
momentum.
Rather thao use Wilder's original method, we lik.e to use
123
Date
S&P-500
Up-Day
Momentum
(SJ
11/29/91
12/02/92
12/03/91
12/04/91
12/05/91
375.22
381 .40
380.96
380.07
377.39
6.18
0.00
0.00
0.00
0.00
0.44
0.89
2.68
6.1 8
4.01
Total
Down-Day
Momentum
(S.)
RSI
60.65
S&P-500
11/29/91
12/02/92
12/03/91
12/04/91
12/05/91
375.22
381 .40
360.96
380.07
377.39
To tal
Relative
Momentum
Absolute
Momentum
- 2.68
6.1 8
0.44
0.69
2.68
2.17
10.19
6.18
- 0.44
-0.89
RSI
60.65
124
(5.19)
Here RSIH and RSIL are the highest and lowest valucs ofRSI
over a gvcn look-back peri od. Whenever RSJ makcs a new
lo~, ~e stochRSI wi!J be at O. Tbis is an example of a failure
swmg .m a down move. During an upmove, thc RS I will make
new highs (ovcr thc caJculation period) and the stochRSI wiU
be near 1.0. Thus, you could use stochRSI both as an overboughtfoversold oscilJ~tor and to follow trends in RSI. Divergences are also eVIdent on the stochRSI plots, which
125
126
New Momentum
Oscillators
00
ro
t---------------------1 ~00
o
..
..
corrected or consolidatcd. These signals are unrnistakably evident in the stochRSI plot, but not as obviously cvident in
the RSI plot. Thus, thc stochRSI analysis is more useful in
this example.
In Figure 5.20 we show a close-up oflhe T-Bond September
1993 contract from Fcbruary throush June, J 993. The
stochRSI showed each significan! overbousht and ovcrsold
condition over this period, Oagsing a sbon-terrn tradeable
rnove. The divergence in March that we spotted in Figure
5.19 can be seen on an expandcd scale bere. Buying around
lhe 0.2 leve! and selling around lhe 0.8 leve! would have been
profitablc. lt's importan! to reenter the market if stochRSl
again rises above 0.80 or falls below 0.20. Say you sold the
bonds when the stochRSI feU below 0.80 in early June; the
uptrend resumed and the stochRSI again moved above 0.80
FIGURE 5.20
128
129
NEW Sf.ANOAROti1y OlQ1N1 C-310 470 -+ 1 1 1 ()io'367 070 11=370 110 t.El.733
tqi,111111 37110ll
ti
1 111 1
1,
"1 1
11
,,.,
111
'I
11
1
11
1
EllDJ
3lll OlJ
3400:0
:m.m:J
l:llllDl
11
111 1
310JXIJ
mOJJ
-~~
..
.. ........
-o~llXXI oxeo
........
93
,..
. /\
~
....
0.80 and then fcll below that level. A buy sigoal occurred
when this indicator was bclow 0.20 and then exceeded that
level. Note the reversa! co nditi.on: lf the stochRSI reverses
after crossing abo ve 0.20 or rises after falling below 0.80, theo
you should reverse the bu y or seU signal. The reversa! condjtion occurs because stochRSI can stay above 0.80 during
uptrcnds and below 0.20 during downtrends for the duration
of the trend.
Nexl, we' U look at the S&P-500 index in Figure 5.22 coveriog the pcriod November, 1991 lhrough February, 1992.
Note how the 14-day stochRSI in this figure weakened aU
lhrough December, 1991 as the index moved up to the 330
area and then entered a light trading range. The stochRSI
settled at O showing a downtrend just before the decline to
""
v.~
91
F..
the 310 area, and it moved offits bottom two days before the
index itself bottomed. The index and stochRSI rose strongly
in January until tbe stochRSl was at 1.00, indicating a strong
uptrend. Tbe stochRSI reroained above 0.80 for the entire
rise into February, again fafog off withio a day of the high
for the move. As the index entered a trading range, the
stochRSI fell off again dropping to O.
The important idea to grasp here is tbat tbe stochRSI feU
when the iodex entcred a trading range. The stochRSI is qwck
to react at both bottoms and tops, so you can trade it only
when it is al ao extreme, showing a trend. Acc.ordingly, you
can buy when it rises above 0.80 and sell when it faUs below
0.20. You can then e.lose your posilion (or reverse) when the
values fall below 0.80 or rise above 0.20. Or, you can trade
ooly when the stochRSI falls below 0.50 or rises above 0.50.
130
131
well before thc huge scll off. lt also picked most of the small
rally that followed. This examplc also prove~ the tim_elne~s
and scnsitivity of stochRSl. You couJd expenment with d1fferent time periods of calculations and action Jevels, and
hence use stochRSI for long- as well as shon-term analysis.
But what about using stochRSI to analyze a very short time
frame? Figure 5.24 illustrates the use of the stochRSI with
hourly data using the September, 1993 T-Bond contract. Here
we smoothed the stochRSI heavily to remove the noise nherent in using such a shon time frame for analysis. We computed a 20-period RSI and a 20-period st?Ch~stic oscilla~or.
Then, we smoothed the 20-period stochasuc with a 10-penod
simple moving average and plotted the slow stochRSI over
the hourly price bars shown in this figure.
The momentum rose as prices rose and fell when prices
entered a trading range. You can see thal the stochRSI can
fall even when the market is moving sideways. A sideways
........
...'
...
'
o ,
'
......'"
,...
o.
o.o
U.P
........
111.
111.0
..' .'
'"
IUO
UI
111
11
110.
11
A GUR.E 5.23 The stochRSI os shown overa longer lime period using
weekly data for Amgen
.....,...
.."..'
o
o.
11.
...
...' ..
l lf .O
11e,. ~
U'-~
111.c
111
..
u: 11
111.'t
110
A CURE 5.24 Hourly data for the T Bond September, 1993 futures
contract 1s used 10 illuslrate the achon of the stochRSI.
132
133
.'
-......
,..
~
..
.."
e
IH.a
11111.0
..
'
111.0
1
.o
11>.
11-..
lll.9
u1.o
llf.
11
u .o
.....
1111.11
12.t
tu.o
111.1
111
'5.
11
..
..
111.0
110.1
110.0
FIGURE 5.25 The 20-peod RSI for Lhe hourly T-Bond September
1993 contraC1.
'
01 ...u ..)
....
......
....
....
....
....
....
........
....
... ,.,.
..,.
.1''
LOO
::
~ _#."l.
~
..'
.
'
.,..'.].
....
....
:. !'
'.
...
.o
...
"
"
,.
'
'..;'
..
"
"
......
..,.
......,
....
....
...."
....
"
134
Vui.ible
~nglh
Dynamic Momenlum
135
(5.20)
(5.21)
We arbitrarily restricted the maximum and mnimum numbcr of days used for DMl calculations to 30 and 5. These
limits fit our trading hori.zon, but you can use other Iimits as
you wish.
Be aware that the conversion is nonlinear, arising from
lhe deftnition of the volatility index itself. The values of tbe
percentage change in V, will give you a sense of the nonlincarity. The conversion is more sensitive wheo the index is
less than 1 and less sensitive to cb.anges when lhe index is
greater than 2.
As shown, we buih the variable length DMI around the
static RSI length of 14 days. Heoce, when the V, is approximately 1, the DMI and RSI will have s.i milar values. As 1be
index drops bclow 1, DMl and RSI values will diverge
quickly. As the index increases above 1, DM I and RSI values
will diverge slowly. Hence, the exact narure o f the DMJ-RSI
136
curves will depend on tbe volatility in tbe data over the given
test period. Tbe divergence will increase as volatility increases.
137
3050 .-~~~~~~~~~~~::;:-~~~~-,
3000
2950
2900
80
?O t---''fli~:--~~--17;--~~~-tt~.-::;-~~-::"iti-j
80
so
40
RSI
2850
2800
2750
2700
2650
2600 '-~~-=~...:.~~~__:;.;.;.;~~..-:::....~~~~~-'
910201
910304
910402
910430
910&29
910626
9107U
FIGURE 5.28 The Oow Jones Industrial Average and the eective
length of the OM l.
in early March, la te March, mid-April, lat~ May. and midJune. lo eacb instance, the market was makiog qwck moves.
This example makes it clear that you can use the DMI to
identify overbougbt or oversold conditions. DMI usually
leads RS! into the extreme regions by one or more days (see
Figure 5.27), and lbjs lead time can prove valuable 10 many
you. Because DMI lengtb is more closely tied to. markct dynamics, 1be extremes tba1 DMI shows are more likely to lead
to profitable antitrcnd trades tban the extremes shown by
stochRSI. In o tber words, DMI is more sclective about sbowing extremes. You can also use DMI to develop flexible parameter trading models.
JO r-~~~~;f-~~~__:~~~~~l'--~---l
20
OMI
10
o
910201
910304
OVCASOlO REGIOH
910402
910430
810529
910828
FIGURE 5.27 A comparison o f unsmoothed RSI and the OMI for the
Dow Jones Industrial Average data.
138
139
8000
80 000 .....-~-:-:--~~~~---::--~~~""1
8000
55 000
7000
60.000
8000
~
i\~
445 000
:i ..oooo
~
'tvl
35.000
30000
S&PS~ 1na...
425 000
Oyn1mlc momencvrn lnde:c (OMI>
5000
CIOll
20.000
tS.000
/_
da~ l./'
00
L~...::..:.:.:....~~-:--:~~::::-::;-:;-:;:;;-;-:;;~
"""""...,""" .... -....,;t""""""2~...,~~
""""""~~~~~~~r~~~~~~~~~sss~~
~a~~sSS~~is~S~aaa~a~aaoa
on the other hand, sbowed overbougbl and oversold conditions on four occasions over this period.
In Figure 5.30, we overlaid the DMJ on the S&P-500 index
daily close. You can see that tbe extremes in DM 1 correspond
to extremes in the index as well. One trading stra1egy is to
buy when the DMJ goes below 20 and then rises up. You
would sell wben the DMI is above 80 and then crosses below
tbis level. This approach would have given tradeable signals
three times in February-March, 1993. Each signal preceded
substantial moves by about two days. T his provides you with
is a significan! edge in trading.
Fina lly, Figure 5.31 shows the e ffective length, that is, nurnber of days, used for these D MI calculations. Low values
correspond to periods of higb volatility. Convcrsely, high valucs fo r the effective lcngth imply periods of low volatility.
You can see from Figure 5.30 that quick moves in thc market
reduced the effective length of DMI calculations.
35.00
3000
.,,
2500
20.00
1500
10.00
5.00
2000
10.00
~~s-es-~J~ - ~~~8----~~~
FIGURE 5.29 A companson of 1he 14-day RSI and the 14-day OMI
for the S&P-500 mdex data 1n Figure 5.30.
0.00
30.00
.,
..
!' \
---
1 '
'
140
Summary
141
6
Market Thrust and
Thrust Oscillator
This chapter introduces powerful new indicators that combine ilems of data unique to the stock market and unavailable
for the futures markets. Tbese four data itcms are the number
of advancing (AJ) and declining issues (DI) and advancing
(AV) and declining vol ume (DV). Tbe Al and DI count issues
without regard to the extent of price change or the market
capitalization.
Market analysts usually look at thesc data in a variety of
ways. On days when the market is up strongl y, Al > Dl by
more !han 1000, and AV> DV by a factor of 3:1 or more.
Thus, the advancing volume could easily be four times (or
more) than the declining volume. Similarly, there is general
agrccment that, on days when the market is very weak, Al
< DI by severa! hundred issues, and DV > AV by a ratio of
3:1 or more.
On days when thc markcts move decisively in either direction, wc see a pattern in which Al > DI, AV > DV or Al
< DI, AV < DV. There also can be other combinations of
these variables, which produce ambiguous cases bccause it's
common to analyze tbe advanciog and declining issues as one
143
144
(6. 1)
Marl~I
145
146
MT-linc,.,..,..,,,
(6.4)
(6.3)
147
(6.5)
TABLE 6.1
Da y
1
2
3
4
6
7
8
9
10
11
12
13
14
15
16
17
Al
AV
DI
DV
1000
100
1000
100
800
600
800
1,000,000
100,000
1,000,000
200,000
800,000
600,000
700,000
900.000
700,000
600,000
500,000
500,000
600,000
700,000
400.000
500,000
400,000
100
1000
100
1000
400
600
400
400
100,000
1,000,000
200,000
1,000.000
400,000
600,000
500,000
300,000
500,000
600,000
700,000
700,000
600,000
500,000
800,000
700,000
800,000
800
600
700
600
700
500
500
soo
400
400
600
500
600
500
700
700
700
800
800
TRl'J
MT
TO
990 0.98
1.0
- 990 - 0.98
1.0
980
0.96
2.0
0.50 - 980 -0.%
480 0.60
1.0
1.0
0.47
1.40
3&0
0.71
0.67
600
0.71
120 0.17
120 0.17
1.40
1.40 - 120 -0.1 7
o
1.96
o
0.71 -120 -0.17
o
0.52
o
1.43 - 360 -0.47
0.70 -360 - 0.47
t.00 - 480 -0.60
o.o
148
Varialions of MT and TO
Consider the special case wben thcre are uni t volume tlows
in!o advancing and declining issues, where AV - DV = J. In
this case, the AV (or DV} 1enn mereJy acts as a constant
muJtiplier, and lhe MT-line simply collapses to 1he usual cumulative Advance-Decline (A-0 ) line, since:
MT - (A l - D l)(AV- DV - 1),
MT-line_, - (AJ - DI) + MT-line,.......r
(6.6)
Therefore, we can expec1 a broad similari1y betwceo the cumulative MT-line and the A-D line. Tbc two ofteo diverge
at tuming points when volumc rotales predominantJy into
advancing or declining issues. T herefore, the MT-line is more
useful an the A-D line at 1uming points.
Ncxt, considcr the special case wben AV = DV which converts the TO into a variant of tbe advance-declin~ ratio since:
TO - (A l - 01)/ (AI + 01) (if A V- DV}.
(6.7)
(6.8)
149
150
shares was greater. Note how Thrust reads -980, and TO 0.96, capturing the beansh mood ofthe day. Day 5 is neutral
says TRlN; MT says it is moderately strong on the upside.
This shows that TRIN does not consistentl y show bearish or
bullish market activity.
TRlN correctly calls day 8 a bullsh day, as do MT and
TO, but calls day 9 more bullish lhan day 8. Thrust correctly
calls day 9 mildly bullish, bccause up volume cxceeded down
volume. Here an equal number of stocks advanced and declined. TRlN is bearisb on day 10, while Thrust is unchanged
from day 9, correctly calling the mildly bullsh day.
Day 11 provides a striking contrast betweeo TRIN and MT
and TO. TRJN calls it strongly bearish; MT and TO call it a
standoff. It is thc same story for day 14: TRIN says great
bullish demand, Thrust is dead neutral. Here thc divergent
interna! dynam ics of TRIN give conflicting values.
Relacive volumc nows, defined as Dl/DV and Al/ AV, simply do noc consisteotly show if lhe market action was evenly
split or onesidcd among the four variables. In contrast, because oftbeir definitions, bolh MT and TO clearly show when
the action was evenly split or one-sided among these variables, and provide a more consisten! picturc of market accion
than TRIN.
In essence, using thc product of Al*AV and Dl*DV elminates tbe confusion that rcsu lts from taking the DV/ DI and
AV/ Al racios found in TRIN. What's more, wc can easily use
moving averages to smooth MT and TO in ordcr to reduce
noise and sec the underlyi ng cbanges more accurately.
:ro
,_
... .
,._
.""
Now we'll look at real markct data using thc thrust oscillator.
TO is bounded betwccn + l and -1 (or + 100 and - 100),
having the same scale for up and down days. Conversely
TRLN is unbounded on down days.
'
,_ ' -
"
-.
.....
.,... l.t"
-
...._. ~,_
HM O
".
fo
,..
- '-
~
-
U H
-"".
'-' o
--
U N. O
"' "'
1 ..
... ...
.. ,_,. -
K V. .~
,...,..~
......
1-
1 . ...
UM O
151
~~
'-
"..
H C
.
.
I".., .
l:
..
';'I
~u
~ !/ ,.
...:
11011
- ,_
-.. "-
'f- t:'t:
..
'
"""'
'
1111. . ...
.....-
...
_:
FIGURE 6.1 A 10-day smple movng aver~ge o the Arms index and
the Dow Jones Industrial Average in 1987-88.
152
....._
.......
-...
"''"
~
...
""
.....,....
....
..........
'"'.
.... .
...." .
....
,,....
....
.... .
l lft
...
..
..
--
......
...
.......
.. '
...
.....
KO
FIGURE 6.2 The 21-day simple moving average of the Lhrust oscillator for the same period as in Figure 6. 1.
FIGU RE 6.3 The OJIA and the 21-day smoothed thrust oscillalor at
the markel bouom in t 990.
.....
-
.....
... ..
.....
.,,, .
.....
,.,...
.... . I~ l~li
.,... fl
...........
...
._
UJI 1
.~. .
._.
..
10
dr1
1111
'
~11'
...
FIGU RE 6.4 The 2 1-day smoolhed Arms index for the same period
peaked al 1.20 al 1he marke1 bottom m 1990.
155
December was panicularly mely. Note also the strong market tbrust in January- February, 1990 tbat took the 21-day
smootbed TO to thc 0.50 area. Sucb market thrusts often
signal highcr prices 12 montbs into the future.
In Figure 6.6, we compare tbe 21-day smoothed TO (lower
curve) to the 21-day smootbed TRIN (upper curve). Notice
that TO and TRIN mirror one another. Tbe smootbed TRIN
sbowed oversold conditions near values 1. 1; TO and TRJN,
bowever, do not peak at the same values, wbicb is what we
expect in a trading range. For example, smoothed TRIN
dropped to the 0.70 area in both February and AugusL However, TO was at 0.50 in February, and only 0.20 in August.
Tbus, thc 1hrus1 was much stronger in February tban in August.
Similarly, 1he smoothed TRlN rose above 1.1 on seven
occasions, whereas 1he TO approached its leve! of - 0.20 level
....
l HH
1 . .. .
......
..........
..........
.""
...
.. l ..
MMt
FIGURE 6.5 The DJIA and the 21 -day smoothed thrus1 oscillator wilh
1he marke1 m a narrow range.
,-.-.
llU
....
il'U '
,..
l.. . ..
f1'
MT
O!fll
f l(
f 'lll
6.5.
157
_15_6_ _ _ _ _ _ _ _ _ __.:_M
:.:::arket Thrust nd Thrust Oscillator
just three times during the year. For example, the May peak
in smoothed TRIN dropped the TO to just the -0. 1 leveL
Finally, the smoothed TRIN stayed flat in early December,
but thc TO moved up sooner off its bottom.
In shon, this comparison in trending and lat markets
shows that the TO is a better osciUator than TRIN because
its signals are more consistent and timcly.
.,...
~~----'----
_.
__ __
__,
__.
___..__--"
Average
Buy
784.00
842.00
813.00
1165.00
1985.00
2644 .00
2611.00
2452.00
21.17
-2.26
28.54
6.09
1.76
3.40
10.65
22.31
28.44
9.14
46.71
9.61
8.06
7.03
15.89
25.49
6.89
48.81
39.1 1
45.75
30.83
22.73
24.63
30.02
3.74
11.46
19.05
31 .10
158
159
SUMMARY
. . . . . i-
..... ""
.""""'"""
.....
.. ..
Summary
. . . . . i-
f\.
/\
...
V /
\''\ -1-
!"\
t;. ,......
r'\.!;\
\ [f.~
""
rv
V
10
Kt
MC
~l
Mt
, ...
la
H -
MI
. . . H<
,, ... ,_,,
dttttttrtttJ1Y
:::~l
...
Uf'
MI
. . . . .(
l .lt
fU
-1
rlf."
t'IMI.
114
tO
9(1
. . (:
IJUI
f U
curring since 1980 did so with an overbought condition followed by a drop in thc 50-day smoothed TO below its own
50-day simple moving average. This means that downside
thrust increased after thc top and led to an intermediate-tolong tenn bottom with the 30-day TO average dropping to
tbe - 0.2 to - 0.3 area. In Figure 6.8, you can see the market
top in 1987. In particular, note how the market was overbougbt (30-day RSI o ver O. 70) and the 50-day smootbed TO
(uppergraph) fell below its 50-day movingaverage. Tbis shon
position began in late August, weU before the major selling
episodes. The graph also shows that this approach is not perfect and should be used wi tb caution.
We hope you can see from the previous discussion that market
thrust is a powerful new way to analyze stock market action.
Tbe thrust oscillator TO provides more consistent readings
tban TRrN and can be smoothed witb averages without distortion. The TO combines an ad vanee/decline oscillator with
a volume oscillator to provide a uniquely seositive market
indicator.
7
Controlling Risk:
The Key to Profitability
162
163
164
165
20
..
15
....~
12
'5
1
:z
- ~---1
o
<500
----
501750
751-tOOO
tOC>11250
t2511500
166
60
'!! 40
20
ol__~~~~~====::::::::====~
c.1000
10~2000
2000..3000
3000-4000
167
>-4000
FIGURE 7.2
make a sudden move against your PoSition. Tbe exact location of a stop is often critical to tbe outcome of tbe trade.
Because oftbis, we'll discuss many different ways to set stops
1ha1 you'll find useful at one Point or another.
One of severa] methods for setting trailing stops is to change
the stop jusi once a week. Say you use the Wednesday close
10 calculate the profit over thc past weck. You can theo advance the stop x percent ofthe weekly prolit, say 40 percent.
This mechanical stop will advance when you have a profit,
but will not step back when you bave a weekly loss.
Another simple approacb is to use a fixed do llar trailing
stop. For instance, you couJd use a $1 ,500 trailing stop, measuring tbe amount from tbe highest high or lowest low during
tbe trade. The exact dollar amount could be an arbitrary
amount from $500 to $5,000 based on your trading style.
One classic method is 10 set tbe stop j ust below a swing
low or high. A look at the price chart will show tbe most
reccnt significant high or low. Tbougb sucb a stop can be just
bcyond the recent high or low, be aware that such stops can
be gu n ned by floor brokers.
11 2 ! -- - - - - - > , 110
108
FIGU RE 7.3
168
169
Juo
FIGURE 7.4 A volatility-based stop used with the Japanese Yen fulures con1 raC1.
170
conditions. We suggest that you use the historical trading performance of your system to plot a template for assessing
trades.
You can plot the day-by-<lay evolution of each trade from
your system by ploning the best and worst equity of each day
on separate charts. If your system has provided more than
30 trades, you can develop a fairly represen ta ti ve trade profile.
First, you would average the best equity of eacb trade on tbe
first day of tbe trade to get a single numbcr for day J. Tben
you would do so for day 2 and so on. (Thcrc may be fewer
tradcs open as the number of days incrcascs.) Then you would
average th e equity across only open tradcs.
Your plot sbould be similar io the one shown in Figure 7.5,
whicb is from a system for trading 10-year Treasury notes.
The middle line shows the best equi ty for tbe average trade,
which peales nine trading days into its cxistence at about
S 1,500. We also plotted the lines one standard deviation ( + 1
sigma and - 1 sigma) away on cilher side of the average.
Tbese plots should account for about 67 percent ofthe trades
from tbis model. The upper line shows that the more prof-
Average winning Hd
2500
1&00
1000
--
171
......=:~-- ...- - - -
10
t1
12
13
O.yo In Trst
FIGURE 7.5 A typical trade template for a system trading the 10-year
Treasury Note contract.
10
11
12
13
01y1 WI 1t1de
FI GURE 7.6 Two actual trades overla1d on the typical trade template
for the 10-year Treasury Note contract.
t72
~~~~~~~~~~~~~~
t73
Hl - C0 + A,
H2 - C0
L l - Co
L2 - Co
+ 2*A,
- A,
- 2*A.
(7.1)
We denote tbe absolute momentum by lmomentum~ today's close by C,,, and yesterday' s close by C ,. The estimat~s
for tomorrow's highs and lows are H 1, H2, Ll, and L2. This
calc11lation says our best estmate for tomorrow's close is today's close. The total range is four times A ( 4*A}, being a
span of 2*A on eitber side of tbe close. You can vary tbe
n umber of days used in the calculations as well as tbe multiples used to project the span. This approach has tbe advantage that, as market volatiUty changes, the projected span will
change as well.
Now that we know how to co mpute price targets, we'U
illustrate how you can develop scenarios and risk control
plans. Assume tbat the markct has been trending lower and
you have a sbort position. A strong rally tomorrow wo11ld
close near or above H2. You also feel that such a strong rally
probably signals a reversa! of tbe dowotrend. In sucb a situation, you could cboose to pul in a stop order to close out
your shortand go longa fewticksabove H2. But, iftbe ~~et
traded between H2 and H I, you would hold tbe short poSJtJOn
and do nothing. Similarly, you would continue to hold your
short position if tbe market tradcd between LI and L2. In
addition, however, you feel only a selling clmax could pusb
the market below L2, so you could decide to cover your sbort
174
175
m
m
m
m
...10.0..Y AllEAAGE Of AESOtllll: OM.Y CHANCE
Cjl I CLOSE Of YESTERDAY
q 14
,,
2""
21
12
changing volatility. Tbis fcature is useful in tradiog the volatile futures markets.
In surn, it is clear that our proactive risk control stratcgy
based on projected price ranges can be successfully implementcd in the futures markets.
176
177
00 L=82 D
EO
e
Jul
FIGU RE 7.8
contrad .
93
Wben you devclop your price targets H2, HJ , LJ, and L2,
observe whether thcy faJI ncar any significant retracemcnt
points such as 33 pcrccnt, 50 pcrcent, or 67 percent for the
current move. Thcse are callcd Fibonacci retracements, and
often provide points of support and resistance. Figure 7.9
shows these Fibonacci retraccmcnts in the Higb-Grade Copper September, 1993 con tract. The major move from A to B
produced a 50 percent retraccmcnt to C. Smaller moves from
A to g produced a 64 pcrcent rctracement to h. Similarly, !he
move from d to f produced a 47 pcrcent retracement to e.
The conclusion? Pay particular attcntion if your projected
targets happen 10 be near Fibonacci retracement levels.
Aftcr you calculate a 1railing stop, avoid placing it near a
round number such as 62.00 or 62.50, but place i1 7-33 ticks
past the round number. For cxample, set your sell stop at
60.87 and hope tha1 the market will find support near 61.00.
You may notice that reversals in the US T-Bond market usually oocur less than 7 ticks beyond the previous higb or Iow.
Thercfore try 10 place your stop say 11 ticks beyood the
previous high or low. Often, tbosc fcw extra ticks may be the
dHfercncc belwecn having a winning trade or avoiding a losiog onc.
Evcn though wc have continuaJJy mentioncd stops (stop
ordcrs) for risk control, using market orders may be tbe best
way to cxit a pcsitioo during the trading day. Once you decide
to closc a pcsitioo, get out as fast as pcssible.
Study the type of orders your broker wilJ accept, since they
can influence how you actuaJJy enter orders for your risk
control plan. A well-placed order will help you trade bener.
Anotber ordcr placement tactic is to use tbe facsimile machine
178
8
How to Use This Book
The best way to use this book is to integrate our ideas into
your own trading proccss. We'U give you examples ofhow it
can be done by combining indicators to find new, elfective
trading systcms. For instance, we'll show you how to devclop
an adaptivc trading systcm combining the momentum oscillator, CMO, with VIDYA. We'll also show how you can use
the ideas of composite momentum and linear regression
analysis for market rotation-moving out of quiet markets
i11to those experieocing major moves. We hope these examples may stimulate otber ideas in your own mind on how to
combine tbe new indicators in this book.
180
181
c Mo - K2Rs 1. - 100)1q
A - t4CMOj (t <- 0.5),
(8.1)
VIDYA, - A. Close + (1 - A) . VTDY\..
182
Market
Bean o il
Bn1ish pound
Canadian $
Cocoa
Coffee
Cotto n
Crude o il
Deutsch mark
Eurodollars
Gold
Heatmg 011
HG copper
Japanese yen
Uve cattle
Uve hogs
Po rk bellies
Sil ver
Soybeans
Sugar
Swiss franc
T no te 10-yr
US bond
USDX
Whcal
Average
Stand. Dev
Avg/stdev
Max
Min
Ma.x/min
o.os
-1078
16682
1480
- 2170
11177
-515
- 790
8000
1025
- 2350
109
4372
6887
-1 548
4175
- 5492
- 4955
3150
2824
15050
6481
2981
1920
2206
2785.04
5529.34
o.so
16682
-5492
3.04
- 2778
31381
2860
-350
5841
-4510
- 1490
5087
3325
3250
177
4687
134 12
1312
5106
- 163
- 4115
2663
2983
21262
1191 8
- 1087
1470
5431
4306.68
7987.04
0.54
31381
-4530
6.93
0.20
-3598
30537
2860
-2070
718
-7310
- 610
32 12
2525
5550
-2013
5787
14325
820
4900
- 1306
-5075
800
5 10
5775
8812
- 6206
1290
4043
2571.84
7521.35
0.34
30537
-7310
4.1 8
183
184
:
:
:
:
05/26/83
$50
S75
SJ.000
05/30/88
$23,400.00
$48,662.50
$ - 25,262.50
33
36%
12
21
58,262.50
S- 2, 175.00
$4,055.2 1
$- 1,202.98
3.37
$709.09
2
5
79
13
$-5,950.00
$-6,850.00
1.92
l
$9,850.00
237%
185
186
S49,775.00
$77,737.50
$ - 27,962.50
27
37%
10
17
$1 7,012.50
$- 3,037.50
$ 7,773.75
S-1 ,644.85
4.72
$1,843.52
3
4
102
10
S- 8, 112.50
2.78
1
$11 ,712.50
424%
CMO-Ori~n
187
8.4.
As you can see, this model does follow long-term trcnds,
judgiog by the oumber oftradiog days io the average winning
trade. There are oo guarantees tbat it will work as well io the
future io every market, bowever. It is simply a good 1reodfollowing tool thal adapts to market volatility, and il's probable that this model could be profitable io trending markets
that have occasional volatile periods.
We could now test tbe model over successive 12-month
periods to develop data on the variabilily in tbe payoff ratio
and winning percentage. Or, for planning purposes, we could
assume a 35 percent winning percentage and a payolf ratio
of 2.50.
Two weaknesses in this model deserve mention. One is its
perfonnance in sideways markets or trendless markcts tbat
produce whipsaws. The other is its performance in very volatile markets, which can stop the trade out at a loss before
188
Table 8.4
Model Name
Notes
Cale Dates
Comm1ssion
Slippage
Margin
:
:
:
:
:
;
Data
Total net profi1
Cross profit
Cross loss
Total # of trades
Percent profitable
Number winning trades
Numbcr losing lrades
Largest winning lrade
Largest losing trade
Average winning trade
Average losing 1rade
Ratio avg win/avg loss
Avg trade (win & loss)
Max consecu live wlnners
Max consecutive losers
Avg . bars in winners
Avg # bars In losers
Max closed-out d rawdown
Max 1ntraday drawdown
Profit factor
Max # o con1raC1s held
Account size required
Return on account
CSI 39
Continuous ContraC1
SJ 1,237.50
$56,900.00
$ - 25,662.00
$29,737.50
$57,912.50
$- 26, 175.00
29
27%
6
21
30
23%
7
23
$17,300.00
S-2,900.00
$7, 1 12.50
s - 1,222.02
5.82
$1,077.16
$17,312.50
$ - 2,900.00
$8,723.21
$ - 1,225.00
6.75
$991.25
3
7
98
6
S- 8,687.50
S-9,275.00
2.21
1
$12,275.00
254%
3
7
109
9
$- 8,637.50
$-9,225.00
2.05
s 12,225.00
243%
Market Rotallon
189
MARKET ROTATION
190
19 t
M<arket Rottion
Table 8.S
Market
Swoss franc
British Pound
Solver
Coffee
Gold
Japanese yen
Deutsc he mark
Sugar
Caule
Heating 011
Crude 011
Eurodollar
Cocoa
Soybeans
Frozen O J
Corn
Hogs
Conon
W heat
us bond
Canadian $
Copper
US S index
Bellies
lumber
Composile
Momen1um
SIo pe
18-Day
($)
($)
8288
7550
6875
611 1
5340
4736
4525
3069
1092
916
530
94
- 190
-338
- 645
-938
- 1497
-1560
- 1875
- 1906
-2180
- 4763
-6350
- 8496
- 14850
232
234
97
98
97
205
99
91
12
-17
- 23
60
- 3
- 14
- 68
- 28
- 40
- 51
- 48
74
- 24
- 152
-181
-138
-173
R-Sqd
18-Day
0.74
0.73
0.60
0.29
0.67
0.63
0 .49
0.41
0.09
0.26
0.44
0.28
0.0 1
0 .20
0.61
0.67
0.62
0.39
0.76
0. 11
O. IS
0.73
0.77
0.56
0.26
ADX
18-Day
48.72
54.67
34.94
30.61
45.89
40.61
41.72
18.22
31.33
12.00
18.44
25 .22
14.33
22.06
28.17
45.00
2 l.11
15.28
34.78
26.33
17.94
64.83
51.44
11.78
56.33
Table 8.6
Rank
1
2
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
0 7/ 13/ 93
Com
poste CMOIO CM020 CM030 CM050
39.30
37.12
32.48
24.36
15.95
14.89
13.96
13.65
3.58
- t.56
33.33
55.56
46.15
47.83
63.64
37.93
18.18
4.76
24. 14
0.00
- 1.95
-2.17
- 2.90
- 3.53
-8.26
- 9.49
- 12.16
- 13.25
- 15.56
- 19. 15
12.50
- 36.84
- 14.29
- 30.30
-t.96
- 7.14
- 8. 11
- 41.67
- 10.00
-42.86
- 19.41
- 20.47
- 24.09
- 24. 16
-27.96
- 28.23
-29.40
33.96
- 4$.74
- 55.56
-1 4.29
- 12.82
- 31.25
- 37.78
- 49.02
- 65.22
- 33.33
- 73.91
66.67
- 53.85
37.50
14.81
42.22
10.38
46.84
6.90
41.46
1.62
12.50
1.34
5.66
3.60
8.11
4.67
21.95
7.40
20.63 - 6.96
-4 62 - l .21
-1 0.59
15. 15
7.69
8.94
- 12.62
-8. 20
- 18.64
3.23
- 24.44
-17.81
- 23.33
- 34.69
- 34.94
-36.08
- 28.74
-20.00
- 47.37
-29.63
- 52.38
-77.27
- 1.97 - 3.81
2.32
6.02
- 1.67
0.00
1.68
2.19
- 3.66 -7.50
- 3.79 - 1 1.27
-6.67 - 1.89
-.43 - 13.29
-6.22 - 3. 17
- 5. 13
- .56
- 7.74 -16.83
-8.21 - 9.76
-7.79 -6.81
- 6.37 - 3.78
-10.97 -1.22
- 4.53 - 14.10
-9.10 - 9.62
- 11. 12
1.02
-13.14 - 24.53
- 18.53
35.56
192
41.94
19.57
16.22
3.28
- 8.33
5.17
15.53
5.66
- 9.59
7.98
+ CMOS0)/4
Mari.el Rolation
193
194
How to
Us~
This Book
llem
Slo""
r'
VIOYA
ns tic k
CMO
St"rhRSI
MT
Thrust Ose.
ITOl
MFE
Volatility
SIOD
Price Tantets
Trade
template
CMOdriven Market
VIDYA Rotation
Stock
Marke1
Timing
Open
Trade
ManagePrice
Trend
ment Extremes iness
.
.
Bibliography
stop, pricc projection, and the trade template. Tbey can belp
you place unique stops critica! to profitability.
Dcpending on your time frame, you could identify price
extremes using the regression slope or momcntum oscillators
(CMO,. stocbRSI, or DMI). On occasion, you may want to
determine presence or absence of trendiness in a market by
using the linear regression r2, the absCMO, or even a volatility
stop. You could derive many other combinations by adapting
thcse flexible and powerful indicators to your trading style.
We encourage you to do so, for your unique combination
could boost your profits by giving you a mucb oeeded analytical edge in today's tougb markets.
Altman. Roger. ''Relative M omentum lndcx: Modifying RSl." Techmcal Analysis o/Stocks & Commod11ies, Vol 11, February, 1993 p .
30-35.
Arms. Richard. The Arms lndex: An lntroducuon to the Volume Analys1s of the Stock and Bond Markt'ls. llomewood. lL: Dow Joncslrwin, 1989.
Babcock Jr., Bruce. Business One lrwin Guide to Trading Sysrems.
Ho mewood, ll.:, Business O nc lrwin , 1989.
Balsara, Nauzer J. Money Managem ellf Strategiesfor Futures Traders.
New York: John Wiley & Sons, lnc., 1992.
Bla u, William. "Tbe True Strength Tndex," Tec/111ical Analysis ofS1ocks
& Commodities, Vol 9 No. 11, November, 199 1 pp. 18-31.
Burke, Gibboos, "Gaio without Pain: Money Management in Acon,"
Futures Dec 1992, (Vol XXI 1114) pp. 36-38.
Cbande, Tushar S. " Adapng Moving Avcrages to Market Volality."
Technical Analysis of Stocks & Commodities, Vol 10, No. 3 ( Mar
1992), pp. 108- 114. Copyright 1992 Tecbnical Analysis, lnc. Used
with permission.
Cbandc. Tusbar S. "Forecasttng Tomorrow's Trading Day." Technical
11nalysis of Scocks & Commod111es. Vol. 1O. No. 5 (May 1992), pp.
220-223. Copyright 1992 Technical Analysis, lnc. Uscd v.itb permission.
C hande, Tushar S. " Market Tbrust." Technica/ llnalysis of Suxks &
195
196
Bbliography
197
Bblogrphy
lndex
25, 28
and StochRSI measurcs, 124
and stock rounioo, 193
and VTDYA bands, 54, 68
Arms, Richard, 144
Arnu Jodex, ix, 144
See a/so TRIN
Average Dircctional lndcx
(ADX)
and Chande momcntum
oscillator (CMO), 108- 10
and dynamic momentum
index (DMI), 140
B
Balsara, Nauzer J., 163, 178
Blau, William, 108
British Pound Futures, 164-65,
166
e
Cand.lestick Analysis, 73-75
and quanfying candlcstick
shadows, 82- 84, 90
See a/so Quantitative
candlestick
Cbandc Momentum Oscillator
(CMO), 11
fea1ures of, 108- 14, 118
199
200
lnde
CMO (Contmued)
and lhrust oscillator, 151-58
filtcnng tradc noise from,
See a/so Stock market
115-18, 141
Oynamtc Momcotum lndcx
and rclat1ve strcnglh index
(DMI). 12
(RSI), 94-95, 141, 180
and Dow Joncs 1ndustrial
and S&P-SOO index, 97-106
Average, 136-37
and stock rotation, 192-94
and S&P-500 index, 137-39
and treasury bonds, 95-97,
trading strategies with, 14()....41
109- 18
and volaLi li ty, 134-36, 141
nnd trendiness, 104-7, 140
and VIDYA, 58, 60-62, 68-70,
179-89
E
a nd wheDL fu lures, IOS-9
Clinton , Bill, 65
Engineers, 162
Cotree Futures, 25-28, 79-81
Commodity Channel lndex
(CCI)
F
limitations of, 93-94
and similarities among
Failurc Rates of lndicators, 1- 3
indicators, 3, 5- 7, 16-17
Fibonacci RctraccmentS, 176
Commodity Systems, lnc. (CSJ),
Flexible Paramcter Trading
7, 9, 25-27, 181-83, 187
Models, 137
Compctition, 162
Forecast Oscillator, 4D-42
Contingcncy Planning, 13-14,
39-40
Copper Futures, 176-77
G
Can on Fu lures, 74
General Elcctric (GE), 87- 89,
107
o
Gold f'uturcs, 20-23, 174-76
Dcclining lssucs (DI), 143-45
Dcchning Volume (DV), 143-45
Deutscbc Mark ContraCtS, 81 ,
83-84, 90-91
Dircctional Mocmcnt System,
3
Dow Janes Industrial Average
and dynamic momentum
indcx (DMl), 136-37
201
lndex
J
Japancse Yen Furures, 168-69,
174-75
L
Lcverage, 16 1, 178
Linear Regression Analysis, ix
and colfee futures, 25-28
and oorrelation amoog
indicators, 7-9
and dynamic mornentum
index (DMI), 140
and lntel Corporatioo stock,
29-34
and rnarlcet rotatioo, 179, 190,
193-94
and price fo=m, 1O, 20,
34-42
and trcndiness, 106
tutorial, 43-48
use of, 19- 25, 43-48
and VlDYA, 60-63, 70-72
wcaknesses in, 34
N
New Concepts in Technical
Trading Sysiems (Wilder),
97
Noise Filters, 115-18, 13 1, 141
p
Philip Morris Stock, 4-9, 89
Plus Direeonal Movement
(DX+),5, 16
Price Oscillator, 5. 7-8, 17
Price Paneros, 17-18
a,nd limitations of mornentum
oscillators, 93
and nsk control plans, 174
Price Targets, x
lndex
202
Q
Quantitative Candlestick
(Qstick)
calculation o, 90-91
dcfined. ix, 11
and momentum, 75-77, 798 1, 90
a nd 1987 cras h, 77-79, 87
and stock prices, 85-89, 193
a nd trading strategics, 76-77,
84
tutora!, 90-92
R
Ra ndom Prices, Generating of
tutoria l, 17- 18
Rc lati ve SlrCngth Lndcx (RSl)
nnd Amgen stock priccs,
130-31
and Chande momcntum
oscillitor {CMO), 94-95.
14 1, 180
and correla on among
indicators, 9
dcfincd, 119-23
a nd Dow Jones Industrial
Average, 136-37
a nd d)namic mo mc ntum
index {DMl), 134-36, 141
a nd intraday momcntum
i.ndex (!MI), 80-81
s
S&P-500 lndex
and Chande momentum
oscillator (CMO), 104-6
203
lndelC
T
Thrust Oscillaior, 12
compa.red to market lhrust
and T RJN, 149-50, 159
dcfined, 145-49
and Dow Jones Industrial
Average, 150-58
and S&P-500 index, 156
trading stratcgies with, 15658, 193
Trade Template, 169-72, 189,
194
Trading Bands
and candlcstick analysis, 74
and VlDYA, 54-55, 58-59, 63,
67-68
Trcasury Bonds
and candlestick analysis,
82-83
and Chandc momentum
oscillator (CMO), 95-97,
109-18
204
V
Variable lndcx Oynamic Average
(VTDYA)
and advan1ages over s tati c
indica1ors, 49
calculation of, 63-72
and Chandc momcntum
oscillator (CMO), 11- 12, 58,
60-63, 68-70, 179-89
defined, ix, l I
dynamic range of, 52-54, 63
lndex
w
Wheat Futures, 108- 9
Whipsaw Trades, 130, 169, 187
White, Adam, 33
Wilder. J. WeUes, Jr., 29
and calculation of relative
strengtb index (RS1), 97,
121, 122-23
William's %R
See Stochastic oscillator
11
fcontinued lrom ftOl'I flPI
irnpr()'.1! idl>nl1lic.ition
d QllClestick p;lltems.
ponde<
yooout.
In IOday's fterccly compelitive lutu.-es and linancial maii<el,. 'IOl6 success depends on masterir6
fast-~ marl<et informatioo and ~ it IO
maloe s"ifl, smart, strategic ~ and irMst
ment decisoons. Desigrled IO ~ ~'OU harde the
~d ~ and the bad. The New Technical
T1ader IS the proacli\-e, single sourre guide for
pl~ the wirri1g percentages and reapSig ~
ser trad1ng prof'its with minin'a.Jm risk.
Aboot the authors
Tl.JSHAR $. CHANDE is a scie1'llis4 ifl\~lo~ and,
lf you need to reliably project prices under different market conditiorb. estmate risk on new positions, create s~ to cut your losses,
0< plot momentum swing failures, The New Technical Trader is
your fist place to tum. lt's the frontline guide for developing objective consistency in your decisioo making and gaining a valuable
edge in today's commodities, currency, and stock markets.