Description
In this paper we conduct tests for two different
trading rules, namely, the Dual Moving
Average (DMA) model and the Channel
Breakout (CHB) rule. These rules are tested
across five futures contracts the S&P 500,
British Pound, US T-Bonds, COMEX Gold and
Corn using daily data over the period 1990 to
1998. Overwhelmingly, we find that the trading
rules are unable to produce (gross or net)
profits at any statistical level. While positive
gross and net profits were available in four of
the five markets, the profits were neither
economically or statistically significant
Accounting Research Journal
Profitability of Trading Rules in Futures Markets
J ohn Anderson Robert Faff
Article information:
To cite this document:
J ohn Anderson Robert Faff, (2005),"Profitability of Trading Rules in Futures Markets", Accounting Research J ournal, Vol. 18
Iss 2 pp. 83 - 92
Permanent link to this document:http://dx.doi.org/10.1108/10309610580000677
Downloaded on: 24 January 2016, At: 20:59 (PT)
References: this document contains references to 0 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 354 times since 2006*
Users who downloaded this article also downloaded:
Kathleen Herbohn, (2005),"J ob Satisfaction Dimensions in Public Accounting Practice", Accounting Research J ournal, Vol.
18 Iss 2 pp. 63-82http://dx.doi.org/10.1108/10309610580000676
Prodosh Simlai, (2014),"Firm characteristics, distress risk and average stock returns", Accounting Research J ournal, Vol. 27
Iss 2 pp. 101-123http://dx.doi.org/10.1108/ARJ -06-2012-0046
Ping-Sheng Koh, (2005),"Institutional Ownership and Income Smoothing: Australian Evidence", Accounting Research
J ournal, Vol. 18 Iss 2 pp. 93-110http://dx.doi.org/10.1108/10309610580000678
Access to this document was granted through an Emerald subscription provided by emerald-srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service
information about how to choose which publication to write for and submission guidelines are available for all. Please
visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio of
more than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of online
products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on Publication
Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.
*Related content and download information correct at time of download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Profitability of Trading Rules in Futures Markets
83
Profitability of Trading Rules in Futures
Markets
John Anderson
School of Economics and Finance
Queensland University of Technology
and
Robert Faff
Department of Accounting and Finance
Monash University
Abstract
In this paper we conduct tests for two different
trading rules, namely, the Dual Moving
Average (DMA) model and the Channel
Breakout (CHB) rule. These rules are tested
across five futures contracts – the S&P 500,
British Pound, US T-Bonds, COMEX Gold and
Corn using daily data over the period 1990 to
1998. Overwhelmingly, we find that the trading
rules are unable to produce (gross or net)
profits at any statistical level. While positive
gross and net profits were available in four of
the five markets, the profits were neither
economically or statistically significant.
1. Introduction
Speculative trading activities in futures markets
are increasingly undertaken by active hedge
funds and other tactical asset allocation
participants. These trades are often with a
reliance on quantitative trading models
generally falling into the category of being
either momentum-type models, where the
trader aims to take positions consistent with the
direction of the prevailing ‘trend’, or some sort
of overbought/oversold model that aims to
identify periods of short-term disequilibria with
the trader aiming to sell into ‘overbought’
markets and buy into ‘oversold’ markets.
Keywords: Futures Markets, Trading Rules, Technical
Analysis.
Acknowledgement: Special thanks to Robert Brooks of
RMIT for assistance in this research project and also to an
anonymous referee.
Given the substantial literature in the area of
trading rule performance, see inter alia Dale
and Workman (1981), Peterson and Leuthold
(1982), Lukac (1985), Lukac et al. (1988),
Simon (1999) and Merrick (2000), the growth
of speculative trading activity means that it is
important to empirically re-evaluate the
performance of such models periodically to
determine whether existing theories of futures
market equilibrium remain valid. With this in
mind, this paper presents two well-known
momentum type models, namely the Dual
Moving Average Crossover rule and the
Channel Breakout Rule, across a broad range of
markets and asset classes, namely futures
contracts over equities (S&P 500), bonds (US
T-Bonds), currencies (British Pound), soft
commodities (Corn) and precious metals
(COMEX Gold).
The remainder of this paper is structured as
follows. Section 2 provides a brief literature
review, while Section 3 discusses some of the
broad methodological issues that must be
resolved, such as position sizes, transaction
costs and how parameter values for each model
are selected. Section 4 provides a description of
the models and how the Buy/Sell decisions
are made. Section 5 presents the results in
terms of profitability. Conclusions are drawn in
Section 6.
2. Brief Literature Review
Dale and Workman (1981) studied Treasury
Bill futures to see whether the number of price
trend reversals agreed with the number
predicted under a random walk model using a
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 18 NO 2 (2005)
84
variation of a serial correlation technique. The
dataset examined was daily Treasury Bill
futures price data between 1974 and 1978. The
study concluded that inter alia the number of
trend reversals observed in the price series did
not concur with the random walk model.
Further they found that the effect of
commissions would negate the economic value
of applying a mechanical trading rule during
the test period.
Peterson and Leuthold (1982) examined the
use of filter-rules in the Hog futures market
between 1973 and 1977. Twenty filters were
tested using percentage filters between 1% and
10% and dollar-based filters between 50c and
$5. Their results indicate that all filter levels
provided a Mean Gross Profit per trade (MGP)
of greater than zero. Lukac et al. (1988),
presents the results for twelve technical trading
models for twelve different futures markets for
the years 1975 to 1984. Transaction costs set at
$100 per round-turn trade were deducted to
more accurately reflect brokerage plus poor
order execution. The key finding from Lukac’s
study is that seven of the twelve models tested
produced positive average monthly returns of
which some results are statistically significant
at the 0.10 level and below.
Another avenue explored by researchers into
the efficiency of futures markets is in the area
of spreads between either markets (inter-
commodity spreads) or between contract
expiries in the same instrument (intra-
commodity spreads). These studies have
generally found significant profits in defiance
of theoretical predictions. An example of work
in this area is Poitras (1987) who examined the
cost-of-carry relationship between Gold and
Eurodollar futures. He found that the observed
cost-of-carry relationship was weakly bounded
and that gold futures occasionally violated this
cost-of-carry relationship.
Simon (1999) reported transitory deviations
in the long-run equilibrium for the soybean
market via the ‘Crush Spread’ between January
1985 and February 1995. This spread involves
trading between the relationships that should
exist between soybeans and their derivative
products, soybean mash and soybean oil. By
determining whether the crush spread has a
cointegrating relationship within a GARCH
framework and a reversion process to a 5-day
mean, he found that a reversion process was
present and profitable trading opportunities
were present. Similarly, interesting price
change behaviour has been reported in the
spread between interest rate futures of different
expiries, referred to as ‘Calendar Spreads’.
Merrick (2000) examined the interest rate
futures traded on the London International
Financial Futures Exchange (LIFFE) between
1989 and 1998. He found that deviations
existed between the high-order polynomial
approximation representation of the term
structure of interest rates and those rates
observed on the LIFFE’s short-term interest
rate contract. He did however note that
transactions costs on futures contracts had
lowered substantially during the test period and
suggests that this may have resulted in a
marked reduction in the observed deviations.
3. Methodological Issues
3.1 Position Sizes
The analysis conducted in this paper utilises
position sizes of one futures contract only per
trading signal. Therefore each Buy/Sell signal
results in only one Long/Short position being
opened. This maintains simplicity in the results
by having a consistent one position per contract
taken, regardless of dollar-value sensitivities
and so on that can be significant between
different contracts. This approach is consistent
with the empirical literature in examining
market efficiency in futures markets and is
adopted by inter alia Peterson and Leuthold
(1982), Lukac et al. (1988), Taylor (1992) and
Taylor (1993).
To avoid potentially large positions being
taken, only one contract has been permitted to
be open at any time. Therefore, if a Buy signal
at time t is generated by the trading rule
followed by another Buy signal at t+n days, the
subsequent signals are ignored until the
position is either closed out or an opposing
signal is generated.
Adopting these approaches to position sizing
provides several benefits. It makes the returns
from the trading rules easier to compare rather
than trying to match dollar values between
different contracts. When examining trading
rules that take relatively longer-term positions,
it avoids potential liquidity problems where
consecutive Buy/Sell conditions may persist for
several months leading to large position
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Profitability of Trading Rules in Futures Markets
85
accumulations that may be difficult to liquidate
in a timely manner.
3.2 Transaction Costs
1
We assume transaction costs of US$100
‘round-turn’, which means that a transaction
cost of US$50 is deducted on trade entry and
an additional US$50 on trade exit. This figure
is consistent with the literature where
comparable amounts were also used by Lukac
et al. (1988) and Taylor (1993). Without
explicit deductions for transaction costs, tests
could arguably bias performance towards
trading rules that trade frequently with only
small profits per trade. Therefore the trading
rule parameters perceived as less profitable
due to their longer term trading characteristics
may have actually provided greater profitability
once brokerage charges were deducted.
Another method for assessing the impact of
transaction costs is to quote some typical retail
brokerage rate of US$25, for example, as was
used in Thomas III (1986). This approach
raises considerable concern as to the expected
trade execution price. While the allowance for
brokerage charges may be realistic, this amount
would fail to take poor order execution into
account where trades may actually occur
several points away from where anticipated –
presumably at a price not advantageous to the
client. Consequently, simulated profits may be
significantly overstated compared to the results
observed if the trading had actually occurred in
a real-time market.
Therefore, the allowance of transaction costs
of US$100 per round-turn transaction should
provide a more realistic assessment of how the
trading rules would have performed in practice.
It allows for brokerage and also the fact that
poor order execution may occur. If the
brokerage allowance has been set too high in
this study to better reflect real trading
performance it should, at worst, lead to a more
1 It should be noted that execution costs have been shown
to be substantial in the microstructure literature. While
our research method assumes a small transaction size, in
reality large traders would be subject to execution and
liquidity effects. This biases our findings toward
overstating the profitability of trading rules, however, as
shown later, the trading rules perform poorly despite
this effect. We thank an anonymous referee for pointing
this out.
conservative assessment of reported trading
rule profitability.
3.3 Transaction Timing
All transactions are assumed to have been
undertaken at the closing price on the day on
which the trading signal was generated. This
has the presumption that the trader would be
aware on the day the trading signal was
generated at what price level a trade would
need to be executed and any closing price
beyond the Buy/Sell threshold for the trading
rule would be readily identifiable in time for
the order to be placed.
The market microstructure literature has
documented a characteristic ‘U-Shape’ pattern
for both volatility and trading volumes in
numerous markets. Examples include
Lauterbach and Monroe (1989) with respect to
Gold futures, Abhyankar et al. (1997) in the
UK equities market and Daigler (1997) in
futures contracts over the S&P 500, MMI stock
index and US T-Bonds. Given that price
volatility and trading volumes tend to be higher
in the first and last half-hour periods of the
trading sessions, it would be reasonable to
assume that the trader would find sufficient
liquidity to enter/exit a futures position at these
times.
It is conceded that a position being executed
at the exact closing price for all trades would be
unrealistic. Therefore, under the assumption
that price movements are largely random it is
further assumed that some trades would be
executed at a price favourable and
unfavourable to the trader with equal likelihood
so producing a ‘swings-and-roundabouts’ effect
on trade entry and exit. Any minor and
persistent bias towards unfavourable trade
entry/exit should be adequately accounted for
in the considerable allowance for transaction
costs.
3.4 Parameter Selection and Optimisation
Methods
Parameters for the trading models are based on
a periodic optimisation process. The
optimisation is conducted by testing a range of
parameters and selecting those parameter
values for which the highest profitability was
reported in the immediately prior year. That is,
to avoid any problems associated with ‘in
sample’ testing and reporting simulated profits
derived from perfect hindsight, all reported
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 18 NO 2 (2005)
86
profitability is obtained from tests conducted
‘out-of-sample’. This is achieved by
conducting a ‘walk-forward’ testing procedure
as follows. In the first year of our dataset, 1989,
we test all parameters in the testing range to
determine those that produce maximum profit.
These selected parameters are then applied into
the 1990 dataset and the profits recorded. We
then roll forward one year at a time, repeating
this procedure. Adopting this walk-forward
approach to the testing helps to ensure that the
trading rule parameters selected are comparable
to the information set that would have been
available to the trader operating in the
market at that time. Neftci (1991) persuasively
argued that for trading rules to be truly testable,
the Buy/Sell decision criteria must be
mathematically pre-defined given an
information set available at time t.
3.5 Stop-Loss Provisions
Stop-Loss mechanisms essentially aim to
reduce the losses from any trading decision that
proves to be unprofitable ab initio by exiting a
trade once a pre-determined maximum loss
occurs after the trade is opened. The stop-loss
technique could be demonstrated by an
example of a trader using a trading system that
produces a buy signal at 93.50. The trader buys
a futures contract at 93.50, but also adds a stop-
loss order of 8 points. The broker would then
be instructed to sell/liquidate the position as
soon as the market trades at ? 93.42. Therefore
if the trade is unprofitable from the outset the
trader should be able to set a maximum loss of
8 points representing the largest dollar amount
the trader is willing to risk/lose on that
individual trade.
2
2 One drawback of this stop-loss technique with daily
data is that by using closing prices only, the following
day’s closing price may be significantly below the 93.42
at which the position was to be liquidated. For example,
if a bought position was taken at 93.50 and an 8 point
stop-loss was used, the closing price of the following
day may be at 93.20. As the trading model has
previously specified that the position should be closed
out at ? 93.42, a sell order would be executed - but at
93.20. This results in an expected maximum loss of 8
points becoming a realised 30 point loss. Therefore the
impact on results can see losses on individual
transactions considerably higher than the stop-loss
might suggest which may make results more volatile
than initially anticipated.
The stop-loss mechanisms built into the test
models used in this paper are only used to
avoid large losses occurring while waiting for
an opposing or liquidating signal. This simple
stop-loss mechanism provides an
uncomplicated method of exiting trades that are
unprofitable shortly after entry. Stop-loss
values ranging between $100 and $1,000 (in
$100 increments) have been tested against each
different set of parameters for the trading
systems examined.
4. Trading Rules
4.1 Dual Moving Average Crossover
The Dual Moving Average (DMA) crossover
rule relies on two moving averages as part of its
Buy/Sell signal generating process. Its
description has been well documented in the
practitioner literature by authors including
Babcock (1989) and Bernstein (1995). The
Dual Moving Average Crossover aims to
identify changes in trend where short-term
‘noise’, or erratic price movements, are
removed by the smoothing effect of the moving
average’s derivation of the mean price for the
previous n days. When the two moving
averages used in the calculations have both
changed direction, for example from down to
up, the trend is regarded as having changed
direction. The trader will buy (sell) in
expectation that the new trend indicated by the
increasing (decreasing) value of the two
moving averages would continue. In essence,
this is a short-term momentum strategy.
The method of calculating the moving
averages uses the equations based on that
provided by Neftci (1991). These are defined
here in equations (1) and (2).
!
"
#
$
%
&
=
'
?
=
?
1
0
1
m
S
S t m
P
m
MA
(1)
!
"
#
$
%
&
=
'
?
=
?
1
0
1
n
S
S t n
P
n
MA
(2)
where MA
m
(MA
m
) is a moving average of
length m
; and P
t
the price of a security at
time t.
The basic rationale of the DMA model is
that the short-term fluctuations in the price data
are smoothed making the longer-term trends
more readily identifiable. When the shorter
moving average MA
m
crosses above the longer
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Profitability of Trading Rules in Futures Markets
87
moving average MA
n
from below (above), it
implies that the short-term trend is for prices to
increase (fall) and so generating a buy (sell)
signal accordingly. MA
m
(MA
n
) has values
tested commencing with a 2-day (5-day)
moving average to a 20-day (25-day) moving
average at 2-day (5-day) increments.
4.2 Channel Breakout Rule
This model is sourced from Babcock (1989)
and is one of the simpler trading rules
described in the market literature. The Channel
Breakout system aims to identify when the
price moves out of a trading range in
expectation that the upward (downward) move
will continue. A move above (below) a
previously established trading range is taken as
a buy (sell) signal.
A Channel is said to be of n days length,
where n is specified by the user. The Breakout
is regarded as the closing price on the most
recent trading day being at a higher high (lower
low) than any other day in the n day channel.
When the market makes the n day high (low),
the trader takes a bought (sold) position in the
expectation of that breakout continuing to form
a sustained up (down) trend. The bought (sold)
position is held until the criteria for taking an
opposing sold (bought) position are met, as
described above, or the stop-loss is triggered.
The parameters tested for this system are from
a 5-day Channel to a 25-day Channel
(incrementing one step at a time) with stop-
losses from $100 to $1,000 (incrementing $100
at a time).
5. Results
5.1 Dual Moving Average Model
3
Panel A of Table 1 reveals the performance of
the Dual Moving Average trading rules when
applied to the S&P 500 futures contract during
the test period. At the gross profit level, only
two years were marginally profitable, with the
worst performing year being 1997 with gross
losses of $64,238 reported. Once transaction
costs were deducted the losses became more
severe with the maximum loss of $70,538
3 As this model requires optimisation, the parameters for
1990 were obtained by optimising the parameters on the
data from 1989 and the values applied in 1990 represent
those parameters producing the highest gross profits in
the previous year.
being reported in 1997. For some years the
maximum drawdown observed was relatively
high, for example $67,400 and $51,150 for
1997 and 1998, respectively. Therefore, it
appears that not only were large losses
observed but also the funding costs were high
given that only one contract was being traded at
any given time highlighting the poor
performance of the DMA model in the S&P
500 futures market during the test period. The
lengths of the moving averages could be
described as highly erratic. They ranged from
quite large values, such as MA
m
= 20 and MA
n
= 25 in 1991, through to quite small values,
such as MA
m
= 2 and MA
n
= 15 in 1998. This
would provide little comfort to the trader trying
to identify a priori which parameters to use and
suggests considerable instability in the returns
from the futures contract across time.
Panel B presents the results for the DMA
model when applied in the US T-Bond futures
market and again, few years provide profits.
Other performance characteristics, such as
maximum drawdown were considerably less
severe than that observed in the S&P 500.
Although the drawdown was lower, the
percentage of profitable trades was generally
well below that expected by chance.
Furthermore, as was the case with the DMA
results for the S&P 500, the parameters appear
to be equally unstable for the US T-Bond
futures with no real preference for longer or
shorter moving average lengths across time.
Panel C presents the profit results for the
British Pound. As with previous studies in
foreign exchange markets (for example Taylor,
1992), gross and net profits were available to
speculators in British Pound futures during the
test period. Specifically, net profits were
available in five of the nine years examined
with 1992 providing the highest net profit of
$18,888. The maximum drawdown figures
were comparable to those observed in the US
T-Bond market and parameter instability was
evident during the test period. Panels D and E
report results for the COMEX gold and corn
contracts. These results largely mirror those
discussed above.
Table 2 presents the statistical significance
of the reported gross and net profits for all five
contracts. Generally, it reveals that while
aggregate gross profits were available in three
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 18 NO 2 (2005)
88
Table 1
Dual Moving Average Profit Results
1990 1991 1992 1993 1994 1995 1996 1997 1998
Panel A: S&P 500
Parameter Values 2/10:300 10/25:900 16/25:100 10/20:100 4/5:1000 12/15:800 8/20:300 4/5:300 10/15:200
Number of Trades 38 12 9 21 63 18 16 63 20
Percent Profitable 18% 8% 11% 10% 38% 17% 0% 16% 15%
Max Drawdown 7,575 10,388 1,450 7,775 9,238 8,750 12,550 67,400 51,150
Gross Profit 2,738 -1,988 -413 -7,625 -6,175 425 -12,550 -64,238 -34,925
Net Profit -1,062 -3,188 -1,313 -9,725 -12,475 -1,375 -14,150 -70,538 -36,925
Panel B: US T-Bonds
Parameter Values 16/20:600 20/25:1000 8/25:900 10/16:300 2/5:600 18/20:1000 2/25:300 4/25:100 2/15:200
Number of Trades 19 12 9 18 52 19 16 12 20
Percent Profitable 37% 25% 0% 33% 38% 37% 25% 0% 10%
Maximum Drawdown 8,950 5,531 7,363 2,119 5,919 7,375 5,938 4,157 5,519
Gross Profit -2,200 3,563 -7,363 3,350 -4,925 -719 8,168 -3,969 -3,088
Net Profit -4,100 2,363 -8,263 1,550 -10,125 -2,619 6,568 -5,169 -5,088
Panel C: British Pound
Parameter Values 4/10:700 16/25:1000 4/25:800 2/25:200 12/15:400 10/15:1000 20/25:800 12/25:800 4/15:600
Number of Trades 22 10 5 26 16 23 18 14 23
Percent Profitable 36.0% 30.0% 60.0% 15.0% 38.0% 30.0% 39.0% 14.0% 17.0%
Maximum Drawdown 3,188 8,188 1,688 7,513 2,013 11,525 4,300 6,350 7,775
Gross Profit 18,888 11,675 19,388 -1,575 4,075 -9,025 3,363 -5,625 -7,450
Net Profit 16,688 10,675 18,888 -4,175 2,475 -11,325 1,563 -7,025 -9,750
Panel D: COMEX Gold
Parameter Values 8/20:100 12/15:700 4/10:300 10/25:600 16/25:100 4/10:700 18/20:800 10/15:100 2/5:800
Number of Trades 12 18 29 6 8 30 15 15 56
Percent Profitable 8% 50% 28% 50% 0% 23% 53% 27% 30.0%
Maximum Drawdown 3,010 1,520 2,400 1,410 1,830 4,260 1,140 500 6,810
Gross Profit -1,870 1,850 910 6,970 -1,810 -4,120 2,430 5,270 -6,560
Net Profit -3,070 50 -1,990 6,370 -2,610 -7,120 930 3,770 -12,160
Panel E: Corn
Parameter Values 4/5:400 14/25:100 2/10:100 18/25:300 6/15:200 14/15:600 8/10:200 2/25:300 8/20:400
Number of Trades 62 16 27 9 21 21 29 9 14
Percent Profitable 50% 0% 30% 44% 24% 43% 28% 22% 50%
Maximum Drawdown 2,113 1,738 950 1,225 1,738 3,150 2,988 1,238 700
Gross Profit 388 -1,738 350 1,238 -438 450 1,863 -388 3,638
Net Profit -5,812 -3,338 -2,350 338 -2,538 -1,650 -1,037 -1,288 2,238
from five markets, namely the British Pound,
Gold and Corn futures markets, none proved to
be statistically significant at any recognised
levels. Once transaction costs were
incorporated only one market was able to
produce positive net profits, namely British
Pound futures. However, the net profitability
for the British Pound failed to be statistical
significantly different from the zero.
5.2 Channel Breakout Rule
Table 3, Panel A shows the profitability of the
Channel Breakout Rule (CHB) rule when
applied to the S&P 500 futures contract during
the test period. The CHB produced inconsistent
profitability with minor profits and losses being
reported between 1990 and 1996. The dramatic
increase in profitability reported for 1997 and
1998 provided the most profitable years of the
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Profitability of Trading Rules in Futures Markets
89
Table 2
Dual Moving Average Statistical Results
Panel A: S&P 500
Gross Profits -124,751 Net Profits -150,751
t-statistic -0.096 t-statistic -0.062
p-value 0.466 p-value 0.478
Panel B: US T-Bonds
Gross Profits -7,183 Net Profits -24,883
t-statistic -0.64 t-statistic -0.16
p-value 0.29 p-value 0.44
Panel C: British Pound
Gross Profits 33,714 Net Profits 18,014
t-statistic 0.33 t-statistic 0.61
p-value 0.39 p-value 0.30
Panel D: COMEX Gold
Gross Profits 3,070 Net Profits -15,830
t-statistic 0.82 t-statistic -0.37
p-value 0.25 p-value 0.37
Panel E: Corn
Gross Profits 5,363 Net Profits -15,437
t-statistic 0.28 t-statistic -0.05
p-value 0.40 p-value 0.48
test period returning $21,013 and $35,925,
respectively. When considering the broader
performance characteristics, the CHB rule
performed particularly poorly during the test
period. Although the maximum drawdown was
considerably less than that observed for the
DMA rule in the same market, the percentage
of profitable trades was remarkably poor with
the highest result of 29% recorded for both
1991 and 1998. Of the 18 trades simulated
between 1992 and 1994, not one trade was
profitable. It is however curious that even with
a percentage of profitable trades of only 17% in
1997 the model was still able to produce profits
after transaction costs.
Panels B to E present the results for the
CHB for US T-Bonds, British Pound, gold and
corn, respectively. Their results are broadly
similar to the S&P and so only brief further
discussion is made. For example, when
examining the profitability of the CHB in US
T-Bonds it is apparent that the trading rule was
able to produce net profits in six of the nine
years tested, with the highest net profit being
$12,075 (1995). In the case of the British
Pound very mixed profit results are reported
with net losses reported in five of the nine years
examined, though one large net profit spike of
$31,825 appeared in 1992. For the COMEX
Gold futures, one very striking observation is
the very low percentage of profitable trades -
zero percent of trades were profitable in all but
three years from nine examined.
Table 4 presents the statistical significance
of the reported gross and net profits for all five
contracts. Despite the simplicity of the Channel
Breakout trading rule it was able to produce
profits in most markets, even after allowances
for transaction costs, but not at any recognised
significance level.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 18 NO 2 (2005)
90
Table 3
Channel Breakout Rule Profit Results
1990 1991 1992 1993 1994 1995 1996 1997 1998
Panel A: S&P 500
Parameter Values 18:100 6:100 10:100 22:100 22:100 7:300 18:800 6:100 14:100
Number Trades 6 7 9 4 5 9 7 6 7
Percent Profitable 17% 29% 0% 0% 0% 11% 0% 17% 29%
Maximum Drawdown 3,375 1,700 6,413 950 1,875 3,300 5,663 9,275 3,125
Gross Profit -3,000 3,325 -6,013 -900 -1,675 -3,238 -5,663 21,613 36,625
Net Profit -3,600 2,625 -6,913 -1,300 -2,175 -4,138 -6,363 21,013 35,925
Panel B: US T-Bonds
Parameter Values 22:200 21:600 5:300 15:600 13:700 18:1000 6:300 7:800 12:800
Number Trades 6 7 10 3 5 3 9 5 2
Percent Profitable 0% 43% 10% 33% 20% 67% 0% 40% 50%
Maximum Drawdown 2,531 2,231 3,501 1,756 5,094 1,531 4,381 1,600 594
Gross Profit -2,531 3,419 3,438 9,337 144 12,375 -4,194 4,350 6,344
Net Profit -3,131 2,719 2,438 9,037 -356 12,075 -5,094 3,850 6,144
Panel C: British Pound
Parameter Values 5:400 22:100 5:1000 6:900 14:100 8:500 18:100 13:600 6:1000
Number Trades 7 12 5 9 5 6 4 6 9
Percent Profitable 29% 8% 80% 0% 20% 0% 25% 17% 11%
Maximum Drawdown 2,188 5,825 2,175 8,563 1,163 4,125 500 3,125 5,000
Gross Profit 14,325 -3,463 32,325 -8,563 1,238 -3,750 8,138 -1,800 -4,975
Net Profit 13,625 -4,663 31,825 -9,463 738 -4,350 7,738 -2,400 -5,875
Panel D: Gold
Parameter Values 5:900 17:100 24:100 7:600 10:1000 17:100 25:800 5:900 14:100
Number of Trades 7 5 5 2 2 1 1 10 10
Percent Profitable 0% 0% 0% 50% 0% 0% 100% 20% 0%
Maximum Drawdown 6,500 520 660 740 1,450 330 450 3,370 1,400
Gross Profit -6,500 -480 -660 4,640 -1,210 -300 3,450 -1,300 -1,330
Net Profit -7,200 -980 -1,160 4,440 -1,410 -400 3,350 -2,300 -2,330
Panel E: Corn
Parameter Values 18:100 7:200 23:800 17:300 12:1000 9:200 24:700 8:1000 5:800
Number of Trades 6 7 2 4 3 5 3 6 7
Percent Profitable 0% 29% 50% 50% 33% 40% 67% 33% 57%
Maximum Drawdown 875 1,625 700 788 1,513 900 1,188 1,238 1,600
Gross Profit -863 -800 2,538 1,588 363 2,800 6,300 -525 1,525
Net Profit -1,463 -1,500 2,338 1,188 63 2,300 6,000 -1,125 825
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Profitability of Trading Rules in Futures Markets
91
Table 4
Channel Breakout Rule Statistical Results
Panel A: S&P 500
Gross Profits 41,074 Net Profits 35,074
t-statistic 0.38 t-statistic 0.45
p-value 0.37 p-value 0.35
Panel B: US T-Bonds
Gross Profits 32,682 Net Profits 27,682
t-statistic 0.08 t-statistic 0.13
p-value 0.47 p-value 0.45
Panel C: British Pound
Gross Profits 33,475 Net Profits 27,175
t-statistic 0.41 t-statistic 0.50
p-value 0.36 p-value 0.33
Panel D: COMEX Gold
Gross Profits -3,690 Net Profits -7,990
t-statistic -0.71 t-statistic -0.45
p-value 0.28 p-value 0.35
Panel E: Corn
Gross Profits 12,926 Net Profits 8,626
t-statistic 0.10 t-statistic 0.27
p-value 0.47 p-value 0.41
5.3 Discussion
A couple of brief comments are worth making
at this stage. Our findings reveal that neither
rule is able to deliver significant profitable
outcomes. This is notable since, given the fact
that we have ignored execution costs, as
indicated earlier; such a simplification induces
a bias towards overstatement of the success of
the trading rules. Further, the fact that we use
daily data may also impact our analysis. For
example, as highlighted earlier daily data
present an issue regarding the stop-loss
technique in that an expected maximum loss
may understate the magnitude of the actual
realised loss. As such, the impact on results can
see losses significantly higher than the stop-
loss might suggest which may make results
more volatile than initially anticipated. A final
observation relates to a possible learning (or
maturity) effect in our results. For example,
applying the Channel Breakout rule to the corn
contract is suggestive of a trend upwards in the
percentage of profitable trades, with 0%
profitable in 1990 to 57% in 1998. However,
generally the findings are quite mixed in this
regard, thereby preventing a strong conclusion
to be drawn.
6. Conclusion
This paper has presented tests for two different
trading rules, namely, the Dual Moving
Average (DMA) model and the Channel
Breakout (CHB) rule. These rules were tested
across five markets, those being the futures
contracts over the S&P 500, British Pound, US
T-Bonds, COMEX Gold and Corn using daily
data over the period 1990 to 1998.
Despite their supposed profitability reflected
by their persistent appearance in the literature,
they failed to produce profits at any statistical
level during the test period on the five markets
considered. In the case of the Dual Moving
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 18 NO 2 (2005)
92
Average (DMA) trading rules, gross profits
were reported in three from five markets, but
this fell to one from five once transaction costs
were incorporated. Neither the gross nor net
profits were statistically significant for any of
the markets examined. Similarly, the CHB rule
failed to produce statistically significant profits
for any of the markets examined. While gross
and net profits were available in four from five
markets, the profits were not significantly
different from zero. These findings are
consistent with market efficiency.
The parameter values derived during the
optimisation process remained generally highly
unstable for their respective models. Whether
this instability reflects some form of non-
stationarity of returns, time-varying volatility
characteristics (for example possibly consistent
with ARCH/GARCH effects) or some other
econometric property remains to be determined
and are offered as possible explanations of the
parameter instability issue.
References
Abhyankar, A., Ghosh, D., Levin E. and Limmack R.J.
(1997), ‘Bid-Ask Spreads, Trading Volume and
Volatility: Intra-Day Evidence from the London Stock
Exchange’, Journal of Business Finance &
Accounting, vol. 24, April, pp. 343–362.
Babcock, B.J. (1989), The Dow-Jones Irwin Guide To
Trading Systems, Richard D. Irwin Inc., Illinois USA.
Bernstein, J. (1995), The Compleat Day Trader, McGraw-
Hill, NewYork, USA.
Daigler, R.T. (1997), ‘Intraday Futures Volatility and
Theories of Market Behavior’, Journal of Futures
Markets, vol. 17(1), pp. 45–74.
Dale, C. and Workman, R. (1981), ‘Patterns of Price
Movements in Treasury Bill Futures’, Journal of
Economics and Business, vol. 33, pp. 81–87.
Lauterbach, B. and Monroe, M. (1989), ‘Evidence on the
Effect of Information and Noise Trading on Intraday
Gold Futures Returns’, Journal of Futures Markets,
vol. 9(4), pp. 297-305.
Louis P., Lukac, B., Brorsen, W. and Irwin, S.H. (1988), ‘A
Test of Futures Market Disequilibrium Using Twelve
Different Technical Trading Systems’, Applied
Economics, vol. 20, pp. 623–639.
Merrick, J.J. (2000), ‘Pascal Spreading of Short-Term
Interest Rate Contracts’, Journal of Futures Markets,
vol. 20(10), pp. 889–910.
Neftci, S.H. (1991), ‘Naive Trading Rules in Financial
Markets and Wiener-Kolmogorov Prediction Theory:
A Study of “Technical Analysis”’, Journal of Business,
vol. 64(4), pp. 549–571.
Poitras, G. (1987), ‘Golden Turtle Tracks: In Search of
Unexploited Profits in Gold Spreads’, Journal of
Futures Markets, vol. 7(4), pp. 397–412.
Simon, D.P. (1999), ‘The Soybean Crush Spread: Empirical
Evidence and Trading Strategies’, Journal of Futures
Markets, vol. 19(3), pp. 271–289.
Taylor, S. J. (1992), ‘Rewards Available to Currency
Futures Speculators: Compensation for Risk or
Inefficient Pricing?’, in Stein, J.L. and Goss, B.A.
(eds.), Supplement to the Economic Record, Special
Issues on Futures Markets–Price Discovery and Price
Determination.
Taylor, S. J. (1993), ‘Trading Futures Using a Channel
Rule: A Study of the Predictive Power of Technical
Analysis with Currency Examples’, Manuscript
received prior to publication in The Journal of Futures
Markets.
Thomas III, L.R. (1986), ‘Random Walk Profits in
Currency Futures Trading’, Journal of Futures
Markets, vol. 6(1), pp. 109–125.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
This article has been cited by:
1. Shan Wang, Zhi-Qiang Jiang, Sai-Ping Li, Wei-Xing Zhou. 2015. Testing the performance
of technical trading rules in the Chinese markets based on superior predictive test. Physica A:
Statistical Mechanics and its Applications 439, 114-123. [CrossRef]
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_749335157.pdf
In this paper we conduct tests for two different
trading rules, namely, the Dual Moving
Average (DMA) model and the Channel
Breakout (CHB) rule. These rules are tested
across five futures contracts the S&P 500,
British Pound, US T-Bonds, COMEX Gold and
Corn using daily data over the period 1990 to
1998. Overwhelmingly, we find that the trading
rules are unable to produce (gross or net)
profits at any statistical level. While positive
gross and net profits were available in four of
the five markets, the profits were neither
economically or statistically significant
Accounting Research Journal
Profitability of Trading Rules in Futures Markets
J ohn Anderson Robert Faff
Article information:
To cite this document:
J ohn Anderson Robert Faff, (2005),"Profitability of Trading Rules in Futures Markets", Accounting Research J ournal, Vol. 18
Iss 2 pp. 83 - 92
Permanent link to this document:http://dx.doi.org/10.1108/10309610580000677
Downloaded on: 24 January 2016, At: 20:59 (PT)
References: this document contains references to 0 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 354 times since 2006*
Users who downloaded this article also downloaded:
Kathleen Herbohn, (2005),"J ob Satisfaction Dimensions in Public Accounting Practice", Accounting Research J ournal, Vol.
18 Iss 2 pp. 63-82http://dx.doi.org/10.1108/10309610580000676
Prodosh Simlai, (2014),"Firm characteristics, distress risk and average stock returns", Accounting Research J ournal, Vol. 27
Iss 2 pp. 101-123http://dx.doi.org/10.1108/ARJ -06-2012-0046
Ping-Sheng Koh, (2005),"Institutional Ownership and Income Smoothing: Australian Evidence", Accounting Research
J ournal, Vol. 18 Iss 2 pp. 93-110http://dx.doi.org/10.1108/10309610580000678
Access to this document was granted through an Emerald subscription provided by emerald-srm:115632 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service
information about how to choose which publication to write for and submission guidelines are available for all. Please
visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio of
more than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of online
products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on Publication
Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.
*Related content and download information correct at time of download.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Profitability of Trading Rules in Futures Markets
83
Profitability of Trading Rules in Futures
Markets
John Anderson
School of Economics and Finance
Queensland University of Technology
and
Robert Faff
Department of Accounting and Finance
Monash University
Abstract
In this paper we conduct tests for two different
trading rules, namely, the Dual Moving
Average (DMA) model and the Channel
Breakout (CHB) rule. These rules are tested
across five futures contracts – the S&P 500,
British Pound, US T-Bonds, COMEX Gold and
Corn using daily data over the period 1990 to
1998. Overwhelmingly, we find that the trading
rules are unable to produce (gross or net)
profits at any statistical level. While positive
gross and net profits were available in four of
the five markets, the profits were neither
economically or statistically significant.
1. Introduction
Speculative trading activities in futures markets
are increasingly undertaken by active hedge
funds and other tactical asset allocation
participants. These trades are often with a
reliance on quantitative trading models
generally falling into the category of being
either momentum-type models, where the
trader aims to take positions consistent with the
direction of the prevailing ‘trend’, or some sort
of overbought/oversold model that aims to
identify periods of short-term disequilibria with
the trader aiming to sell into ‘overbought’
markets and buy into ‘oversold’ markets.
Keywords: Futures Markets, Trading Rules, Technical
Analysis.
Acknowledgement: Special thanks to Robert Brooks of
RMIT for assistance in this research project and also to an
anonymous referee.
Given the substantial literature in the area of
trading rule performance, see inter alia Dale
and Workman (1981), Peterson and Leuthold
(1982), Lukac (1985), Lukac et al. (1988),
Simon (1999) and Merrick (2000), the growth
of speculative trading activity means that it is
important to empirically re-evaluate the
performance of such models periodically to
determine whether existing theories of futures
market equilibrium remain valid. With this in
mind, this paper presents two well-known
momentum type models, namely the Dual
Moving Average Crossover rule and the
Channel Breakout Rule, across a broad range of
markets and asset classes, namely futures
contracts over equities (S&P 500), bonds (US
T-Bonds), currencies (British Pound), soft
commodities (Corn) and precious metals
(COMEX Gold).
The remainder of this paper is structured as
follows. Section 2 provides a brief literature
review, while Section 3 discusses some of the
broad methodological issues that must be
resolved, such as position sizes, transaction
costs and how parameter values for each model
are selected. Section 4 provides a description of
the models and how the Buy/Sell decisions
are made. Section 5 presents the results in
terms of profitability. Conclusions are drawn in
Section 6.
2. Brief Literature Review
Dale and Workman (1981) studied Treasury
Bill futures to see whether the number of price
trend reversals agreed with the number
predicted under a random walk model using a
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 18 NO 2 (2005)
84
variation of a serial correlation technique. The
dataset examined was daily Treasury Bill
futures price data between 1974 and 1978. The
study concluded that inter alia the number of
trend reversals observed in the price series did
not concur with the random walk model.
Further they found that the effect of
commissions would negate the economic value
of applying a mechanical trading rule during
the test period.
Peterson and Leuthold (1982) examined the
use of filter-rules in the Hog futures market
between 1973 and 1977. Twenty filters were
tested using percentage filters between 1% and
10% and dollar-based filters between 50c and
$5. Their results indicate that all filter levels
provided a Mean Gross Profit per trade (MGP)
of greater than zero. Lukac et al. (1988),
presents the results for twelve technical trading
models for twelve different futures markets for
the years 1975 to 1984. Transaction costs set at
$100 per round-turn trade were deducted to
more accurately reflect brokerage plus poor
order execution. The key finding from Lukac’s
study is that seven of the twelve models tested
produced positive average monthly returns of
which some results are statistically significant
at the 0.10 level and below.
Another avenue explored by researchers into
the efficiency of futures markets is in the area
of spreads between either markets (inter-
commodity spreads) or between contract
expiries in the same instrument (intra-
commodity spreads). These studies have
generally found significant profits in defiance
of theoretical predictions. An example of work
in this area is Poitras (1987) who examined the
cost-of-carry relationship between Gold and
Eurodollar futures. He found that the observed
cost-of-carry relationship was weakly bounded
and that gold futures occasionally violated this
cost-of-carry relationship.
Simon (1999) reported transitory deviations
in the long-run equilibrium for the soybean
market via the ‘Crush Spread’ between January
1985 and February 1995. This spread involves
trading between the relationships that should
exist between soybeans and their derivative
products, soybean mash and soybean oil. By
determining whether the crush spread has a
cointegrating relationship within a GARCH
framework and a reversion process to a 5-day
mean, he found that a reversion process was
present and profitable trading opportunities
were present. Similarly, interesting price
change behaviour has been reported in the
spread between interest rate futures of different
expiries, referred to as ‘Calendar Spreads’.
Merrick (2000) examined the interest rate
futures traded on the London International
Financial Futures Exchange (LIFFE) between
1989 and 1998. He found that deviations
existed between the high-order polynomial
approximation representation of the term
structure of interest rates and those rates
observed on the LIFFE’s short-term interest
rate contract. He did however note that
transactions costs on futures contracts had
lowered substantially during the test period and
suggests that this may have resulted in a
marked reduction in the observed deviations.
3. Methodological Issues
3.1 Position Sizes
The analysis conducted in this paper utilises
position sizes of one futures contract only per
trading signal. Therefore each Buy/Sell signal
results in only one Long/Short position being
opened. This maintains simplicity in the results
by having a consistent one position per contract
taken, regardless of dollar-value sensitivities
and so on that can be significant between
different contracts. This approach is consistent
with the empirical literature in examining
market efficiency in futures markets and is
adopted by inter alia Peterson and Leuthold
(1982), Lukac et al. (1988), Taylor (1992) and
Taylor (1993).
To avoid potentially large positions being
taken, only one contract has been permitted to
be open at any time. Therefore, if a Buy signal
at time t is generated by the trading rule
followed by another Buy signal at t+n days, the
subsequent signals are ignored until the
position is either closed out or an opposing
signal is generated.
Adopting these approaches to position sizing
provides several benefits. It makes the returns
from the trading rules easier to compare rather
than trying to match dollar values between
different contracts. When examining trading
rules that take relatively longer-term positions,
it avoids potential liquidity problems where
consecutive Buy/Sell conditions may persist for
several months leading to large position
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Profitability of Trading Rules in Futures Markets
85
accumulations that may be difficult to liquidate
in a timely manner.
3.2 Transaction Costs
1
We assume transaction costs of US$100
‘round-turn’, which means that a transaction
cost of US$50 is deducted on trade entry and
an additional US$50 on trade exit. This figure
is consistent with the literature where
comparable amounts were also used by Lukac
et al. (1988) and Taylor (1993). Without
explicit deductions for transaction costs, tests
could arguably bias performance towards
trading rules that trade frequently with only
small profits per trade. Therefore the trading
rule parameters perceived as less profitable
due to their longer term trading characteristics
may have actually provided greater profitability
once brokerage charges were deducted.
Another method for assessing the impact of
transaction costs is to quote some typical retail
brokerage rate of US$25, for example, as was
used in Thomas III (1986). This approach
raises considerable concern as to the expected
trade execution price. While the allowance for
brokerage charges may be realistic, this amount
would fail to take poor order execution into
account where trades may actually occur
several points away from where anticipated –
presumably at a price not advantageous to the
client. Consequently, simulated profits may be
significantly overstated compared to the results
observed if the trading had actually occurred in
a real-time market.
Therefore, the allowance of transaction costs
of US$100 per round-turn transaction should
provide a more realistic assessment of how the
trading rules would have performed in practice.
It allows for brokerage and also the fact that
poor order execution may occur. If the
brokerage allowance has been set too high in
this study to better reflect real trading
performance it should, at worst, lead to a more
1 It should be noted that execution costs have been shown
to be substantial in the microstructure literature. While
our research method assumes a small transaction size, in
reality large traders would be subject to execution and
liquidity effects. This biases our findings toward
overstating the profitability of trading rules, however, as
shown later, the trading rules perform poorly despite
this effect. We thank an anonymous referee for pointing
this out.
conservative assessment of reported trading
rule profitability.
3.3 Transaction Timing
All transactions are assumed to have been
undertaken at the closing price on the day on
which the trading signal was generated. This
has the presumption that the trader would be
aware on the day the trading signal was
generated at what price level a trade would
need to be executed and any closing price
beyond the Buy/Sell threshold for the trading
rule would be readily identifiable in time for
the order to be placed.
The market microstructure literature has
documented a characteristic ‘U-Shape’ pattern
for both volatility and trading volumes in
numerous markets. Examples include
Lauterbach and Monroe (1989) with respect to
Gold futures, Abhyankar et al. (1997) in the
UK equities market and Daigler (1997) in
futures contracts over the S&P 500, MMI stock
index and US T-Bonds. Given that price
volatility and trading volumes tend to be higher
in the first and last half-hour periods of the
trading sessions, it would be reasonable to
assume that the trader would find sufficient
liquidity to enter/exit a futures position at these
times.
It is conceded that a position being executed
at the exact closing price for all trades would be
unrealistic. Therefore, under the assumption
that price movements are largely random it is
further assumed that some trades would be
executed at a price favourable and
unfavourable to the trader with equal likelihood
so producing a ‘swings-and-roundabouts’ effect
on trade entry and exit. Any minor and
persistent bias towards unfavourable trade
entry/exit should be adequately accounted for
in the considerable allowance for transaction
costs.
3.4 Parameter Selection and Optimisation
Methods
Parameters for the trading models are based on
a periodic optimisation process. The
optimisation is conducted by testing a range of
parameters and selecting those parameter
values for which the highest profitability was
reported in the immediately prior year. That is,
to avoid any problems associated with ‘in
sample’ testing and reporting simulated profits
derived from perfect hindsight, all reported
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 18 NO 2 (2005)
86
profitability is obtained from tests conducted
‘out-of-sample’. This is achieved by
conducting a ‘walk-forward’ testing procedure
as follows. In the first year of our dataset, 1989,
we test all parameters in the testing range to
determine those that produce maximum profit.
These selected parameters are then applied into
the 1990 dataset and the profits recorded. We
then roll forward one year at a time, repeating
this procedure. Adopting this walk-forward
approach to the testing helps to ensure that the
trading rule parameters selected are comparable
to the information set that would have been
available to the trader operating in the
market at that time. Neftci (1991) persuasively
argued that for trading rules to be truly testable,
the Buy/Sell decision criteria must be
mathematically pre-defined given an
information set available at time t.
3.5 Stop-Loss Provisions
Stop-Loss mechanisms essentially aim to
reduce the losses from any trading decision that
proves to be unprofitable ab initio by exiting a
trade once a pre-determined maximum loss
occurs after the trade is opened. The stop-loss
technique could be demonstrated by an
example of a trader using a trading system that
produces a buy signal at 93.50. The trader buys
a futures contract at 93.50, but also adds a stop-
loss order of 8 points. The broker would then
be instructed to sell/liquidate the position as
soon as the market trades at ? 93.42. Therefore
if the trade is unprofitable from the outset the
trader should be able to set a maximum loss of
8 points representing the largest dollar amount
the trader is willing to risk/lose on that
individual trade.
2
2 One drawback of this stop-loss technique with daily
data is that by using closing prices only, the following
day’s closing price may be significantly below the 93.42
at which the position was to be liquidated. For example,
if a bought position was taken at 93.50 and an 8 point
stop-loss was used, the closing price of the following
day may be at 93.20. As the trading model has
previously specified that the position should be closed
out at ? 93.42, a sell order would be executed - but at
93.20. This results in an expected maximum loss of 8
points becoming a realised 30 point loss. Therefore the
impact on results can see losses on individual
transactions considerably higher than the stop-loss
might suggest which may make results more volatile
than initially anticipated.
The stop-loss mechanisms built into the test
models used in this paper are only used to
avoid large losses occurring while waiting for
an opposing or liquidating signal. This simple
stop-loss mechanism provides an
uncomplicated method of exiting trades that are
unprofitable shortly after entry. Stop-loss
values ranging between $100 and $1,000 (in
$100 increments) have been tested against each
different set of parameters for the trading
systems examined.
4. Trading Rules
4.1 Dual Moving Average Crossover
The Dual Moving Average (DMA) crossover
rule relies on two moving averages as part of its
Buy/Sell signal generating process. Its
description has been well documented in the
practitioner literature by authors including
Babcock (1989) and Bernstein (1995). The
Dual Moving Average Crossover aims to
identify changes in trend where short-term
‘noise’, or erratic price movements, are
removed by the smoothing effect of the moving
average’s derivation of the mean price for the
previous n days. When the two moving
averages used in the calculations have both
changed direction, for example from down to
up, the trend is regarded as having changed
direction. The trader will buy (sell) in
expectation that the new trend indicated by the
increasing (decreasing) value of the two
moving averages would continue. In essence,
this is a short-term momentum strategy.
The method of calculating the moving
averages uses the equations based on that
provided by Neftci (1991). These are defined
here in equations (1) and (2).
!
"
#
$
%
&
=
'
?
=
?
1
0
1
m
S
S t m
P
m
MA
(1)
!
"
#
$
%
&
=
'
?
=
?
1
0
1
n
S
S t n
P
n
MA
(2)
where MA
m
(MA
m
) is a moving average of
length m
; and Pt
the price of a security at
time t.
The basic rationale of the DMA model is
that the short-term fluctuations in the price data
are smoothed making the longer-term trends
more readily identifiable. When the shorter
moving average MA
m
crosses above the longer
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Profitability of Trading Rules in Futures Markets
87
moving average MA
n
from below (above), it
implies that the short-term trend is for prices to
increase (fall) and so generating a buy (sell)
signal accordingly. MA
m
(MA
n
) has values
tested commencing with a 2-day (5-day)
moving average to a 20-day (25-day) moving
average at 2-day (5-day) increments.
4.2 Channel Breakout Rule
This model is sourced from Babcock (1989)
and is one of the simpler trading rules
described in the market literature. The Channel
Breakout system aims to identify when the
price moves out of a trading range in
expectation that the upward (downward) move
will continue. A move above (below) a
previously established trading range is taken as
a buy (sell) signal.
A Channel is said to be of n days length,
where n is specified by the user. The Breakout
is regarded as the closing price on the most
recent trading day being at a higher high (lower
low) than any other day in the n day channel.
When the market makes the n day high (low),
the trader takes a bought (sold) position in the
expectation of that breakout continuing to form
a sustained up (down) trend. The bought (sold)
position is held until the criteria for taking an
opposing sold (bought) position are met, as
described above, or the stop-loss is triggered.
The parameters tested for this system are from
a 5-day Channel to a 25-day Channel
(incrementing one step at a time) with stop-
losses from $100 to $1,000 (incrementing $100
at a time).
5. Results
5.1 Dual Moving Average Model
3
Panel A of Table 1 reveals the performance of
the Dual Moving Average trading rules when
applied to the S&P 500 futures contract during
the test period. At the gross profit level, only
two years were marginally profitable, with the
worst performing year being 1997 with gross
losses of $64,238 reported. Once transaction
costs were deducted the losses became more
severe with the maximum loss of $70,538
3 As this model requires optimisation, the parameters for
1990 were obtained by optimising the parameters on the
data from 1989 and the values applied in 1990 represent
those parameters producing the highest gross profits in
the previous year.
being reported in 1997. For some years the
maximum drawdown observed was relatively
high, for example $67,400 and $51,150 for
1997 and 1998, respectively. Therefore, it
appears that not only were large losses
observed but also the funding costs were high
given that only one contract was being traded at
any given time highlighting the poor
performance of the DMA model in the S&P
500 futures market during the test period. The
lengths of the moving averages could be
described as highly erratic. They ranged from
quite large values, such as MA
m
= 20 and MA
n
= 25 in 1991, through to quite small values,
such as MA
m
= 2 and MA
n
= 15 in 1998. This
would provide little comfort to the trader trying
to identify a priori which parameters to use and
suggests considerable instability in the returns
from the futures contract across time.
Panel B presents the results for the DMA
model when applied in the US T-Bond futures
market and again, few years provide profits.
Other performance characteristics, such as
maximum drawdown were considerably less
severe than that observed in the S&P 500.
Although the drawdown was lower, the
percentage of profitable trades was generally
well below that expected by chance.
Furthermore, as was the case with the DMA
results for the S&P 500, the parameters appear
to be equally unstable for the US T-Bond
futures with no real preference for longer or
shorter moving average lengths across time.
Panel C presents the profit results for the
British Pound. As with previous studies in
foreign exchange markets (for example Taylor,
1992), gross and net profits were available to
speculators in British Pound futures during the
test period. Specifically, net profits were
available in five of the nine years examined
with 1992 providing the highest net profit of
$18,888. The maximum drawdown figures
were comparable to those observed in the US
T-Bond market and parameter instability was
evident during the test period. Panels D and E
report results for the COMEX gold and corn
contracts. These results largely mirror those
discussed above.
Table 2 presents the statistical significance
of the reported gross and net profits for all five
contracts. Generally, it reveals that while
aggregate gross profits were available in three
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 18 NO 2 (2005)
88
Table 1
Dual Moving Average Profit Results
1990 1991 1992 1993 1994 1995 1996 1997 1998
Panel A: S&P 500
Parameter Values 2/10:300 10/25:900 16/25:100 10/20:100 4/5:1000 12/15:800 8/20:300 4/5:300 10/15:200
Number of Trades 38 12 9 21 63 18 16 63 20
Percent Profitable 18% 8% 11% 10% 38% 17% 0% 16% 15%
Max Drawdown 7,575 10,388 1,450 7,775 9,238 8,750 12,550 67,400 51,150
Gross Profit 2,738 -1,988 -413 -7,625 -6,175 425 -12,550 -64,238 -34,925
Net Profit -1,062 -3,188 -1,313 -9,725 -12,475 -1,375 -14,150 -70,538 -36,925
Panel B: US T-Bonds
Parameter Values 16/20:600 20/25:1000 8/25:900 10/16:300 2/5:600 18/20:1000 2/25:300 4/25:100 2/15:200
Number of Trades 19 12 9 18 52 19 16 12 20
Percent Profitable 37% 25% 0% 33% 38% 37% 25% 0% 10%
Maximum Drawdown 8,950 5,531 7,363 2,119 5,919 7,375 5,938 4,157 5,519
Gross Profit -2,200 3,563 -7,363 3,350 -4,925 -719 8,168 -3,969 -3,088
Net Profit -4,100 2,363 -8,263 1,550 -10,125 -2,619 6,568 -5,169 -5,088
Panel C: British Pound
Parameter Values 4/10:700 16/25:1000 4/25:800 2/25:200 12/15:400 10/15:1000 20/25:800 12/25:800 4/15:600
Number of Trades 22 10 5 26 16 23 18 14 23
Percent Profitable 36.0% 30.0% 60.0% 15.0% 38.0% 30.0% 39.0% 14.0% 17.0%
Maximum Drawdown 3,188 8,188 1,688 7,513 2,013 11,525 4,300 6,350 7,775
Gross Profit 18,888 11,675 19,388 -1,575 4,075 -9,025 3,363 -5,625 -7,450
Net Profit 16,688 10,675 18,888 -4,175 2,475 -11,325 1,563 -7,025 -9,750
Panel D: COMEX Gold
Parameter Values 8/20:100 12/15:700 4/10:300 10/25:600 16/25:100 4/10:700 18/20:800 10/15:100 2/5:800
Number of Trades 12 18 29 6 8 30 15 15 56
Percent Profitable 8% 50% 28% 50% 0% 23% 53% 27% 30.0%
Maximum Drawdown 3,010 1,520 2,400 1,410 1,830 4,260 1,140 500 6,810
Gross Profit -1,870 1,850 910 6,970 -1,810 -4,120 2,430 5,270 -6,560
Net Profit -3,070 50 -1,990 6,370 -2,610 -7,120 930 3,770 -12,160
Panel E: Corn
Parameter Values 4/5:400 14/25:100 2/10:100 18/25:300 6/15:200 14/15:600 8/10:200 2/25:300 8/20:400
Number of Trades 62 16 27 9 21 21 29 9 14
Percent Profitable 50% 0% 30% 44% 24% 43% 28% 22% 50%
Maximum Drawdown 2,113 1,738 950 1,225 1,738 3,150 2,988 1,238 700
Gross Profit 388 -1,738 350 1,238 -438 450 1,863 -388 3,638
Net Profit -5,812 -3,338 -2,350 338 -2,538 -1,650 -1,037 -1,288 2,238
from five markets, namely the British Pound,
Gold and Corn futures markets, none proved to
be statistically significant at any recognised
levels. Once transaction costs were
incorporated only one market was able to
produce positive net profits, namely British
Pound futures. However, the net profitability
for the British Pound failed to be statistical
significantly different from the zero.
5.2 Channel Breakout Rule
Table 3, Panel A shows the profitability of the
Channel Breakout Rule (CHB) rule when
applied to the S&P 500 futures contract during
the test period. The CHB produced inconsistent
profitability with minor profits and losses being
reported between 1990 and 1996. The dramatic
increase in profitability reported for 1997 and
1998 provided the most profitable years of the
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Profitability of Trading Rules in Futures Markets
89
Table 2
Dual Moving Average Statistical Results
Panel A: S&P 500
Gross Profits -124,751 Net Profits -150,751
t-statistic -0.096 t-statistic -0.062
p-value 0.466 p-value 0.478
Panel B: US T-Bonds
Gross Profits -7,183 Net Profits -24,883
t-statistic -0.64 t-statistic -0.16
p-value 0.29 p-value 0.44
Panel C: British Pound
Gross Profits 33,714 Net Profits 18,014
t-statistic 0.33 t-statistic 0.61
p-value 0.39 p-value 0.30
Panel D: COMEX Gold
Gross Profits 3,070 Net Profits -15,830
t-statistic 0.82 t-statistic -0.37
p-value 0.25 p-value 0.37
Panel E: Corn
Gross Profits 5,363 Net Profits -15,437
t-statistic 0.28 t-statistic -0.05
p-value 0.40 p-value 0.48
test period returning $21,013 and $35,925,
respectively. When considering the broader
performance characteristics, the CHB rule
performed particularly poorly during the test
period. Although the maximum drawdown was
considerably less than that observed for the
DMA rule in the same market, the percentage
of profitable trades was remarkably poor with
the highest result of 29% recorded for both
1991 and 1998. Of the 18 trades simulated
between 1992 and 1994, not one trade was
profitable. It is however curious that even with
a percentage of profitable trades of only 17% in
1997 the model was still able to produce profits
after transaction costs.
Panels B to E present the results for the
CHB for US T-Bonds, British Pound, gold and
corn, respectively. Their results are broadly
similar to the S&P and so only brief further
discussion is made. For example, when
examining the profitability of the CHB in US
T-Bonds it is apparent that the trading rule was
able to produce net profits in six of the nine
years tested, with the highest net profit being
$12,075 (1995). In the case of the British
Pound very mixed profit results are reported
with net losses reported in five of the nine years
examined, though one large net profit spike of
$31,825 appeared in 1992. For the COMEX
Gold futures, one very striking observation is
the very low percentage of profitable trades -
zero percent of trades were profitable in all but
three years from nine examined.
Table 4 presents the statistical significance
of the reported gross and net profits for all five
contracts. Despite the simplicity of the Channel
Breakout trading rule it was able to produce
profits in most markets, even after allowances
for transaction costs, but not at any recognised
significance level.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 18 NO 2 (2005)
90
Table 3
Channel Breakout Rule Profit Results
1990 1991 1992 1993 1994 1995 1996 1997 1998
Panel A: S&P 500
Parameter Values 18:100 6:100 10:100 22:100 22:100 7:300 18:800 6:100 14:100
Number Trades 6 7 9 4 5 9 7 6 7
Percent Profitable 17% 29% 0% 0% 0% 11% 0% 17% 29%
Maximum Drawdown 3,375 1,700 6,413 950 1,875 3,300 5,663 9,275 3,125
Gross Profit -3,000 3,325 -6,013 -900 -1,675 -3,238 -5,663 21,613 36,625
Net Profit -3,600 2,625 -6,913 -1,300 -2,175 -4,138 -6,363 21,013 35,925
Panel B: US T-Bonds
Parameter Values 22:200 21:600 5:300 15:600 13:700 18:1000 6:300 7:800 12:800
Number Trades 6 7 10 3 5 3 9 5 2
Percent Profitable 0% 43% 10% 33% 20% 67% 0% 40% 50%
Maximum Drawdown 2,531 2,231 3,501 1,756 5,094 1,531 4,381 1,600 594
Gross Profit -2,531 3,419 3,438 9,337 144 12,375 -4,194 4,350 6,344
Net Profit -3,131 2,719 2,438 9,037 -356 12,075 -5,094 3,850 6,144
Panel C: British Pound
Parameter Values 5:400 22:100 5:1000 6:900 14:100 8:500 18:100 13:600 6:1000
Number Trades 7 12 5 9 5 6 4 6 9
Percent Profitable 29% 8% 80% 0% 20% 0% 25% 17% 11%
Maximum Drawdown 2,188 5,825 2,175 8,563 1,163 4,125 500 3,125 5,000
Gross Profit 14,325 -3,463 32,325 -8,563 1,238 -3,750 8,138 -1,800 -4,975
Net Profit 13,625 -4,663 31,825 -9,463 738 -4,350 7,738 -2,400 -5,875
Panel D: Gold
Parameter Values 5:900 17:100 24:100 7:600 10:1000 17:100 25:800 5:900 14:100
Number of Trades 7 5 5 2 2 1 1 10 10
Percent Profitable 0% 0% 0% 50% 0% 0% 100% 20% 0%
Maximum Drawdown 6,500 520 660 740 1,450 330 450 3,370 1,400
Gross Profit -6,500 -480 -660 4,640 -1,210 -300 3,450 -1,300 -1,330
Net Profit -7,200 -980 -1,160 4,440 -1,410 -400 3,350 -2,300 -2,330
Panel E: Corn
Parameter Values 18:100 7:200 23:800 17:300 12:1000 9:200 24:700 8:1000 5:800
Number of Trades 6 7 2 4 3 5 3 6 7
Percent Profitable 0% 29% 50% 50% 33% 40% 67% 33% 57%
Maximum Drawdown 875 1,625 700 788 1,513 900 1,188 1,238 1,600
Gross Profit -863 -800 2,538 1,588 363 2,800 6,300 -525 1,525
Net Profit -1,463 -1,500 2,338 1,188 63 2,300 6,000 -1,125 825
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
Profitability of Trading Rules in Futures Markets
91
Table 4
Channel Breakout Rule Statistical Results
Panel A: S&P 500
Gross Profits 41,074 Net Profits 35,074
t-statistic 0.38 t-statistic 0.45
p-value 0.37 p-value 0.35
Panel B: US T-Bonds
Gross Profits 32,682 Net Profits 27,682
t-statistic 0.08 t-statistic 0.13
p-value 0.47 p-value 0.45
Panel C: British Pound
Gross Profits 33,475 Net Profits 27,175
t-statistic 0.41 t-statistic 0.50
p-value 0.36 p-value 0.33
Panel D: COMEX Gold
Gross Profits -3,690 Net Profits -7,990
t-statistic -0.71 t-statistic -0.45
p-value 0.28 p-value 0.35
Panel E: Corn
Gross Profits 12,926 Net Profits 8,626
t-statistic 0.10 t-statistic 0.27
p-value 0.47 p-value 0.41
5.3 Discussion
A couple of brief comments are worth making
at this stage. Our findings reveal that neither
rule is able to deliver significant profitable
outcomes. This is notable since, given the fact
that we have ignored execution costs, as
indicated earlier; such a simplification induces
a bias towards overstatement of the success of
the trading rules. Further, the fact that we use
daily data may also impact our analysis. For
example, as highlighted earlier daily data
present an issue regarding the stop-loss
technique in that an expected maximum loss
may understate the magnitude of the actual
realised loss. As such, the impact on results can
see losses significantly higher than the stop-
loss might suggest which may make results
more volatile than initially anticipated. A final
observation relates to a possible learning (or
maturity) effect in our results. For example,
applying the Channel Breakout rule to the corn
contract is suggestive of a trend upwards in the
percentage of profitable trades, with 0%
profitable in 1990 to 57% in 1998. However,
generally the findings are quite mixed in this
regard, thereby preventing a strong conclusion
to be drawn.
6. Conclusion
This paper has presented tests for two different
trading rules, namely, the Dual Moving
Average (DMA) model and the Channel
Breakout (CHB) rule. These rules were tested
across five markets, those being the futures
contracts over the S&P 500, British Pound, US
T-Bonds, COMEX Gold and Corn using daily
data over the period 1990 to 1998.
Despite their supposed profitability reflected
by their persistent appearance in the literature,
they failed to produce profits at any statistical
level during the test period on the five markets
considered. In the case of the Dual Moving
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
ACCOUNTING RESEARCH JOURNAL VOLUME 18 NO 2 (2005)
92
Average (DMA) trading rules, gross profits
were reported in three from five markets, but
this fell to one from five once transaction costs
were incorporated. Neither the gross nor net
profits were statistically significant for any of
the markets examined. Similarly, the CHB rule
failed to produce statistically significant profits
for any of the markets examined. While gross
and net profits were available in four from five
markets, the profits were not significantly
different from zero. These findings are
consistent with market efficiency.
The parameter values derived during the
optimisation process remained generally highly
unstable for their respective models. Whether
this instability reflects some form of non-
stationarity of returns, time-varying volatility
characteristics (for example possibly consistent
with ARCH/GARCH effects) or some other
econometric property remains to be determined
and are offered as possible explanations of the
parameter instability issue.
References
Abhyankar, A., Ghosh, D., Levin E. and Limmack R.J.
(1997), ‘Bid-Ask Spreads, Trading Volume and
Volatility: Intra-Day Evidence from the London Stock
Exchange’, Journal of Business Finance &
Accounting, vol. 24, April, pp. 343–362.
Babcock, B.J. (1989), The Dow-Jones Irwin Guide To
Trading Systems, Richard D. Irwin Inc., Illinois USA.
Bernstein, J. (1995), The Compleat Day Trader, McGraw-
Hill, NewYork, USA.
Daigler, R.T. (1997), ‘Intraday Futures Volatility and
Theories of Market Behavior’, Journal of Futures
Markets, vol. 17(1), pp. 45–74.
Dale, C. and Workman, R. (1981), ‘Patterns of Price
Movements in Treasury Bill Futures’, Journal of
Economics and Business, vol. 33, pp. 81–87.
Lauterbach, B. and Monroe, M. (1989), ‘Evidence on the
Effect of Information and Noise Trading on Intraday
Gold Futures Returns’, Journal of Futures Markets,
vol. 9(4), pp. 297-305.
Louis P., Lukac, B., Brorsen, W. and Irwin, S.H. (1988), ‘A
Test of Futures Market Disequilibrium Using Twelve
Different Technical Trading Systems’, Applied
Economics, vol. 20, pp. 623–639.
Merrick, J.J. (2000), ‘Pascal Spreading of Short-Term
Interest Rate Contracts’, Journal of Futures Markets,
vol. 20(10), pp. 889–910.
Neftci, S.H. (1991), ‘Naive Trading Rules in Financial
Markets and Wiener-Kolmogorov Prediction Theory:
A Study of “Technical Analysis”’, Journal of Business,
vol. 64(4), pp. 549–571.
Poitras, G. (1987), ‘Golden Turtle Tracks: In Search of
Unexploited Profits in Gold Spreads’, Journal of
Futures Markets, vol. 7(4), pp. 397–412.
Simon, D.P. (1999), ‘The Soybean Crush Spread: Empirical
Evidence and Trading Strategies’, Journal of Futures
Markets, vol. 19(3), pp. 271–289.
Taylor, S. J. (1992), ‘Rewards Available to Currency
Futures Speculators: Compensation for Risk or
Inefficient Pricing?’, in Stein, J.L. and Goss, B.A.
(eds.), Supplement to the Economic Record, Special
Issues on Futures Markets–Price Discovery and Price
Determination.
Taylor, S. J. (1993), ‘Trading Futures Using a Channel
Rule: A Study of the Predictive Power of Technical
Analysis with Currency Examples’, Manuscript
received prior to publication in The Journal of Futures
Markets.
Thomas III, L.R. (1986), ‘Random Walk Profits in
Currency Futures Trading’, Journal of Futures
Markets, vol. 6(1), pp. 109–125.
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
This article has been cited by:
1. Shan Wang, Zhi-Qiang Jiang, Sai-Ping Li, Wei-Xing Zhou. 2015. Testing the performance
of technical trading rules in the Chinese markets based on superior predictive test. Physica A:
Statistical Mechanics and its Applications 439, 114-123. [CrossRef]
D
o
w
n
l
o
a
d
e
d
b
y
P
O
N
D
I
C
H
E
R
R
Y
U
N
I
V
E
R
S
I
T
Y
A
t
2
0
:
5
9
2
4
J
a
n
u
a
r
y
2
0
1
6
(
P
T
)
doc_749335157.pdf