TSFDC: A trading strategy based on forecasting directional change

• Date: 01 July 2018
• Published date: 01 July 2018
• Author: Amer M. Bakhach,V.L. Raju Chinthalapati,Edward P.K. Tsang
• DOI: http://doi.org/10.1002/isaf.1425

RESEARCH ARTICLE

TSFDC: A trading strategy based on forecasting directional

change

Amer M. Bakhach

1

|Edward P.K. Tsang

1

|V.L. Raju Chinthalapati

2

1

Centre for Computational Finance and

Economic Agents, University of Essex,

Colchester, UK

2

Dept. Accounting and Finance, University of

Greenwich, Greenwich, UK

Correspondence

Amer M. Bakhach, Centre for Computational

Finance and Economic Agents, University of

Essex, Colchester, UK

Email: abakhaa@essex.ac.uk

Summary

Directional Change (DC) is a technique to summarize price movements in a financial

market. According to the DC concept, data is sampled only when the magnitude of

price change is significant according to the investor. In this paper, we develop a

contrarian trading strategy named TSFDC. TSFDC is based on a forecasting model

which aims to predict the change of the direction of market's trend under the DC

context. We examine the profitability, risk and risk‐adjusted return of TSFDC in the

FX market using eight currency pairs. The results suggest that TSFDC outperforms

the buy and hold approach and another DC‐based trading strategy.

KEYWORDS

Algorithmic trading, directional change,FX trading

1|INTRODUCTION

Directional Change (DC) is an approach to summarize prices

movement (Ao & Tsang, 2013). Under the DC framework, the market

is simplified as alternating uptrend and downtrend. A trend is identi-

fied as a change in market price larger than, or equal to, a specific

threshold. This threshold, named theta, is set by the observer and usu-

ally expressed as percentage. A trend ends whenever a price change of

the same threshold, theta, is observed in the inverse direction. For

example, a market downtrend ends when we observe a price rise of

magnitude theta; in this case we say that the market changes its direc-

tion to an uptrend. Similarly, a market's uptrend ends when we

observe a price drop of magnitude theta. Many studies (e.g. Aloud,

Fasli, Tsang, Dupuis, & Olsen, 2013; Aloud, Tsang, & Olsen, 2012;

Glattfelder, Dupuis, & Olsen, 2011; Masry, 2013) have reported that

the DC framework is helpful in studying the foreign exchange (FX)

markets. However, developing trading strategies based on the DC

framework is still in its early stages.

In general, the literature encompasses a large number of trading

strategies. Many of these trading strategies are based on forecasting

models. Some of these forecasting models have the traditional objec-

tive of predicting the change of the direction of market's trend (e.g.

Chen & Chen, 2016; Davis & Episcopos, 2001; Galeshchuk &

Mukherjee, 2017; Göçkena, Özçalıcıb, Borua, & Dosdo˘gru, 2016; Li,

Deng, & Luo, 2009). In 2016, Bakhach, Tsang, and Jalalian (2016) pro-

posed a forecasting model, under the DC context, which aims to

answer the question of whether the current trend will continue until

a specific magnitude, expressed as a percentage, is reached before

the trend ends. They also reported that, in some cases, the accuracy

of their proposed forecasting model was over 80%. However, they

did not present any trading strategy. The establishment of such trad-

ing strategy is important in the sense of giving some empirical guaran-

tee that the proposed forecasting method can be used in the real‐

world (Cavalcante, Brasileiro, Souza, Nobrega, & Oliveira, 2016).

In this paper we present a novel trading strategy named TSFDC.

TSFDC relies on the forecasting model introduced by Bakhach, Tsang,

and Jalalian (2016) to decide when to initiate a trade. We examine the

performance of TSFDC in the FX market using eight currency pairs.

We evaluate the profitability, risk, and risk‐adjusted performance of

TSFDC. We compare the performance of TSFDC to other DC‐based

trading strategies.

The paper continues as follows: Section 2 describes the concept

of Directional Changes. Section 3 provides a brief summary of the

forecasting model introduced in Bakhach, Tsang, and Jalalian (2016)

We present TSFDC and its trading rules in Section 4. We discuss the

selection and preparation of the datasets and the employed evaluation

metrics in Section 5. The details of the experiments, conducted to

evaluate the performance of TSFDC, are provided in Section 6.

Section 7 reports and discusses the results of these experiments. We

compare our trading strategy with other DC‐based trading strategies

in Section 8. Finally, we summarize the major findings of this paper in

Section 9.

Received: 15 July 2017 Revised: 9 February 2018 Accepted: 21 February 2018

DOI: 10.1002/isaf.1425

Intell Sys Acc Fin Mgmt. 2018;25:105–123. Copyright © 2018 John Wiley & Sons, Ltd.wileyonlinelibrary.com/journal/isaf 105

2|DIRECTIONAL CHANGES

2.1 |The DC Framework: The main concept

In this section, we explain how market prices are summarized based on

the DC concept (Ao & Tsang, 2013; Guillaume et al., 1997). Directional

change (DC) is an approach to summarize price changes. Under the DC

framework, the market is represented as alternating uptrends and

downtrends. The basic idea is that the magnitude of price changes dur-

ing an uptrend, or a downtrend, must be at least equal to a specific

threshold theta usually expressed as a percentage. Let us consider a

market in a downtrend. Let P

EXT

be the lowest price in this downtrend

and P

c

be the current price. We say that the market switches its direc-

tion from downtrend to uptrend whenever P

c

becomes greater than

P

EXT

by at least theta (where theta is the threshold predetermined by

the observer; usually expressed as a percentage). Similarly, if the mar-

ket is in uptrend, P

EXT

would refer to the highest price in this uptrend.

We say that the market switches its direction from an uptrend to a

downtrend if P

c

is lower than P

EXT

by at least theta (the threshold

predetermined by the observer).

Figure 1 illustrates the mid‐prices movements of GBP/CHF

sampled minute by minute from 1/1/2013 19:05 to 1/2/2013 02:05

(UK time). Whereas, Figure 2 illustrates how these mid‐prices are

simplified as uptrends and downtrends under the DC context. The

detection of a new uptrend or a new downtrend is a formalized

inequality, as shown in (1).

Pc−PEXT

PEXT















≥theta (1)

If (1) holds,then the time at whichthe market traded at P

EXT

is called

an ‘extremepoint’(e.g. points A and D in Figure 2), andthe time at which

the market tradesat P

c

is called a DC confirmation point, orDCC point

for short (e.g. points A

0.1

and D

0.1

in Figure 2). Note that whilst an

extremepoint is the end of one trend,it is also the start ofthe next trend,

which has an opposite direction.An extreme point is only recognizedin

hindsight; precisely at the DCC point.For example, in Figure 2, at point

A

0.1

we confirm that point A is an extreme point. Similarly, inFigure 2,

at point D

0.1

we confirm thatpoint D is an extreme point.

Under the DC framework, a trend is dissected into a DC event

and an overshoot (OS) event. A DC event starts with an extreme point

and ends with a DCC point. We refer to a specific DC event by its

starting point, i.e. extreme point, and its DCC point. For example, in

Figure 2 the DC event which starts at point B and ends at point B

0.1

is denoted as [BB

0.1

]. An OS event starts at the DCC point and ends

at the next extreme point.

The DC summary of a given market is the identification of the

DC and OS events, governed by the threshold theta. Figure 2

illustrates an example of a DC summary of prices in Figure 1. Here,

theta is a percentage that the observer considers substantial.

Note that for a given time series and a predetermined threshold,

the DC summary is unique. However, we may generate multiple

DC summaries for the same considered prices series by selecting

multiple thresholds. One observer may consider 0.1% an important

change, while another observer may consider a price‘s change of 0.2%

as important. Observers who use different thresholds will observe

different DC events and trends.Any price's change less than the identi-

fied threshold will not be considered as a trend when summarizing

market prices.

The chosen threshold determines what constitutes a directional

change. For example, Figure 2 and Figure 3 provide two distinct DC

summaries, using two different thresholds, for the same prices series.

If a greater threshold been chosen, then less directional changes

would have been concluded between prices. For instance, in

Figure 2 the DC summary of threshold 0.1% reveals 4 downtrends

and 3 uptrends. Whereas, in Figure 3 the DC summary of threshold

0.2% detects 2 downtrends and 1 uptrend.

The DC concept is similar to the zigzag indicator (Azzini, Pereira, &

Tettamanzi, 2009; Raftopoulos, 2003). The zigzag approach models

price movement as alternating uptrend and downtrend. The price

change during an uptrend or a downtrend must be at least equal to

a specific threshold. The main difference between the DC approach

and the zigzag indicator is that a trend, under the DC methodology,

is dissected into: 1) a DC event of fixed percentage equal to the

selected threshold and 2) an OS event represented by the remaining

part of the trend before it reverses. This segmentation of a trend into

DC and OS event, under the DC framework, has been proved to be

helpful to analyse and characterize financial markets (Aloud et al.,

FIGURE 1 GBP/CHF mid‐prices sampled minute by minute from 1/1/2013 19:05 to 1/2/2013 02:05 (UK)

106 BAKHACH ET AL.