SSRN Id1762118

Download as pdf or txt
Download as pdf or txt
You are on page 1of 25

Optimal Trend Following Trading Rules∗

Min Dai, Qing Zhang and Qiji Jim Zhu

July 19, 2011

Abstract
We develop an optimal trend following trading rule in a bull-bear switching market, where the
drift of the stock price switches between two parameters corresponding to an uptrend (bull mar-
ket) and a downtrend (bear market) according to an unobservable Markov chain. We consider
a finite horizon investment problem and aim to maximize the expected return of the terminal
wealth. We start by restricting to allowing flat and long positions only and describe the trading
decisions using a sequence of stopping times indicating the time of entering and exiting long
positions. Assuming trading all available funds, we show that the optimal trading strategy is
a trend following system characterized by the conditional probability in the uptrend crossing
two threshold curves. The thresholds can be obtained by solving the associated HJB equations.
In addition, we examine trading strategies with short selling in terms of an approximation.
Simulations and empirical experiments are conducted and reported.
Keywords: Trend following trading rule, bull-bear switching model, partial information, HJB
equations
AMS subject classifications: 91G80, 93E11, 93E20


Dai is from Department of Mathematics, National University of Singapore (NUS) 10, Lower Kent Ridge Road,
Singapore 119076, [email protected], Tel. (65) 6516-2754, Fax (65) 6779-5452, and he is also affiliated with Risk
Management Institute and Institute of Real Estate Studies, NUS. Zhang is from Department of Mathematics, The
University of Georgia, Athens, GA 30602, USA, [email protected], Tel. (706) 542-2616, Fax (706) 542-2573. Zhu
is from Department of Mathematics, Western Michigan University, Kalamazoo, MI 49008, USA, [email protected], Tel.
(269) 387-4535, Fax (269) 387-4530. Dai is supported by the Singapore MOE AcRF grant (No. R-146-000-138-112)
and the NUS RMI grant (No.R-146-000-124-720/646). We thank seminar participants at Carnegie Mellon University,
Wayne State University, University of Illinois at Chicago, and 2010 Mathematical Finance and PDE conference for
helpful comments.

Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=1762118
http://ssrn.com/abstract=1762118
1 Introduction

Roughly speaking, trading strategies can be classified as i) the buy and hold strategy, ii) the contra-
trend strategy, and iii) the trend following strategy. The buy and hold strategy can be justified
because the average return of stocks is higher than the bank rate1 . An investor that performs
the contra-trending strategy purchases shares when prices fall to some low level and sells when
they go up to a certain high level (known as buy-low-sell-high). As the name suggests, the trend
following strategy tries to enter the market in the uptrend and signal investors to exit when the
trend reverses. In contrast to the contra-trend investors, a trend following believer often purchases
shares when prices go up to a certain level and sells when they fall to a higher level (known as
buy-high-sell-higher).
There is an extensive literature devoted to the contra-trend strategy. For instance, Merton [14]
pioneered the continuous-time portfolio selection with utility maximization, which was subsequently
extended to incorporate transaction costs by Magil and Constantinidies [13] (see also Davis and
Norman [5], Shreve and Soner [19], Liu and Loeweinstein [12], Dai and Yi [3], and references
therein). The resulting strategies turn out to be contra-trend because the investor is risk averse
and the stock market is assumed to follow a geometric Brownian motion with constant drift and
volatility. Recently Zhang and Zhang [24] showed that the optimal trading strategy in a mean
reverting market is also contra-trend. Other work relevant to the contra-trend strategy includes
Dai et al. [1], Song et al. [20], Zervors et al. [23], among others.
The present paper is concerned with a trend following trading rule. Traders who adopt this
trading rule often use moving averages to determine the general direction of the market and to
generate trade signals [21]. However, to the best of our knowledge, there is not yet any solid
theoretical framework supporting the use of moving average2 . Recently, Dai et al. [4] provided a
theoretical justification of the trend following strategy in a bull-bear switching market and employed
the conditional probability in the bull market to generate the trade signals. However, the work
imposed a less realistic assumption3 : Only one share of stock is allowed to be traded. In the present
paper, we will remove this restriction and develop an optimal trend following rule. We also carry
1
Recently Shiryaev et al. [18] provided a theoretical justification of the buy and hold strategy from another angle.
2
There does exist research on statistical analysis for trading strategies with moving averages. See, for example,
[6].
3
The same assumption was imposed in [24], [20], and [23].

Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=1762118
http://ssrn.com/abstract=1762118
out extensive simulations and empirical analysis to examine the efficiency of our strategy.
Following [4], we model the trends in the markets using a geometric Brownian motion with
regime switching and partial information. More precisely, two regimes are considered: the uptrend
(bull market) and downtrend (bear market), and the switching process is modeled as a two-state
Markov chain which is not directly observable4 . We consider a finite horizon investment problem,
and our target is to maximize the expected return of the terminal wealth. We begin by considering
the case that only long and flat positions are allowed. We use a sequence of stopping times to
indicate the time of entering and exiting long positions. Assuming trading all available funds, we
show that the optimal trading strategy is a trend following system characterized by the conditional
probability in the uptrend crossing two time-dependent threshold curves. The thresholds can be
obtained through solving a system of HJB equations satisfied by two value functions that are
associated with long and flat positions, respectively. Simulation and market tests are conducted to
demonstrate the efficiency of our strategy.
The next logical question to ask is whether adding short will improve the return. Due to
asymmetry between long and short as well as solvency constraint, the exact formulation with short
selling still eludes us. Hence, we instead utilize the following approximation. First, we consider
trading with the short and flat positions only. Using reverse exchange traded funds to approximate
the short selling we are able to convert it to the case of long and flat. Then, assuming there are
two traders A and B. Trader A trades long and flat only and trader B trades short and flat only.
Combination of the actions of both A and B yields a trading strategy that involves long, short
and flat positions. Simulation and market tests are provided to investigate the performance of the
strategy.
The rest of the paper is arranged as follows. We present the problem formulation in the next
section. Section 3 is devoted to a theoretical characterization of the resulting optimal trading
strategy. We report our simulation results and market tests in Section 4. In Section 5, we examine
the trading strategy when short selling is allowed. We conclude in Section 6. All proofs, some
technical results and details on market tests are given in Appendix.
4
Most existing literature in trading strategies assumes that the switching process is directly observable, e.g. Jang
et al. [9] and Dai et al. [2].

Electronic copy available at: https://ssrn.com/abstract=1762118


2 Problem Formulation

Let Sr denote the stock price at time r satisfying the equation

dSr = Sr [µ(αr )dr + σdBr ], St = X, t ≤ r ≤ T < ∞, (1)

where αr ∈ {1, 2} is a two-state Markov chain, µ(i) ≡ µi is the expected return rate5 in regime
i = 1, 2, σ > 0 is the constant volatility, Br is a standard Brownian motion, and t and T are the
initial and terminal times, respectively.
The process αr represents the market mode at each time r: αr = 1 indicates a bull market
(uptrend) and αr = 2 a bear market (downtrend). Naturally, we assume µ1 > 0 and µ2 < 0.
µ ¶
−λ1 λ1
Let Q = , (λ1 > 0, λ2 > 0), denote the generator of αr . So, λ1 (λ2 ) stands for
λ2 −λ2
the switching intensity from bull to bear (from bear to bull). We assume that {αr } and {Br } are
independent.
Let
t ≤ τ10 ≤ v10 ≤ τ20 ≤ v20 · · · ≤ τn0 ≤ vn0 ≤ · · · , a.s.,

denote a sequence of stopping times. For each n, define

τn = min{τn0 , T } and vn = min{vn0 , T }.

A buying decision is made at τn if τn < T and a selling decision is at vn if vn < T , n = 1, 2, . . ..


In addition, we impose that one has to sell the entire shares by the terminal time T .
We first consider the case that the investor is either long or flat. If she is long, her entire wealth
is invested in the stock account. If she is flat, all of her wealth is in the bank account that draws
interests. Let i = 0, 1 denote the initial position. If initially the position is long (i.e, i = 1), the
corresponding sequence of stopping times is denoted by Λ1 = (v1 , τ2 , v2 , τ3 , . . .). Likewise, if initially
the net position is flat (i = 0), then the corresponding sequence of stopping times is denoted by
Λ0 = (τ1 , v1 , τ2 , v2 , . . .).
Let 0 < Kb < 1 denote the percentage of slippage (or commission) per transaction with a buy
order and 0 < Ks < 1 that with a sell order.
5
Here we assume no dividend payments. If the stock pays a constant dividend yield, we then re-invest the dividends
in the stock. So, our assumption is without loss of generality.

Electronic copy available at: https://ssrn.com/abstract=1762118


Let ρ ≥ 0 denote the risk-free interest rate. Given the initial time t, initial stock price St = S,
initial market trend αt = α ∈ {1, 2}, and initial net position i = 0, 1, the reward functions of the
decision sequences, Λ0 and Λ1 , are the expected return rates of wealth:

Ji (S,
 α, t, Λi )
(  " #I{τn <T } )

 Y∞
Sv 1 − Ks



 Et log eρ(τ1 −t) eρ(τn+1 −vn ) n  , if i = 0,

 Sτn 1 + Kb
 n=1
( " # ∞ " #I{τn <T } )
= Y

 Sv1 ρ(τ2 −v1 ) Sv 1 − Ks

 Et log  e (1 − Ks ) eρ(τn+1 −vn ) n  ,

 S Sτn 1 + Kb

 n=2

if i = 1.

It is easy to see that

Ji (S,
 α, t,
(Λi ) " #)
X∞ µ ¶

 S v 1 − K s

 Et ρ(τ1 − t) + log n
+ ρ(τn+1 − vn ) + log I{τn <T } , if i = 0,

 Sτn 1 + Kb

 (" n=1 #

 Sv 1
= Et log + log(1 − Ks ) + ρ(τ2 − v1 )

 S

 " µ ¶ #)

 ∞
X S 1 − K

 v n s
 +
 log
Sτn
+ ρ(τn+1 − vn ) + log
1 + Kb
I{τn <T } , if i = 1,
n=2


X PN
where the term E ξn for random variables ξn is interpreted as lim supN →∞ E n=1 ξn . Our goal
n=1
is to maximize the reward function6 .

Remark 1 Note that the indicator function I{τn <T } is used in the definition of the reward functions
Ji . This is to ensure that if the last buy order is entered at t = τn < T , then the position will
be sold at vn ≤ T . The indicator function I confines the effective part of the sum to a finite time
horizon so that the reward functions are bounded above.

To exclude trivial cases,7 we always assume

σ2 σ2
µ2 − < ρ < µ1 − . (2)
2 2

Note that only the stock price Sr is observable at time r in marketplace. The market trend αr
is not directly observable. Thus, it is necessary to convert the problem into a completely observable
one. One way to accomplish this is to use the Wonham filter [22].
6
It is easy to see that the problem is equivalent to maximizing the expected logarithm utility of the terminal
wealth over allowable trading strategies that incur proportional transaction costs.
7 2 2
Intuitively, one should never buy stock if ρ ≥ µ1 − σ2 and never sell stock if ρ ≤ µ2 − σ2 .

Electronic copy available at: https://ssrn.com/abstract=1762118


Let pr = P (αr = 1|Sr ) denote the conditional probability of αr = 1 (bull market) given the
filtration Sr = σ{Su : 0 ≤ u ≤ r}. Then we can show (see Wonham [22]) that pr satisfies the
following SDE

(µ1 − µ2 )pr (1 − pr ) b
dpr = [− (λ1 + λ2 ) pr + λ2 ] dr + dBr , (3)
σ
br is the innovation process (a standard Brownian motion; see e.g., Øksendal [15]) given by
where B

br = d log(Sr ) − [(µ1 − µ2 )pr + µ2 − σ 2 /2]dr


dB . (4)
σ
br :
It is easy to see that Sr can be written in terms of B

br .
dSr = Sr [(µ1 − µ2 ) pr + µ2 ] dr + Sr σdB (5)

Given St = S and pt = p, the problem is to choose Λi to maximize the discounted return

Ji (S, p, t, Λi ) ≡ Ji (S, α, t, Λi ),

subject to (3) and (5). We emphasize that this new problem is completely observable because the
conditional probability pr can be obtained using the stock price up to time r.
Note that, for a given Λ0 , we have
Z vn Z vn
Sv n br ,
log = f (pr )dr + σdB (6)
Sτn τn τn

where

σ2
f (pr ) = (µ1 − µ2 )pr + µ2 − . (7)
2

Note also that


Z vn
E br = 0.
σdB (8)
τn

Consequently, the reward function J0 (S, p, t, Λ0 ) can be rewritten completely as a function of pr .


Therefore, it is independent of the initial S, so is J1 (S, p, t, Λ0 ). In view of these, the corresponding
value functions are functions of (p, t).
For i = 0, 1, let Vi (p, t) denote the value function with the states p and net positions i = 0, 1 at
time t. That is,
Vi (p, t) = sup Ji (S, p, t, Λi ).
Λi

The following lemma gives the bounds of the values functions. Its proof is given in Appendix.

Electronic copy available at: https://ssrn.com/abstract=1762118


Lemma 1 We have

V0 (p, t) ≥ ρ(T − t), V1 (p, t) ≥ log(1 − Ks ) + ρ(T − t),

and
µ ¶
σ2
Vi (p, t) ≤ µ1 − (T − t), for i = 0, 1.
2

Next, we consider the associated Hamilton-Jacobi-Bellman equations. It is easy to see that, for
t < T and stopping times τ1 and v1 ,

V0 (p, t) = sup Et {ρ(τ1 − t) − log(1 + Kb ) + V1 (pτ1 , τ1 )}


τ1

and
½Z v1 ¾
V1 (p, t) = sup Et f (ps )ds + log(1 − Ks ) + V0 (pv1 , v1 ) ,
v1 t

where f (·) is as given in (7). Let L denote the generator of (t, pt )


µ ¶
1 (µ1 − µ2 )p(1 − p) 2
L = ∂t + ∂pp + [− (λ1 + λ2 ) p + λ2 ] ∂p .
2 σ
Then, the associated HJB equations are
½
min{−LV0 − ρ, V0 − V1 + log(1 + Kb )} = 0,
(9)
min{−LV1 − f (p), V1 − V0 − log(1 − Ks )} = 0,
with the terminal conditions
½
V0 (p, T ) = 0
(10)
V1 (p, T ) = log(1 − Ks ).
Using the same technique as in Dai et al. [4], we can show that Problem (9)-(10) has a unique
bounded strong solution (V0 , V1 ), where Vi ∈ Wq2,1 ([ε, 1−ε]×[0, T ]), for any ε ∈ (0, 1/2), q ∈ [1, +∞).

Remark 2 In this paper, we restrict the state space of p to (0, 1) because both p = 0 and p = 1 are
entrance boundaries (see Karlin and Taylor [10] and Dai et al. [4] for definition and discussions).

Now we define the buying region (BR), the selling region (SR), and the no-trading region (N T )
as follows:

BR = {(p, t) ∈ (0, 1) × [0, T ) : V1 (p, t) − V0 (p, t) = log(1 + Kb )} ,

SR = {(p, t) ∈ (0, 1) × [0, T ) : V1 (p, t) − V0 (p, t) = log (1 − Ks )} ,

NT = (0, 1) × [0, T ) (BR ∪ SR) .

To study the optimal strategy, we only need to characterize these regions.

Electronic copy available at: https://ssrn.com/abstract=1762118


3 Optimal trading strategy

In this section, we present the main theoretical results.

Theorem 2 There exist two monotonically increasing boundaries p∗s (t), p∗b (t) ∈ C ∞ (0, T ) such
that

SR = {p ∈ (0, 1) × [0, T ) : p ≤ p∗s (t)}, (11)

BR = {p ∈ (0, 1) × [0, T ) : p ≥ p∗b (t)}. (12)

Moreover,
ρ − µ2 + σ 2 /2
i) p∗b (t) ≥ ≥ p∗s (t) for all t ∈ [0, T );
µ1 − µ2
ρ − µ2 + σ 2 /2
ii) lim p∗s (t) = ;
t→T − µ1 − µ2
1 1 + Kb
iii) there is a δ > 2
log such that p∗b (t) = 1 for t ∈ (T − δ, T ).
µ1 − ρ − σ /2 1 − Ks

The proof is placed in Appendix. The theoretical results enable us to examine the validity of
the program codes for numerically solving the system of HJB equations.
We call p∗s (t) (p∗b (t)) the optimal sell (buy) boundary. To better understand Theorem 2, we
provide a numerical result for illustration8 . In Figure 1, we plot the optimal sell and buy boundaries
against time. It can be seen that both the buy boundary and the sell boundary are increasing with
time, between the two boundaries is the no trading region (NT), the buy region (BR) is above the
buy boundary, and the sell region (SR) is below the sell boundary. Moreover, the sell boundary p∗s (t)
ρ − µ2 + σ 2 /2 0.0679 + 0.77 + 0.1842 /2
approaches the theoretical value = = 0.9, as t → T = 1.
µ1 − µ2 0.18 + 0.77
Also, we observe that there is a δ such that p∗b (t) = 1 for t ∈ [T − δ, T ], which indicates that it is
never optimal to buy stock when t is very close to T . Using Theorem 2, the lower bound of δ is
estimated as

1 1 + Kb 1 1.001
2
log = 2
log = 0.021,
µ1 − ρ − σ /2 1 − Ks 0.18 − 0.0657 − 0.184 /2 0.999

which is consistent with the numerical result.


We now point out that our trading strategy is trend following. The numerical results illustrated
in Figure 1 show that the thresholds p∗s (·) and p∗b (·) are rather close to constants except when t
8
Analytic solutions of (9)-(10) are not available, but we can use numerical methods to find numerical solutions
(cf. [4]).

Electronic copy available at: https://ssrn.com/abstract=1762118


Figure 1: Optimal trading strategy
1
BR
0.95
p*(t)
b

0.9

NT
0.85
p

0.8
p*(t)
s

0.75
SR

0.7
0 0.2 0.4 0.6 0.8 1
t

Parameter values: λ1 = 0.36, λ2 = 2.53, µ1 = 0.18, µ2 = −0.77, σ = 0.184, Kb = Ks = 0.001, ρ =


0.0679, T = 1.

approaches T . The behavior of the thresholds when t approaches to T is due to our technical require-
ment of liquidating all the positions at T . Because we are interested in long-term investment, we
will approximate these thresholds, as in [4], by constants p∗s = lim p∗s (t) and p∗b = lim p∗b (t).9
T −t→∞ T −t→∞
Assume the initial position is flat and the initial conditional probability p(0) ∈ (p∗s , p∗b ). Then our
trading strategy can be described as follows. As pt goes up to hit p∗b , we take a long position, that
is, investing all wealth in stock. We will not close out the position unless pt goes down to hit p∗s .
According to (3)-(4), we have

(µ1 − µ2 )pr (1 − pr )
dpr = g(pr )dr + d log Sr , (13)
σ2

where ¡ ¢
(µ1 − µ2 )pt (1 − pt ) (µ1 − µ2 )p + µ2 − σ 2 /2
g(p) = − (λ1 + λ2 ) p + λ2 − .
σ2
(13) implies that the conditional probability pt in the bull market increases (decreases) as the stock
price goes up (down). Hence, our strategy suggests that we buy only when stock price is going up
and sell only when stock price is going down. This is a typical trend following strategy!
We conclude this section by a verification theorem, showing that the solutions V0 and V1 of
problem (9)-(10) are equal to the value functions and sequences of optimal stopping times can be
9
The constant thresholds are essentially associated with the infinity horizon investment problem:
1
lim max E(WT ), where WT be the terminal wealth.
T →∞ T

Electronic copy available at: https://ssrn.com/abstract=1762118


constructed by using (p∗s , p∗b ).

Theorem 3 (Verification Theorem) Let (w0 (p, t), w1 (p, t)) be the unique bounded strong solution
to problem (9)-(10) with w0 (p, t) ≥ 0 and p∗b (t) and p∗s (t) be the associated free boundaries. Then,
w0 (p, t) and w1 (p, t) are equal to the value functions V0 (p, t) and V1 (p, t), respectively.
Moreover, let
Λ∗0 = (τ1∗ , v1∗ , τ2∗ , v2∗ , · · ·),

where the stopping times τ1∗ = T ∧ inf{r ≥ t : pr ≥ p∗b (r)}, vn∗ = T ∧ inf{r ≥ τn∗ : pr ≤ p∗s (r)}, and

τn+1 = T ∧ inf{r > vn∗ : pr ≥ p∗b (r)} for n ≥ 1, and let

Λ∗1 = (v1∗ , τ2∗ , v2∗ , τ3∗ , · · ·),

where the stopping times v1∗ = T ∧ inf{r ≥ t : p∗r ≤ p∗s (r)}, τn∗ = T ∧ inf{r > vn−1
∗ : pr ≥ p∗b (r)},
and vn∗ = T ∧ inf{r ≥ τn∗ : pr ≤ p∗s (r)} for n ≥ 2. If vn∗ → T , a.s., as n → ∞, then Λ∗0 and Λ∗1 are
optimal.

The proof is in Appendix.

4 Simulation and market tests

We use both simulations and tests on historical market data to examine the effectiveness of the
theoretical characterization of the trading strategy. To estimate pt , the conditional probability in a
bull market, we use a discrete version of the stochastic differential equation (13), for t = 0, 1, . . . , N
with dt = 1/252,
µ µ ¶ ¶
(µ1 − µ2 )pt (1 − pt )
pt+1 = min max pt + g(pt )dt + log(St+1 /St ), 0 , 1 , (14)
σ2

where the price process St is determined by the simulated paths or the historical market data. The
min and max are added to ensure the discrete approximation pt of the conditional probability in
the bull market stays in the interval [0, 1]. Note that the equation for pt is the same as that in [4]
because it is irrelevant to objective functions.

4.1 Simulations

For simulation we use the parameters summarized in Table 1 and a 40 year time horizon. They are
the same as those used in [4] so that it is easy to compare the results.

10

Electronic copy available at: https://ssrn.com/abstract=1762118


λ1 λ2 µ1 µ2 σ K ρ
0.36 2.53 0.18 -0.77 0.184 0.001 0.0679
Table 1. Parameter values
Solving for the buy and sell thresholds using numerical solutions to the HJB equation, we derive
p∗s = 0.796 and p∗b = 0.948. We run the 5000 round simulation for 10 times and summarize the
mean of the total (annualized) return and the standard deviation in Table 2.
Trend Following Buy and Hold No. of Trades
Mean 75.76(11.4%) 5.62(4.4%) 41.16
Stdev 2.48 0.39 0.29
Table 2. Statistics of ten 5000-path simulations
Comparing to the simulation results in [4] we only observe a slight improvement in terms of
the ratio of mean return of the trend following strategy and the buy and hold strategy. However,
the improvement is not significant enough to distinguish statistically from the results in [4] despite
theoretically the present paper is more solid than [4]. This also reveals that using the conditional
probability in the bull market as trade signals is rather robust against the change of thresholds.
The above simulation results are based on the average outcomes of large numbers of simulated
paths. We now investigate the performance of our strategy with individual sample paths. Table 3
collects simulation results on 10 single paths using buy-sell thresholds p∗s = 0.795 and p∗b = 0.948
with the same data given in Table 1. We can see that the simulation is very sensitive to individual
paths, but our strategy clearly outperforms the buy and hold strategy.
Note that this observation is consistent with the measurement of an effective investment strategy
in marketplace. For example, O’Neil’s CANSLIM works during a period of time does not mean it
works on each stock when applied. How it works is measured based on the overall average when
applied to a group of stocks fitting the prescribed selection criteria.
Trend Following Buy and Hold No. of Trades
67.080 3.2892 36.000
24.804 2.2498 42.000
22.509 0.40591 42.000
1887.8 257.75 33.000
26.059 0.16373 48.000
60.267 1.5325 43.000
34.832 5.7747 42.000
8.6456 0.077789 46.000
128.51 30.293 37.000
224.80 29.807 40.000
Table 3. Ten single-path simulations

11

Electronic copy available at: https://ssrn.com/abstract=1762118


4.2 Market tests

We now turn to test the trend following trading strategy in real markets. Here we conduct the ex-
ante tests.10 The parameters are determined using only information available at the decision time
and updated periodically. More concretely, let us use the test of SP500 as an example to explain
the process. We have SP500 historical closing data since 1962. We use the first 10 years data to
derive statistics for bull (rally at least 20%) and bear (decline at least 20%) markets to serve as
the initial parameters for determining the buy-sell thresholds. We then update the parameters and
thresholds at the beginning of every year if new up or down trends are confirmed ending before the
beginning of the year. We update the parameters using the so called exponential average method
in which the update of the parameters is determined by the old parameters and new parameters
with formula
update = (1 − 2/N )old + (2/N )new,

where we chose N = 6 based on the number of bull and bear markets between 1962–1972. The
exponential average allows to overweight the new parameters while avoiding unwanted abrupt
changes due to dropping old parameters. Then we use the yearly updated parameters to calculate
the corresponding thresholds. Finally, we use these parameters and thresholds to test the SP500
index from 1972-2011. The equity curve of the trend following strategy is compared to the buy
and hold strategy in the same period of time in Figure 2. The upper, middle and the lower curves
represents the equity curves of the trend following strategy, the buy and hold strategy including
dividend and the SP500 index without dividend adjustment, respectively.
As we can see, the trend following strategy not only outperforms the buy and hold strategy
in total return, but also has a smoother equity curve, which means a higher Sharpe ratio. A
similar ex-ante test is done for the Shanghai Stocks Exchange index (SSE). Since we have only
10 year data (2001–2011) for the SSE, we have to estimate the initial parameters. We summarize
the tests on SP500 and SSE in Table 4 showing annualized return along with quarterly Sharpe
ratio in parenthesis, where the estimate for the SSE initial parameters are µ1 = 1, µ2 = −1, and
λ1 = λ2 = 1.11 The equity curves for the SSE test are shown in Figure 3. The SSE index closing
10
Ex-post tests of a trend following strategy were conducted in [4]. In the present paper we carry out the ex-ante
tests which are more convincing.
11
We have also tried other initial parameter values, e.g. µ1 = 0.4, µ2 = −0.4, λ1 = 1, λ2 = 0.5, which yield similar
results as we present here. These results are available upon request.

12

Electronic copy available at: https://ssrn.com/abstract=1762118


3
10

2
10

1975 1980 1985 1990 1995 2000 2005 2010

Figure 2: Trend following trading of SP500 1972–2011 compared with buy and hold

4
10

3
10
2002 2004 2006 2008 2010 2012

Figure 3: Trend following trading of SSE 2001–2011 compared with buy and hold

13

Electronic copy available at: https://ssrn.com/abstract=1762118


price has already adjusted for dividend. Thus, in Figure 3 we have only two curves: the upper
represents the equity curve of the trend following method and the lower is that of the buy and
hold strategy. The yearly parameters and thresholds for both SP500 and SSE tests are included in
Tables in Appendix 6.2.
Index(time frame) Trend following Buy and hold 10 year bonds
SP500 (1972-2011) 11.03%(0.217) 9.8%(0.128) 6.79%
SSE (2001-2011) 14.0%(0.263) 2.58%(0.083) 3%
Table 4. Testing results for trend following trading strategies
The test result for SP500 here is, if not better, at least comparable to the ex-pose test in [4]
showing that trends indeed exists in the price movement of SP500.12 The test on SSE shows that
the trend also exists in China market.
As alluded to in the end of the last section, market trends are the consequence of aggregating
many individual actions. Thus, it seems more stable than the generated paths using pure Brownian
motions. Nevertheless, sensitivity with respect to parameters is still to be expected and putting
test results into practice should be cautious.

5 Allowing shorts

Can we benefit by adding shorts? In practice there are differences between short and long. Notably,
the short risks may lose more than the initial capital, so the no-bankruptcy constraint would have
to be imposed, which makes the problem intractable under our theoretical framework. One way to
circumvent this difficulty is to use the reverse Exchange Traded Funds (ETFs) of the corresponding
indices to determine threshold values for short selling. In such an approximation to short sell Sr
we long a reverse ETF which is equivalent to longing an asset Sˆr = 1/Sr . First we consider the
case when only short and flat positions are allowed. Use qr = 1 − pr to represent the conditional
probability of Sˆr in an uptrend (equivalently Sr is in a downtrend) and denote

µˆ1 = σ 2 − µ2 , µˆ2 = σ 2 − µ1 , λˆ1 = λ2 , λˆ2 = λ1 . (15)

It is easy to verify that the process Sˆr and qr satisfy the system of stochastic differential equations

dSˆr = Sˆr [(µˆ1 − µˆ2 ) qr + µˆ2 ] dr + Sˆr σdB


br , Ŝt = Ŝ, (16)
12
In [4], there is a mistake that the dividends are not treated as reinvestment. As a correction, the returns of the
buy and hold strategy and the trend following strategy in [4] (Table 10) should be respectively 54.6 and 70.9, instead
of 33.5 and 64.98, for SP500(1962-2008)

14

Electronic copy available at: https://ssrn.com/abstract=1762118


h ³ ´ i (µˆ1 − µˆ2 )qr (1 − qr ) b
dqr = − λˆ1 + λˆ2 qr + λˆ2 dr + dBr , qt = q, (17)
σ

where

br = − d log(Sˆr ) − [(µˆ1 − µˆ2 )qr + µˆ2 − σ 2 /2]dr


dB . (18)
σ

Since the form of the system of stochastic equations (16) and (17) is the same as (3) and (5),
we conclude that the optimal trading strategy when allowing only short and flat has the same form
as that of the long and flat only case. Moreover, the thresholds for qr determined by (17) and (18)
can be calculated using the same numerical procedure described in Section 4 with the parameter
(15).
The next logical step is to allow both long and short along with flat positions in trading.
Following the method we have used so far, to analyze trading with long, short and flat positions we
would need to add a new value function V−1 corresponding to start with a short position which will
considerably complicate the analysis. An alternative is to consider the following approximation.
First assume there are two traders A and B. Trader A trades long and flat only and trader B
trades short and flat only. As discussed before, trader A can use the method in Section 3 to find
two thresholds p∗b and p∗s and to make buying and selling decisions when pr cross those thresholds.
Trader B can similarly determine two thresholds qb∗ and qs∗ for buying the inverse ETF and going
flat when qr cross those thresholds. Furthermore, we can use the relationship q = 1 − p to translate
the thresholds for qr to that of pr . Namely, when pr crosses 1 − qb∗ from above trader B should
short the index and when pr crosses 1 − qs∗ from below trader B should go flat. Finally, we combine
the action of A and B. Let us use the parameters in Table 1 as an illustration. In this case the four
thresholds have the following order:

1 − qb∗ = 0.589 < p∗s = 0.774 < 1 − qs∗ = 0.849 < p∗b = 0.947. (19)

Now we consider the net combined position of trader A and B. If we start with a long position one
will sell to flat first when pr crosses p∗s from above. If market deteriorates further then one will sell
short when pr crosses 1 − qb∗ eventually. Symmetrically, if the starting position is short one will
cover the short first as market improves and pr crosses 1 − qs∗ , and a long position will be initiated
when the uptrend further strengthes such that pr crosses p∗b .

15

Electronic copy available at: https://ssrn.com/abstract=1762118


We turn to simulations and tests with market data. We use actual short selling with 50%
margin requirement. Thus, if the trend following strategy signals us to shot at r = s and cover at
r = c, we record a gain of

Ss + Ss (1 − K) − Sc (1 + K)
(see Appendix 7.3) (20)
Ss

rather than simulate inverse ETF with a gain Ss /Sc . This is because the inverse ETF works well
as an approximation of sell only in short term while our trend following trading strategy tends to
have relative long holding period for a position.
Again, we run the 5000 round simulation for 10 times and summarize the mean and standard
deviation in Table 5.

Trend Following Buy and Hold No. of Trades


Mean 180.34 (13.9%) 5.77 (4.5%) 66.46
Stdev 1.61 0.168 0.124

Table 5. Statistics of ten 5000-path simulations allowing shorts

Comparing Table 5 and Table 2 we see that adding shorts does improve the performance consid-
erably. However, it seems that the gain from short selling is less than that from long. This is not
surprising given that the market is biased to the up side in the long run.
We now turn to market tests. Ex-ante test of SSE from 2001 to 2011 with long, short and
flat positions yields an annualized return of 18.48% with a quarterly Sharpe ratio 0.306. This
represents a significant improvement over the trend following strategy using only long and flat
positions. However, a similar test for SP500 using the trend following strategy with long, short
and flat positions from 1972 to 2011 only gives an annualized return of 2.57% which is worse than
the annualized return of 8.57% using the buy and hold strategy. Thus, using the trend following
strategy to sell index short may lose money. This is not entirely surprising. It is well known for
practitioners that shorting a market is treacherous (see e.g. [16]).

6 Conclusion

We consider a finite horizon investment problem in a bull-bear switching market, where the drift
of the stock price switches between two parameters corresponding to an uptrend (bull market) and
a downtrend (bear market) according to an unobservable Markov chain. Our target is to maximize

16

Electronic copy available at: https://ssrn.com/abstract=1762118


the expected return of the terminal wealth. We start by restricting to allowing flat and long
positions only and use a sequence of stopping times to indicate the time of entering and exiting
long positions. Assuming trading all available funds, we formulate the problem as a system of
HJB equations satisfied by two value functions that are associated with long and flat positions,
respectively. The system leads to two threshold curves that stand for the optimal buy and sell
boundaries. We show that the optimal trading strategy is trend following and is characterized by
the conditional probability in the uptrend crossing the buy and sell boundaries. We also examine
trading strategies with short selling in terms of an approximation.
We carry out extensive simulations and empirical experiments to investigate the efficiency of
our trading strategies. Here we have a number of interesting observations. First, simulations show
that, somewhat surprisingly, the optimal trend following strategy only yields similar performance
to the suboptimal trend following strategy derived in [4] where the conditional probability in bull
market was also used to signal trading opportunities. This fact also demonstrates that using
conditional probability in bull market as trade signals is robust and is insensitive to parameter
perturbations. Second, the performances of the trend following strategy on individual simulated
paths are dramatically different although averaging the results over large number of paths is rather
stable. In almost all cases the trend following strategy apparently outperforms the buy and hold
strategy. Third, the ex ante experiments with market data reveals that our strategy is efficient not
only in U.S market (SP500 index) but also in China market (SSE index). Last but not the least,
we observe an interesting divergence of the performances of the trend following trading strategy
with short selling. Adding short selling significantly improves the performance in simulations but
the performance in tests using the market historical data is mixed. In some cases it helps, and in
some cases it actually hinders the performance. This is not entirely surprising because it is well
known among practitioners that making money with short selling is difficult. This is an indication
that up and downtrends in the real market are not symmetric and our regime switching model may
only be a crude approximation to the real markets.

7 Appendix
7.1 Proofs of Results

Proof of Lemma 1. It is clear that the lower bounds for Vi follow from their definition. It remains

17

Electronic copy available at: https://ssrn.com/abstract=1762118


to show their upper bounds. Using (6) and (8) and noticing 0 ≤ pr ≤ 1, we have
µ ¶ ·Z vn ¸
Sv
E log n = E f (pr )dr
Sτn τn
µ ¶ Z vn µ ¶
σ2 σ2
≤ µ1 − dr = µ1 − (vn − τn ).
2 τn 2

Note that log(1 − Ks ) < 0 and log(1 + Kb ) > 0. It follows that


( ∞ ·µ ¶ ¸)
X σ2
J0 (S, p, t, Λ0 ) ≤ Et ρ(τ1 − t) + µ1 − (vn − τn ) + ρ(τn+1 − vn )
2
n=1
½ ¾
σ2
≤ max ρ, µ1 − (T − t)
2
µ ¶
σ2
= µ1 − (T − t),
2

where the last equality is due to (2). We then obtain the desired result. Similarly, we can show the
inequality for V1 . 2
Proof of Theorem 2. Denote Z(p, t) ≡ V1 (p, t) − V0 (p, t). It is not hard to verify Z (p, t) is the
unique strong solution to the following double obstacle problem:

min {max {−LZ − f (p) + ρ, Z − log (1 + Kb )} , Z − log(1 − Ks )} = 0,

in (0, 1) × [0, T ), with the terminal condition Z(p, T ) = log (1 − Ks ) . Apparently ∂t Z|t=T ≤ 0,
which implies by the maximum principle

∂t Z ≤ 0. (21)

Thanks to f 0 (p) = µ1 − µ2 > 0, we again apply the maximum principle to get

∂p Z ≥ 0. (22)

By (22), we immediately infer the existence of p∗s (t) and p∗b (t) as given in (9) and (10). Their
monotonicity can be deduced from (21).
Now let us prove part i. If (p, t) ∈ SR, i.e. Z(p, t) = log (1 − K) , then

0 ≤ −L (log (1 − K)) − f (p) + ρ = − (µ1 − µ2 ) p − µ2 + σ 2 /2 + ρ,

namely,
ρ − µ2 + σ 2 /2
p≤ ,
µ1 − µ2

18

Electronic copy available at: https://ssrn.com/abstract=1762118


which is desired. The proofs of part ii and iii as well as the smoothness of p∗s (t) and p∗b (t) are similar
to those in [4]. 2
Proof of Theorem 3. First, note that −Lw0 − ρ ≥ 0. Using Dynkin’s formula and Fatou’s lemma
as in Øksendal [15, p. 226], we have, for any stopping times t ≤ θ1 ≤ θ2 , a.s.,

Ew0 (pθ1 , θ1 ) ≥ E[ρ(θ2 − θ1 ) + w0 (pθ2 , θ2 )]. (23)

Similarly, using −Lw1 − f (p) ≥ 0, we have


·Z θ2 ¸ · ¸
Sθ2
Ew1 (pθ1 , θ1 ) ≥ E f (pr )dr + w1 (pθ2 , θ2 ) = E log + w1 (pθ2 , θ2 ) . (24)
θ1 Sθ1

We next show, for any Λ1 and k = 1, 2, . . .,


"
Svk+1
Ew0 (pvk , vk ) ≥ E ρ(τk+1 − vk ) + log
Sτk+1
# (25)
+w0 (pvk+1 , vk+1 ) + (log(1 − Ks ) − log(1 + Kb ))I{τk+1 <T } .

In fact, using (23) and (24) and noticing that

w0 ≥ w1 − log(1 + Kb ) and w1 ≥ w0 + log(1 − Ks ),

we have
w0 (pvk , vk )
≥ E[ρ(τk+1 − vk ) + ¡w0 (pτk+1 , τk+1 )] ¢
≥ E[ρ(τk+1 − vk ) + µw1 (pτk+1 , τk+1 ) − log(1 + Kb ) I{τk+1 <T } ] ¶
Sv
≥ E[ρ(τk+1 − vk ) + log k+1 + w1 (pvk+1 , vk+1 ) − log(1 + Kb ) I{τk+1 <T } ]
µ Sτk+1 ¶
Svk+1
≥ E[ρ(τk+1 − vk ) + log + w0 (pvk+1 , vk+1 ) + log(1 − Ks ) − log(1 + Kb ) I{τk+1 <T } ]
Sτk+1
Sv
= E[ρ(τk+1 − vk ) + log k+1 + w0 (pvk+1 , vk+1 ) + (log(1 − Ks ) − log(1 + Kb )) I{τk+1 <T } ].
Sτk+1
Note that the above inequalities also work when starting at t in lieu of vk , i.e.,
· ¸
Sv1
Ew0 (pt , t) ≥ E ρ(τ1 − t) + log + w0 (pv1 , v1 ) + (log(1 − Ks ) − log(1 + Kb ))I{τ1 <T } .
Sτ1

Use this inequality and iterate (25) with k = 1, 2, . . ., and note w0 ≥ 0 to obtain

w0 (p, t) ≥ V0 (p, t).

Similarly, we can show that


· ¸ · ¸
Sv 1 Sv1
Ew1 (pt , t) ≥ E log + w1 (pv1 , v1 ) ≥ E log + w0 (pv1 , v1 ) + log(1 − Ks ) .
St St

19

Electronic copy available at: https://ssrn.com/abstract=1762118


Use this and iterate (25) with k = 1, 2, . . . as above to obtain

w1 (p, t) ≥ V1 (p, t).

Finally, it is easy to check that the equalities hold when τk = τk∗ and vk = vk∗ . This completes
the proof. 2

20

Electronic copy available at: https://ssrn.com/abstract=1762118


7.2 Yearly parameters and thresholds in ex-ante tests
date µ1 µ2 λ1 λ2 σ
02-May-1972 0.2780 -0.3400 0.8700 1.4600 0.1042
02-Jan-1973 0.2780 -0.3400 0.8700 1.4600 0.0800
02-Jan-1974 0.2439 -0.3400 1.0600 1.4600 0.1579
02-Jan-1975 0.2439 -0.3480 1.0600 1.1660 0.2196
02-Jan-1976 0.2439 -0.3480 1.0600 1.1660 0.1526
03-Jan-1977 0.2439 -0.3480 1.0600 1.1660 0.1109
03-Jan-1978 0.2439 -0.3480 1.0600 1.1660 0.0915
02-Jan-1979 0.2439 -0.3480 1.0600 1.1660 0.1258
02-Jan-1980 0.2439 -0.3480 1.0600 1.1660 0.1106
02-Jan-1981 0.2439 -0.3480 1.0600 1.1660 0.1633
04-Jan-1982 0.2439 -0.3480 1.0600 1.1660 0.1357
03-Jan-1983 0.2096 -0.2905 0.7607 0.9727 0.1843
03-Jan-1984 0.2096 -0.2905 0.7607 0.9727 0.1370
02-Jan-1985 0.2096 -0.2905 0.7607 0.9727 0.1288
02-Jan-1986 0.2096 -0.2905 0.7607 0.9727 0.1011
02-Jan-1987 0.2096 -0.2905 0.7607 0.9727 0.1493
04-Jan-1988 0.2217 -0.2905 0.5731 0.9727 0.3231
03-Jan-1989 0.2217 -0.6240 0.5731 1.8316 0.1678
02-Jan-1990 0.2217 -0.6240 0.5731 1.8316 0.1314
02-Jan-1991 0.2217 -0.6240 0.5731 1.8316 0.1588
02-Jan-1992 0.2217 -0.6240 0.5731 1.8316 0.1422
04-Jan-1993 0.2217 -0.6240 0.5731 1.8316 0.0969
03-Jan-1994 0.2217 -0.6240 0.5731 1.8316 0.0860
03-Jan-1995 0.2217 -0.6240 0.5731 1.8316 0.0983
02-Jan-1996 0.2217 -0.6240 0.5731 1.8316 0.0781
02-Jan-1997 0.2217 -0.6240 0.5731 1.8316 0.1179
02-Jan-1998 0.2217 -0.6240 0.5731 1.8316 0.1810
04-Jan-1999 0.2217 -0.6240 0.5731 1.8316 0.2024
03-Jan-2000 0.2217 -0.6240 0.5731 1.8316 0.1805
02-Jan-2001 0.2217 -0.6240 0.5731 1.8316 0.2232
02-Jan-2002 0.2033 -0.5113 0.4091 1.4462 0.2142
02-Jan-2003 0.3674 -0.5692 1.4394 1.5773 0.2619
02-Jan-2004 0.3674 -0.5692 1.4394 1.5773 0.1704
03-Jan-2005 0.3674 -0.5692 1.4394 1.5773 0.1110
03-Jan-2006 0.3674 -0.5692 1.4394 1.5773 0.1039
03-Jan-2007 0.3674 -0.5692 1.4394 1.5773 0.1001
02-Jan-2008 0.3674 -0.5692 1.4394 1.5773 0.1599
02-Jan-2009 0.2919 -0.5748 1.0236 1.3483 0.4107
04-Jan-2010 0.8449 -1.0038 3.4824 2.8989 0.2726
03-Jan-2011 0.8449 -1.0038 3.4824 2.8989 0.1810

Table 6. Yearly parameters for the SP500 index ex-ante test

21

Electronic copy available at: https://ssrn.com/abstract=1762118


Date p∗s p∗b
02-May-1972 0.428 0.754
02-Jan-1973 0.391 0.774
02-Jan-1974 0.524 0.764
02-Jan-1975 0.580 0.770
02-Jan-1976 0.525 0.770
03-Jan-1977 0.481 0.783
03-Jan-1978 0.453 0.794
02-Jan-1979 0.498 0.777
02-Jan-1980 0.480 0.783
02-Jan-1981 0.535 0.769
04-Jan-1982 0.509 0.774
03-Jan-1983 0.566 0.768
03-Jan-1984 0.523 0.771
02-Jan-1985 0.514 0.773
02-Jan-1986 0.481 0.784
02-Jan-1987 0.535 0.769
04-Jan-1988 0.658 0.791
03-Jan-1989 0.656 0.873
02-Jan-1990 0.619 0.879
02-Jan-1991 0.648 0.874
02-Jan-1992 0.631 0.877
04-Jan-1993 0.570 0.890
03-Jan-1994 0.548 0.895
03-Jan-1995 0.572 0.890
02-Jan-1996 0.530 0.900
02-Jan-1997 0.602 0.883
02-Jan-1998 0.667 0.872
04-Jan-1999 0.683 0.872
03-Jan-2000 0.666 0.872
02-Jan-2001 0.698 0.873
02-Jan-2002 0.685 0.860
02-Jan-2003 0.572 0.769
02-Jan-2004 0.510 0.775
03-Jan-2005 0.449 0.797
03-Jan-2006 0.439 0.802
03-Jan-2007 0.433 0.804
02-Jan-2008 0.501 0.777
02-Jan-2009 0.739 0.856
04-Jan-2010 0.448 0.703
03-Jan-2011 0.395 0.725

Table 7. Yearly thresholds for the SP500 index ex-ante test

22

Electronic copy available at: https://ssrn.com/abstract=1762118


date µ1 µ2 λ1 λ2 σ
02-Jan-2001 1.0000 -1.0000 1.0000 1.0000 0.1992
01-Jan-2002 0.7259 -1.0000 1.3846 1.0000 0.2126
01-Jan-2003 0.6694 -0.9199 1.6290 1.1950 0.2333
01-Jan-2004 0.6694 -0.7869 1.6290 1.4478 0.1707
03-Jan-2005 0.5279 -0.7869 1.3429 1.4478 0.1977
02-Jan-2006 0.5279 -0.6607 1.3429 1.2205 0.2124
01-Jan-2007 0.5279 -0.6607 1.3429 1.2205 0.2086
01-Jan-2008 0.6185 -0.6607 1.0374 1.2205 0.3399
05-Jan-2009 4.8903 -1.8062 23.4833 4.8778 0.4485
04-Jan-2010 3.6021 -2.3256 16.1120 7.6729 0.3017
04-Jan-2011 2.7612 -1.7366 12.2969 5.6829 0.2270

Table 8. Yearly parameters for the SSE index ex-ante test

Date p∗s p∗b


02-Jan-2001 0.361 0.682
01-Jan-2002 0.455 0.742
01-Jan-2003 0.474 0.738
01-Jan-2004 0.404 0.721
03-Jan-2005 0.487 0.759
02-Jan-2006 0.464 0.721
01-Jan-2007 0.462 0.722
01-Jan-2008 0.483 0.679
05-Jan-2009 0.186 0.442
04-Jan-2010 0.251 0.592
04-Jan-2011 0.233 0.604

Table 9. Yearly thresholds for the SSE index ex-ante test

7.3 Return on short selling

Assuming that the signals from the trend following strategy indicate short selling at r = s and then
cover at r = c and that the margin requirement for short selling is α, we calculate the return on
the above trade when trading full margin. Let w be the wealth at r = s. Suppose we can short sell
k shares on full margin. Then kSs = α(w + kSs ) or w = kSs 1−α
α . When we cover at r = c the net

gain is kSs (1 − K) − kSc (1 + K) taking into account of the trading cost. Thus, the return is

w + kSs (1 − K) − kSc (1 + K) Ss (1 − α) + Ss (1 − K)α − Sc (1 + K)α


= .
w Ss (1 − α)

When α = 50% = 1/2 we get (20).

23

Electronic copy available at: https://ssrn.com/abstract=1762118


References

[1] M. Dai, H. Jin, Y. Zhong and X. Y. Zhou, Buy low and sell high, Contemporary Quantita-
tive Finance: Essays in Honour of Eckhard Platen, edited by Chiarella, Carl and Novikov,
Alexander, Springer, pp. 317-334, (2010).

[2] M. Dai, H.F. Wang, and Z. Yang, Leverage management in a bull-bear switching market,
SSRN, http://ssrn.com/abstract=1571613, (2010)

[3] M. Dai and F. Yi, Finite horizontal optimal investment with transaction costs: a parabolic
double obstacle problem, Journal of Diff. Equ., 246, pp. 1445-1469, (2009).

[4] M. Dai, Q. Zhang, and Q. Zhu, Trend following trading under a regime switching model, SIAM
Journal on Financial Mathematics, 1, pp. 780-810, (2010).

[5] M.H.A. Davis and A.R. Norman, Portfolio selection with transaction costs, Mathematics of
Operations Research, 15, 676-713, (1990).

[6] M.T. Faber, A quantitative approach to tactical asset allocationno access, 9, pp. 69C79, (2007).

[7] Federal Reserve, Federal Reserve Statistic Release Website,


http : //www.f ederalreserve.gov/releases/h15/data/Annual/H15 T CM N OM Y 10.txt.

[8] A. Friedman, Variational Principles and Free-Boundary Problems, Wiley, New York, (1982).

[9] B.G. Jang, H.K. Koo, H. Liu, and M. Loewenstein, Liquidity Premia and transaction costs,
Journal of Finance, 62, pp. 2329-2366, (2007).

[10] S. Karlin and H.M. Taylor, A Second Course in Stochastic Processes, Academic Press, New
York, 1981.

[11] R.S. Liptser and A.N. Shiryayev, Statistics of Random Processes I. General Theory, Springer-
Verlag, Berlin, New York, 1977.

[12] H. Liu and M. Loeweinstein, Optimal portfolio selection with transaction costs and finite
horizons, Review of Financial Studies, 15, pp. 805-835, (2002).

24

Electronic copy available at: https://ssrn.com/abstract=1762118


[13] M.J.P. Magill and G.M. Constantinides, Portfolio selection with transaction costs, Journal of
Economic Theory, 13, pp. 264-271, (1976).

[14] R.C. Merton, Optimal consumption and portfolio rules in a continuous time model, Journal
of Economic Theory, 3, pp. 373-413, (1971).

[15] B. Øksendal, Stochastic Differential Equations 6th ed. Springer-Verlag, Berlin, New York,
2003.

[16] W. J. O’Neal, How to Make Money in Stocks: A Winning System in Good Times or Bad, 2nd
edition, McGraw-Hill, New York, 1995.

[17] R. Rishel and K. Helmes, A variational inequality sufficient condition for optimal stopping
with application to an optimal stock selling problem, SIAM J. Control and Optim., Vol. 45,
pp. 580-598, (2006).

[18] A. Shiryaev, Z. Xu and X. Y. Zhou, Thou shalt buy and hold, Quantitative Finance, 8, pp.
765-776, (2008).

[19] S.E. Shreve and H.M. Soner, Optimal investment and consumption with transaction costs,
Annals of Applied Probability, 4, 609-692, (1994).

[20] Q.S. Song, G. Yin and Q. Zhang, Stochastic optimization methods for buyinglow- and-selling-
high strategies, Stochastic Analysis and Applications, 27, pp. 52-542, (2009).

[21] Wikipedia on trend following, http://en.wikipedia.org/wiki/Trend following.

[22] W.M. Wonham, Some applications of stochastic differential equations to optimal nonlinear
filtering, SIAM J. Control, 2, pp. 347-369, (1965).

[23] M. Zervos, T.C. Johnsony and F. Alazemi, Buy-low and sell-high investment strategies, Work-
ing paper, (2011).

[24] H. Zhang and Q. Zhang, Trading a mean-reverting asset: Buy low and sell high, Automatica,
44, pp. 1511-1518, (2008).

25

Electronic copy available at: https://ssrn.com/abstract=1762118

You might also like