Assessing the profitability of intraday opening range breakout

strategies

Ulf Holmberg, Carl Lonnbark, Christian Lundstrom

Department of Economics
Ume
a School of Business and Economics
Ume
a University
SE-901 87 Ume
a

Abstract

Is it possible to beat the market by mechanical trading rules based on historical and publicly
known information? Such rules have long been used by investors and in this paper, we test
the success rate of trades and profitability of the Open Range Breakout (ORB) strategy. An
investor that trades on the ORB strategy seeks to identify large intraday price movements
and trades only when the price moves beyond some predetermined threshold. We present
an ORB strategy based on normally distributed returns to identify such days and find that
our ORB trading strategy result in significantly higher returns than zero as well as an
increased success rate in relation to a fair game. The characteristics of such an approach
over conventional statistical tests is that it involves the joint distribution of Low, High, Open
and Close over a given time horizon.

Keywords: Bootstrap, Crude oil futures, Contraction-Expansion principle, Efficient market

hypothesis, Martingales, Technical Analysis.

JEL classification: C49, G11, G14, G17.

The second author gratefully acknowledges the financial support from the Wallander foundation. We thank
Kurt Br
ann
as and Tomas Sjogren for insightful comments and suggestions.
1 Introduction

The Efficient Market Hypothesis (EMH) of Fama (1965, 1970) asserts that current asset prices

fully reflect available information (see also Fama, 1991) implying that asset prices evolve as

random walks in time. Consequently, tests of the EMH have traditionally been designed to

catch deviations from random walk prices and in the massive literature on the subject one

is bound to find support for both acceptances and rejections of the hypothesis (e.g., Malkiel,

1996; Lo, 2001). In particular, an assertion of the EMH is that it should not be possible to

base a trading strategy on historical prices (so-called filter rules or technical trading) and earn

positive expected returns. However, the fact remains that the use of filter rules is a widespread

phenomenon. Barclay Hedge estimates that filter based Hedge Funds within the Managed

Futures category manage over 300 Billion USD in 2011 and is today the largest hedge fund

category with respect to assets under management. Indeed, some filter rule traders appear

to consistently outperform the market (see Schwager, 1989, for a classic reference) and the

subject has been given due attention in the literature (e.g. Brock, Lakonishok, and LeBaron,

1992; Gencay, 1996, 1998). Testing of the profitability of trading rules has traditionally been

carried out based on a (at least) daily investment horizon. However, as discussed in Taylor and

Allen (1992) the use of filter rules among practitioners appears to increase with the frequency

of trading (see also Schulmeister, 2009). In particular, many strategies are typically employed

intraday and to assess their potential profitability one would typically require intraday data.

The relative unavailability of intraday data may thus be a possible explanation for the apparent

lagging behind of the research community.

In this paper we remove this obstacle and propose a quite novel approach on how to assess

the profitability when only records of daily high, low, opening and close are available. Obviously,

there is a plethora of filter rules out there and the one we have in mind in the present paper is

the so-called opening range breakout (ORB), which is typically adopted intraday. This rule is

based on the premise that if the market moves a certain percentage from the opening price level,

the odds favor a continuation of that move. An ORB filter suggests that, long (short) positions

are established at some predetermined price threshold a certain percentage above (below) the

opening price.

1
 To evoke the testing strategy and gain intuition on the way we first note that the rationale

behind using an ORB filter is the believe in so-called momentum in prices (e.g. Jegadeesh and

Titman, 1993). That is, the tendency for rising asset prices to rise further and falling prices to

keep falling. In the behavioral finance literature the appearance of momentum is often attributed

to cognitive biases from irrational investors such as investor herding, investor over- and under

reaction, and confirmation bias (see Barberis et al., 1998; Daniel et al., 1998). However, as

discussed in Crombez (2001) momentum can also be observed with perfectly rational traders.

In pioneering the ORB strategy Crabel (1990) presented the so-called Contraction-Expansion

(C-E) principle. The principle asserts that markets alternates between regimes of contraction

and expansion, or, periods of modest and large price movements, respectively. An ORB strategy

may be viewed as a strategy of identifying and profiting from days of expansion. In passing we

note the resemblance with the stylized fact of volatility clustering in financial return series (e.g.

Engle, 1982).

Now, a seemingly quite reasonable assumption is that markets for the most part are relatively

efficient with prices evolving as random walks in time, or equivalently, returns are martingales.

Thus, a heuristic use of the law of large number implies normally distributed returns. According

to the (C-E) principle these calm days could be considered as periods of contraction during

which the returns are normally distributed. Now, during periods of expansions traders activates

ORB strategies and the profitability of them implies that the martingale property breaks down

with non-normality as a consequence. Building on this reasoning our testing strategy is simply

based on identifying days of large intraday movements and evaluating the expected return on

these days. In particular, if on a given day the price threshold implied by the rule is above

(below) the high (low) price we deduce that a long (short) position was established at some

point during this day. To assess statistical significance we build on Brock et al. (1992) and use

a bootstrap approach adapted to the present case.

The remainder of the paper is organized as follows. In Section 2 we briefly review the

underlying theory and give an account of the ORB strategy. In this section we also outline

our proposed test for profitability. Section 3 gives results for the empirical application and the

fourth section concludes.

2
2 Martingale prices and momentum based trading strategies

We denote by Pto ,Pth , Ptl and Ptc the opening, high, low and, closing price on day t, respectively.

A point in time on day t is given by t + , 0 1. Note that Pto = Pt and Ptc = Pt+1 . The set

t+ contains the information available at time t + . Furthermore, let u ( l ) denote a certain

threshold price level that is such that if the price crosses it from below (above) a momentum

investor acts, i.e. takes a long (short) position. For ORB investors, these threshold price are

often set in terms of some pre-determined (large) relative change, , from the opening price

such that tu = (1 + )Pto and tl = (1 )Pto . For the purpose of this paper we assume that

all positions are closed at the end of the trading day. Hence, no type of money management

techniques such as a stop loss, trailing loss, profit stop are considered.

Within the context of the present paper it is natural to involve the martingale pricing model

(MPT) of Samuelson (1965). If capital markets are efficient with respect to t+ some pre-

scribed formula based on t+ should not result in systematic success implying that prices are

martingales with respect to this information set. In particular,

E[Ptc |t+ ] = Pt+ .

A direct consequence of martingale pricing is that any investment should earn a zero expected

return
c
E[Rt+ |t+ ] = 0,

c
where Rt+ = log (Ptc /Pt+ ). As such, any investment within the MPT framework is a fair

game and from the martingale central limit theorem it follows that the returns are normally

distributed (Brown, 1971).

Now, momentum investments are based on the premise that, if the market moves a certain

percentage from the opening price level, the odds favor a continuation of that move. More

specifically, a profitable momentum based trading strategy implies that

E[Ptc |Pt+ > tu ] > Pt+ and/or E[Ptc |Pt+ < tl ] < Pt+ .

3
 Pt"

c
Pt

ut

Pot

lt

Figure 1: An ORB strategy trader enters a long position if the intraday price exceeds tu .

As such, the breaking down of the martingale property implies that the martingale central limit

theorem no longer applies. Thus, it is natural to define as a daily return that is unlikely to

occur given normally distributed returns

=
+
q , (1)

where are estimates of the mean and standard deviation of Rtc = log (Ptc /Pto ), respec-
and

tively, and q the inverse of the standard normal cumulative distribution function evaluated at

. Figure 1 illustrates a profitable intraday trade based an ORB strategy. The price opens

at Pto and as long as the price stays within normal bounds, i.e. within (tu , tl ), the trader

refrains from action but as soon as Pt+ = tu , the trader initiates a long position, anticipating

a continuation of the price moving in the same direction.

Given that an ORB strategy is based on intraday price movements, as illustrated in Figure

1, it is clear that a perfect test of profitability requires information on the intraday price paths.

The challenge we take on here is that of designing a test with access only to records of daily

opening, high, low and closing prices. Our basic observation is that if the daily high (low) is

higher (lower) than the set tu (tl ), we know with certainty that a buy (sell) signal was triggered

at some point during the day and that a position was initiated at tu (tl ). For the purpose of

this paper we assume a perfect order fill at the threshold price, a zero bid ask spread, as well as

4
zero commissions. Consequently, real-life trading produce slightly different results.
! "
Upon defining the return series Rtlong = log (Ptc /tu ) and Rtshort = log Ptc /tl we may con-

sider the averages

#
long = 1(P h > u )Rtlong
R # t h ,
1(Pt > u )
#
short 1(P l < l )Rtshort
R = # t l ,
1(Pt < l )

long
where 1() is the indicator function. If strategies based on ORB filters are profitable then R
short should be significantly larger than zero. To assess statistical significance we rely on
and R

the bootstrap approach suggested in Brock et al. (1992). Here, we face additional challenges

compared to their work as the case at hand is multivariate with a natural ordering of the level

series. A reasonable procedure that accommodates this restriction proceeds as follows.

Assume that the level series share a common trend (cf. co-integration). Hence, considering

a benchmark series to bootstrap the general levels appears reasonable. The other series may

then be obtained as bootstrapped deviations from the benchmark series. To this end we consider

the daily opening price as the benchmark series and define Rto = log(Pto /Pt1
o ), t = 2, .., T . Also

define deviations Rti = log(Pti /Pto ) for i = {h, l, c} and t = 1, .., T . Collect these returns in

Rt = (Rto , Rth , Rtl , Rtc ) are then drawn randomly with replacement, generating an pseudo-sample

of returns. Based on this sample, an alternative realization of the level series is then generated.
long and R
This procedure is repeated N times to generate sampling distributions of R short

respectively. The sampling distributions are then used in the standard way to test the null of

zero expected returns against the alternative of positive ones.

3 Application

We apply the testing strategy presented above to a time series of U.S. crude oil futures prices

obtained from Commodity Systems Inc covering the period March 30, 1983 to January 26, 2011.

When constructing the time series the switch from the near-by contract to the next typically

occur around the 20th each month, one month prior to the expiration month (see Pelletier, 1997,

5
 100 125 150 175
Open price

75
50

1983 03 30 1990 03 09 1997 02 19 2004 02 13 2011 01 26

Figure 2: The evolution of the daily open price for U.S. crude oil futures adjusted for roll-over
effects from March 30, 1983 to January 26, 2011. Source: Commodity Systems Inc.

Table 1: Descriptives of the daily return series.

Obs. Mean Std.Dev Min Max Skewness Kurtosis Jarque-Bera

6976 0.02 0.72 -6.06 9.90 0.16 10.26 30668

for details on the adjustment of roll-over effects). Commodity futures are as easily sold short

as bought long, and are not subject to short-selling restrictions while the costs associated with

trading (e.g. slippage, bid ask spreads, and commissions) are often relatively low. In Figure

2 we plot the evolution of the level series. The series exhibit a cyclical pattern and follows a

positive long run trend reasonably due to inflation. Notable is also the sharp drop during the

2008 sub-prime crisis.

In Table 1 we give some descriptives for the daily returns series, i.e. Rtc . The series exhibit

positive skewness and excess kurtosis and consequently the Jarque-Bera test strongly rejects

normality.

The values of the s (and consequently the treshold prices) are derived from the sample.

We thus check ex post for the existence of intraday trending of oil futures prices.

As can be read in Table 2, the ORB strategy results in significant positive average returns

suggesting that the fair game argument embedded in the Martingale pricing theory does

not hold true for adverse price movements. Interestingly, as we tighten the criterion used to

determine entry, i.e. if we move further down the tail of a normal distribution, both the success

rate and average returns increase. Figure 3 clarifies this relationship. However, it should be

6
Table 2: Empirical results. The is the tail probability, and gives the associated percentage
return. N is the number of trades. freq gives the proportion of trades that result in positive
gives the average returns.
returns, while R

Long Short
N f req. Rlong p N f req. R short p
10% 0.9388 738 0.6057 0.2019 0.0000 -0.9013 826 0.5424 0.1439 0.0000
5% 1.1996 439 0.6036 0.2180 0.0000 -1.1621 497 0.5714 0.1784 0.0000
full sample 1% 1.6889 188 0.6117 0.2583 0.0001 -1.6513 224 0.6205 0.2442 0.0003
0.5% 1.8680 141 0.6028 0.3108 0.0002 -1.8304 172 0.6454 0.2527 0.0008
0.1% 2.2373 80 0.7125 0.4027 0.0010 -2.1997 98 0.6225 0.2489 0.0147

N f req. Rlong p N f req. R short p

10% 0.7840 260 0.4923 0.0334 0.2539 -0.7574 272 0.5368 0.0871 0.0430
1983-03-30 5% 1.0024 159 0.5157 0.0711 0.1350 -0.9759 156 0.5192 0.1313 0.0401
to 1% 1.4122 72 0.4861 0.1140 0.1246 -1.3857 73 0.5753 0.1978 0.0563
1992-06-29 0.5% 1.5623 57 0.4912 0.0799 0.2467 -1.5357 56 0.5893 0.2420 0.0494
0.1% 1.8716 33 0.5758 0.1656 0.1448 -1.8451 41 0.6342 0.1026 0.2859

N f req. Rlong p N f req. short

R p
10% 0.6069 373 0.5657 0.0374 0.0357 -0.5947 371 0.5148 0.0307 0.0734
1992-06-30 5% 0.7772 195 0.5795 0.0634 0.0196 -0.7650 214 0.5327 0.0228 0.2172
to 1% 1.0966 62 0.5807 0.0843 0.0357 -1.0845 79 0.5317 -0.0258 0.6814
2001-10-11 0.5% 1.2136 53 0.3962 0.0068 0.4546 -1.2015 57 0.5790 -0.0608 0.8091
0.1% 1.4548 20 0.5000 0.0254 0.4061 -1.4426 27 0.3333 -0.0290 0.6420

N f req. Rlong p N f req. R short p

10% 1.2956 245 0.6612 0.2813 0.0000 -1.2216 300 0.5967 0.2483 0.0000
2001-10-12 5% 1.6524 138 0.6522 0.3405 0.0004 -1.5784 177 0.6328 0.2734 0.0001
to 1% 2.3216 50 0.8000 0.5155 0.0063 -2.2477 64 0.6406 0.3879 0.0006
2011-01-26 0.5% 2.5667 44 0.7500 0.4926 0.0062 -2.4927 48 0.6667 0.3892 0.0008
0.1% 3.0718 23 0.8261 0.6397 0.0096 -2.9979 28 0.7143 0.3763 0.0054

noted that by moving down the tail of the normal distribution, we also reduce the number of

trades, reducing the investors potential profits.

Dividing the full data set into three sub-samples, 1983-03-30 to 1992-06-29, 1992-06-30 to

2001-10-11, and finally 2001-10-12 to 2011-01-26 we find that the most recent time period drives

the result. Given the possible resemblance between the ORB strategy and the stylized fact of

volatility clustering in financial returns series, one plausible explanation is the relatively high

volatility in the 2001-10-12 to 2011-01-26 period. After all, ORB is a directional strategy in the

sense that either a long or a short position is established and hence it is basically long volatility

in contrast to hedge fund strategies such as Long Short Equity, Market Neutral strategies or dif-

ferent variants of Arbitrage strategies to mention a few. Market volatility and ORB profitability

should be expected to go hand in hand.

7
 0.5
Long
Short

0.4
Average return

0.3
0.2
0.1
0.0

0.000 0.025 0.050 0.075 0.100

Figure 3: Average returns on the tail probability ().

4 Concluding discussion

We proposed a way of assessing the profitability of intraday ORB strategies when long records of

daily opening, high, low and closing prices are available. In an empirical application we employ

our testing strategy to U.S. crude oil futures. Using the full sample we find a remarkable success

of the of ORB strategies. However, splitting up the full sample into three sub-periods reveals

that this finding is not robust to time and to a large extent explained by the most recent (and

most volatile) period. In this sense, our results relate to the findings in Gencay (1998), that

mechanical trading rules tend to result in higher profits when markets trend or in times of

high volatility.

A point to note is that our testing strategy will underestimate the actual profits since the

closing of the positions is assumed to occur at the daily close. Thus, days when the momentum

does not carry through to the end of the day or even reverses intraday will be included. In

practice, the losses on these days will be limited by so-called stop losses.

Notable is also the our filter results in relatively few trades, which restricts potential profits.

Most likely though the orb trader simultaneously monitors and acts on several markets.

Admittedly, transaction costs in terms of commission fees and bid-ask spreads will consume

some of the profits. However, for the market under consideration these are relatively small. A

reasonable estimate is 0.04%, or 0.08% round trip.

8
References

Barberis, N., A. Shleifer, and R. Vishny (1998): A Model of Investor Sentiment,

Journal of Financial Economics, 49, 307343.

Brock, W., J. Lakonishok, and B. LeBaron (1992): Simple Technical Trading Rules and

the Stochastic Properties of Stock Returns, Journal of Finance, 47, 173164.

Brown, B. M. (1971): Martingale Central Limit Theorems, The Annals of Mathematical

Statistics, 42, 5966.

Crabel, T. (1990): Day Trading With Short Term Price Patterns Day Trading With Short

Term Price Patterns and Opening Range Breakout, Greenville, S.C. : Traders Press.

Crombez, J. (2001): Momentum, Rational Agents and Efficient Markets, Journal of Psy-

chology and Financial Markets, 2, 190200.

Daniel, K., D. Hirshleifer, and A. Subrahmanyam (1998): Investor Psychology and

Security Market Under- and Overreactions,, Journal of Finance, 53, 18391885.

Engle, R. F. (1982): Autoregressive Conditional Heteroscedasticity with Estimates of the

Variance of United Kingdom Inflation, Econometrica, 50.

Fama, E. (1965): The Behavior of Stock Market Prices, Journal of Business, 38, 34105.

(1970): Efficient Capital Markets: A Review of Theory and Empirical Work, The

Journal of Finance, 25, 383417.

(1991): Efficient Capital Markets: II, Journal of Finance, 46, 1575617.

Genc
ay, R. (1996): Non-linear Prediction of Security Returns with Moving Average Rules,

Journal of Forecasting, 15, 165174.

(1998): The Predictability of Security Returns with Simple Technical Trading Rules,

Journal of Empirical Finance, 5, 347359.

Jegadeesh, N. and S. Titman (1993): Returns to Buying Winners and Selling Losers:

Implications for Stock Market Efficiency, Journal of Finance, 48, 6591.

9
Lo, A. (2001): A Non-random Walk Down Wall Street, Princeton Paperbacks.

Malkiel, B. G. (1996): A Random Walk Down Wall Street, W. W. Norton.

Pelletier, B. (1997): Computed Contracts: Their Meaning, Purpose and Application, CSI

Technical Journal, 13, 16.

Samuelson, P. A. (1965): Proof That Properly Anticipated Prices Fluctuate Randomly,

Industrial Management Review, 6, 4149.

Schulmeister, S. (2009): Profitability of technical stock trading: Has it moved from daily to

intraday data? Review of Financial Economics, 18, 190201.

Schwager, J. D. (1989): Market Wizards: Interviews with Top Traders, HarperCollins pub-

lishers Inc.

Taylor, M. P. and H. Allen (1992): The Use of Technical Analysis in the Foreign Exchange

Market, Journal of International Money and Finance, 11, 304314.

10