Beat the Market

An Effective Intraday Momentum Strategy

for S&P500 ETF (SPY)

Carlo Zarattini1 , Andrew Aziz2,3 , Andrea Barbon4

1 Concretum Research, Piazza Molino Nuovo 8, 6900 Lugano, Switzerland

2 Peak Capital Trading, 744 West Hastings Street, Vancouver, BC, Canada V6C 1A5
3 Bear Bull Traders, 744 West Hastings Street, Vancouver, BC, Canada V6C 1A5
4 University of St.Gallen, Dufourstrasse 50, 9000 St. Gallen, Switzerland

Q 1 c.zarattini@concretumresearch.com, 3 andrew@peakcapitaltrading.com, 4 andrea.barbon@unisg.ch

X: 1 @ConcretumR,2 @BearBullTraders ,4 @Andrea Barbon
May 10, 2024

Abstract

This paper investigates the profitability of a simple, yet effective intraday momen-
tum strategy applied to SPY, one of the most liquid ETFs that tracks the S&P
500. Unlike the academic literature that typically limits trading to the last 30 min-
utes of the trading session, our model initiates trend-following positions as soon
as there is an indication of abnormal demand/supply imbalance in the intraday
price action. Building on trading techniques commonly used by active day traders,
which have been discussed in our previous papers, we introduce the use of dy-
namic trailing stops to mitigate downside risks while allowing for unlimited upside
potential. From 2007 to early 2024, the resulting intraday momentum portfolio
achieved a total return of 1,985% (net of costs), an annualized return of 19.6%, and
a Sharpe Ratio of 1.33. We conduct extensive statistical tests to examine whether
the profitability of the strategy is affected by different market volatility regimes
and whether the estimated gamma imbalance of dealers could predict changes in
strategy profitability. We analyze the daily profitability of the intraday momentum
strategy with respect to day-of-the-week effects. Additionally, we evaluate its per-
formance against well-known technical daily patterns to understand its behavior
under various market conditions. Given the short-term nature of the model, we
also assess the impact of commissions and slippage on the overall profitability of
the strategy.

Keywords: Day Trading, Day Trading Systems, Algo Trading, Momentum, Trend-Following, Intraday
Momentum, Delta-Hedging

1
Electronic copy available at: https://ssrn.com/abstract=4824172
1 Introduction

An object in motion tends to stay in motion,

an object at rest tends to remain at rest.
– Isaac Newton

In a very influential paper, Jegadeesh and Titman [1] introduce a groundbreaking insight
into the realm of stock market anomalies: the momentum effect. Contrary to the efficient
market hypothesis, which suggests that past stock performance is not predictive of future
returns, Jegadeesh and Titman demonstrate that stocks exhibiting strong performance
over recent periods tend to continue outperforming in the short to medium term. The
concept of momentum trading has revolutionized investment strategies by revealing the
persistent trend-following behavior of stock prices. It also contradicts the efficient market
hypothesis, which posits that stock prices reflect all available information and that past
performance cannot predict future performance.

Jegadeesh and Titman’s research shows that a portfolio that goes long the best performing
stocks and shorts the worst performing stocks, generates significant positive returns over
3- to 12-month holding periods and the profitability of these portfolios is not explained
by common risk factors. They provide empirical evidence supporting the profitability
of momentum, which has since become a widely studied and implemented investment
strategy in both academic and practical contexts.

Since then, momentum strategies have been implemented over longer time frames, with
research focusing on daily, weekly, or monthly intervals. However, with the rise of high-
frequency trading and increased accessibility to intraday data, there is growing interest
in applying momentum strategies to shorter time frames, such as intraday intervals for
day trading [2, 3, 4, 5]. Gao et al. [6] demonstrates the potential of overnight returns
to forecast the final half-hour returns of the S&P 500, highlighting significant intraday
time-series momentum. Extending these findings, Rosa [7] analyzes additional data post-
2013, noting a decline in predictability, which he attributes to market adaptations or the

2
Electronic copy available at: https://ssrn.com/abstract=4824172
ephemeral nature of previously exploitable inefficiencies. His application of a Markov-
switching model underlines how the intraday momentum’s effectiveness is contingent upon
volatile market regimes.

The study by Baltussen et al. [8] further enriches this discussion by examining how op-
tions dealers’ gamma hedging activities influences intraday price dynamics. This work
illustrates the complex interplay between institutional hedging and equity market move-
ments, shedding light on how such actions can amplify or mitigate intraday price trends,
thus impacting the performance of momentum strategies.

Other global studies have expanded the scope of intraday momentum across various mar-
kets. For instance, Zhang et al. [9] and Jin et al. [10] confirm the presence of intraday
momentum in Chinese equities and commodities, respectively, while Li et al. [11] provide
evidence across multiple developed markets, underscoring the pervasive yet varied nature
of intraday momentum influenced by localized market behaviors and regulations.

In our previous publications, we present findings from various day trading strategies, in-
cluding the Opening Range Breakout (ORB) [2, 3, 4, 12] and Volume Weighted Average
Price (VWAP) Trend Trading [5], which consistently outperform the simple buy-and-hold
strategy on major US stock market indices such as SPY and QQQ. These results are
robust, exhibiting significant Sharpe Ratios and statistical significance for our straight-
forward day trading approaches.

Building upon our prior research, this paper extends the model of Gao et al. [6] and
Baltussen et al. [8]. Our aim is to study the economic and statistical significance of a day
trading strategy that uses SPY, a very liquid ETF that tracks the S&P500, to exploit
intraday trends caused by sticky demand/supply imbalances. Unlike the aforementioned
studies, our model does not limit the trading to the last 30-minutes of the trading session;
empirically, we have seen that intraday trends can start earlier and a strategy confined
to trading only after 15:30 may prove to be suboptimal.

Our model is an instance of a time-series momentum strategy, focusing on a single in-

3
Electronic copy available at: https://ssrn.com/abstract=4824172
strument, namely the SPY, instead of relying on a portfolio of multiple assets sorted by
their relative past performance. A research conducted by Moskowitz et al. [13] on time-
series momentum indicates that the effect is more pronounced at shorter horizons and
tends to revert over longer time frames. It is, therefore, natural to test the profitability
of time-series momentum at high frequencies, which motivates our intraday study.

While academic literature continues to debate the causes of momentum effects, one of
the most convincing explanations is the under-reaction to news, which causes informa-
tion to be slowly incorporated into prices. This under-reaction may arise from a variety
of factors. For instance, Frazzini [14] proposes and tests an explanation based on a be-
havioral bias called the disposition effect, which is the tendency to capitalize gains from
winning stocks and hold onto losing positions. The flows originating from this biased
behavior may slow down the appreciation of stock prices following positive news and,
symmetrically, decelerate the price drops following negative news. Another explanation
for under-reaction, based on investors’ limited attention to news, is proposed by Della
Vigna and Pollet [15]. Such behavior may stem from limited cognitive abilities or be
optional if information acquisition is costly [16]. Finally, Duffie [17] argues that under-
reaction may stem from slow-moving capital, caused by institutional impediments to fast
trading, such as the search costs for counterparties or raising capital by intermediaries.

Our work contributes to the ongoing discourse surrounding market anomalies and the
practical applications of momentum trading in intraday contexts. Through empirical
analysis and rigorous backtesting, we evaluate the effectiveness of intraday momentum
strategies, shedding light on their potential benefits and challenges in the modern trading
landscape.

Furthermore, our evidence of sizeable momentum effects at an intraday frequency sup-

ports an explanation of momentum based on under-reaction to news, in line with previous
literature on the topic.

4
Electronic copy available at: https://ssrn.com/abstract=4824172
2 Data
The database for SPY and VIX (Chicago Board Options Exchange Volatility Index)
has been constructed using 1-minute OHLCV (Open, High, Low, Close, and Volume)
data from IQFeed, covering the period from May 2007 to April 2024. All backtests and
statistical tests have been conducted using Matlab R2023a.

3 Strategy Description
The existence of intraday trends typically results from a persistent and significant im-
balance between buyers and sellers throughout a trading session. Given the high level
of noise in mature equity markets, such as the S&P 500, an intraday momentum strat-
egy should only establish positions when the price action clearly shows a deviation from
demand-supply equilibrium. To effectively navigate this, we have to mathematically de-
fine a region where prices are expected to oscillate under conditions of balanced buying
and selling pressures. We refer to this region as the Noise Area, an equilibrium zone
where markets do not exhibit any exploitable intraday trend.

Practically, the market demonstrates a lack of abnormal imbalances when the intraday
movement from the Open is less than the average movement observed on the same intra-
day time interval over the previous days. We define the Noise Area as the space between
2 boundaries. These boundaries are time-of-day dependent and are computed using the
average movements recorded over the previous 14 days. Mathematically, the Noise Area
can be compupted on day t following these steps:

1. For each day t − i and time-of-day HH:MM, calculate the absolute move from Open
as
Closet−i,HH:M M
movet−i,9:30−HH:M M = −1 , where i = [1, 14],
Opent−i,9:30

2. For each time-of-day HH:MM, calculate the average move over the last 14 days as

14
1 X
σt,9:30−HH:M M = movet−i,9:30−HH:M M
14 i=1

5
Electronic copy available at: https://ssrn.com/abstract=4824172
 3. Using the Open of day t, compute the Upper and Lower Boundaries as

UpperBoundt,HH:M M = Opent,9:30 × (1 + σt,9:30−HH:M M )

LowerBoundt,HH:M M = Opent,9:30 × (1 − σt,9:30−HH:M M )

4. Compute the Noise Area as the area between the Upper and Lower Boundaries

 
NoiseAreat,HH:MM = LowerBoundt,HH:MM , UpperBoundt,HH:MM

To make the evidence of potential imbalances more robust, we include the closing price in
our calculation. This adjustment accounts for overnight gaps, which in themselves often
signal imbalances. For instance, following a gap-down event, the Upper Boundary of the
Noise Area is adjusted upward by a quantity equal to the gap size, and conversely for a
gap-up event, the Lower Boundary is adjusted downward.

With these considerations, the revised mathematical equations for the Upper and Lower
Boundaries, including adjustments for overnight gaps, are as follows:

UpperBoundt,HH:M M = max(Opent,9:30 , Closet−1,16:00 ) × (1 + σt,9:30−HH:M M )

LowerBoundt,HH:M M = min(Opent,9:30 , Closet−1,16:00 ) × (1 − σt,9:30−HH:M M )

This enhancement ensures that our definition of the Noise Area dynamically reflects both
the intraday fluctuations and the impact of the opening market conditions, providing a
more comprehensive tool for identifying true market imbalances beyond mere noise.

Figure 1 provides a graphical representation of the Noise Area. As long as the market
remains within this designated region, it is considered that demand and supply are in
equilibrium, and thus, the strategy refrains from taking any positions. A crossing above
the Upper Boundary, illustrated by the blue line, indicates abnormal buying pressure, po-
tentially heralding an exploitable upward trend. Conversely, a crossing below the Lower

6
Electronic copy available at: https://ssrn.com/abstract=4824172
Figure 1: The time-varying boundary area for identifying the start of a new upward or downward trend
in price. The shaded area represents noise or choppy price action.

Boundary, depicted by the red line, signals abnormal selling pressure, suggesting a pos-
sible downward trend.

A distinctive feature of this model is that the noise boundaries are dynamic and vary
throughout the trading day. This variation means, for example, that the market move-
ment required to indicate a demand/supply imbalance is different in the first 30 minutes
of trading compared to the first 6 hours. Typically, the average intraday movement from
the Open tends to increase over time, peaking at 16:00. As demonstrated in the simplified
example in Figure 1, the movement required from the Open to 10:30 to indicate abnormal
selling pressure was -0.30%, which was half of that required between the Open and 15:30.

The introduction of time-varying boundaries represents a novel feature that distinguishes

this model from well-documented systems like the volatility breakout systems by Cabrel
[18] and Kaufman [19], or the intraday momentum models described in academic papers
by authors such as Rosa [7] and Baltussen et al. [8]. This adaptability enhances the strat-
egy’s responsiveness to varying market conditions throughout the trading day, providing
a more precise and actionable framework for identifying significant market movements.

7
Electronic copy available at: https://ssrn.com/abstract=4824172
 (a) Long (b) Short

Figure 2: The time-varying boundary areas on SPY for 31 January 2022 (a) and 29 April 2022 (b) and
the resulting trades based on our base-model. Please note that all positions are closed at the market
Close (16:00).

As discussed earlier, if the market breaches the boundaries of the Noise Area, our strat-
egy initiates positions in accordance with the prevailing intraday move—going long if the
price is above the Noise Area and short if it is below. To mitigate the risk of overtrad-
ing caused by short-term market fluctuations, trading is restricted to bi-hourly intervals,
specifically at HH:00 and HH:30. This timing mechanism ensures that decisions are
based on more stable price movements rather than transient spikes.

Positions are unwound either at market Close or if there is a crossover to the opposite
boundary of the Noise Area. In the event of such a crossover, the existing position is
closed, and a new one is initiated in the opposite direction to align with the latest evidence
of demand/supply imbalance. Stop losses can also be triggered only on bi-hourly intervals.

Figures 2 illustrates the Noise Areas and the resulting trades on 2 days with clear trends
that occurred in 2022. As shown by the dashed grid on the charts, trades are executed
at 10:30, even though the price had moved outside the Noise Area a few minutes earlier.
This delay in response reduces the risk that trading decisions are triggered by fleeting
market movements. All positions are closed at market Close, adhering to the strategy’s
protocol to limit exposure to overnight market changes.

8
Electronic copy available at: https://ssrn.com/abstract=4824172
The strategy allocates exposure equal to 100% of the equity available at the beginning
of each trading day. We calculate the number of shares that can be traded using the
equation:

 
AU Mt−1
Sharest = ,
Opent,9:30

where AU Mt−1 represents the equity (or Asset Under Management) available at the end
of day t − 1.

To assess the historical performance of our strategy, we simulate a portfolio starting with
an initial capital of $100,000. The backtest incorporates transaction costs, including a
commission fee of $0.0035 per share, which represents the entry-level commission charged
by Interactive Brokers. Additionally, we account for a slippage of $0.001 per share, which
is consistent with a real-market case study recently conducted in our proprietary trading
accounts1 .

These operational parameters ensure that our backtesting reflects realistic trading con-
ditions and provides a credible estimation of the strategy’s potential performance.

Figure 3 displays the equity curve of the strategy in black, juxtaposed with a long passive
exposure in SPY, shown in red. As set out in Table 1, from 2007 to 2024, the active intra-
day momentum strategy realized a total return of 178%, with an average annual return
of 6.2% and a volatility of 10.9%. This resulted in a Sharpe Ratio of 0.61. Despite the
overall slight underperformance compared to the passive SPY exposure, the most con-
cerning aspects are the worst daily return of -10.33% and the negative skewness of the
intraday momentum portfolio (more details are provided in Table A1, at the end of the
paper). Typically, trend-following strategies are favored for their significant asymmetry
between positive and negative daily returns.

By design, the active intraday strategy is vulnerable to sudden market reversals, which
1
More details are provided in Section 4.6

9
Electronic copy available at: https://ssrn.com/abstract=4824172
Figure 3: Comparison of equity curves of intraday momentum strategy vs. SPY buy-hold strategy.
Commission set at $0.0035 as per Interactive Brokers’ entry-level rate.

can be triggered by news releases or other fundamental catalysts that can provoke dra-
matic shifts in market character. A trailing stop set at the opposite side of the Noise
Area exacerbates this vulnerability, as the portfolio can suffer dramatically on such days.

January 20, 2022 serves as a poignant example of the weaknesses inherent in our base
model. As shown in Figure 4, the market strongly rallies in the first 30 minutes, triggering
a long exposure. For some 4 hours, the price remains above the Noise Area, indicating
continued abnormal buying pressure. However, around 14:00, the market begins to turn
south, dropping almost 2% in the last trading hours; SPY closes the day below the Lower
Boundary of the Noise Area, resulting in a 2.19% loss for the active strategy. This sce-
nario highlights the need for adjustments in our base model to mitigate losses during

Table 1: Summary statistics of intraday momentum strategy with stop loss at opposite band and SPY
Buy&Hold. Commission set at $0.0035 as per Interactive Brokers’ entry-level rate. We highlight in bold
coefficients that are statistically significant at 5% level or below.

Total Sharpe Hit

Strategy Stop Size IRR Vol MDD Alpha Beta
Return Ratio Ratio

Momentum Opp.Band 100% 178% 6.2% 10.9% 0.61 54% 21% 7.1% -0.05
SPY (Buy&Hold) 100% 227% 7.2% 20.2% 0.45 54% 56%

10
Electronic copy available at: https://ssrn.com/abstract=4824172
Figure 4: SPY price action on 20 January 2022 illustrates our model strategy’s entry and exits. The
base strategy uses the opposite band as a trailing stop. Following a sharp reversal in SPY in the last 2
hours of trading, the base model realizes a loss of more than 2%.

abrupt market reversals.

A straightforward method to mitigate the risk of sudden market reversals is by utilizing

the Upper Boundary of the Noise Area as a trailing stop for our long positions, and vice
versa for short positions. As we demonstrate in Figure 5(a), this approach leads to the
position being closed at 14:00, just as the market begins to weaken. This adjustment
improves the return from -2.19% to -0.31%. While a tighter stop loss can significantly
reduce downside risks, it is important to note that it may also substantially increase the
number of trades due to false signals, and consequently, the associated transaction costs.

This observation underscores the need to balance the protective benefits of a tighter stop
loss against the potential for increased operational costs, ensuring that the strategy re-
mains economically viable while providing adequate risk management.

Moreover, as documented in a previous paper [5], Volume Weighted Average Price (VWAP)
can serve as a powerful intraday indicator to identify shifts in the demand-supply imbal-
ance. Building upon insights from our previous publication, we have chosen to integrate

11
Electronic copy available at: https://ssrn.com/abstract=4824172
 (a) Stop = UpperBand (b) Stop = max(VWAP,UpperBand)

Figure 5: Improvement in our strategy’s exit by incorporating tighter trailing stop losses. In Figure
(a), the trailing stop is based on the upper band of the Noise Area, while in Figure (b), it is based on
the maximum between the VWAP and the upper band.

VWAP as a trailing stop for intraday trend exposures2 . Once a position is established,
as soon as the price crosses either the curent band or the VWAP, the position is closed.
Mathematically, the trailing stops for long and short postions are defined as:

Long TrailingStopt,HH:M M = max(U Bt,HH:M M , V W APt,HH:M M )

Short TrailingStopt,HH:M M = min(LBt,HH:M M , V W APt,HH:M M )

Figure 5(b) illustrates the revised approach in the case study of January 20, 2022. Includ-
ing VWAP allows us to close the long exposure at 13:00, ultimately achieving a break-even
trade. This adjustment represents a significant improvement, showcasing VWAP’s effi-
cacy as a dynamic stop loss mechanism that aligns closely with market momentum, thus
enhancing the strategy’s risk management capabilities.

To demonstrate the economic advantages of these trailing stop refinements, we have con-
ducted a comprehensive historical backtest, the results of which are displayed in Figure
6 and Table 2. The improvements are substantial: the average annual return increases
to 9.7%, while annual volatility decreases to 7.7%. The Sharpe Ratio doubles, improving
from 0.61 to 1.24, while the Hit Ratio decreased to 43% due tighter stops. As exhibited
2
VWAP is computed only using market hours data. For more details see our previous publication [5].

12
Electronic copy available at: https://ssrn.com/abstract=4824172
 Figure 6: Comparison of equity curves of intraday momentum strategy with a) stop loss at opposite
band, b) current band with VWAP, and c) SPY buy-hold strategy. Commission set at $0.0035 as per
Interactive Brokers’ entry-level rate.

in Table A1, the skewness shifts to positive (from -1.24 to 1.29) and the worst daily re-
turn sees a 5% improvement (from -10.3% to -4.8%). However, the tighter stop loss does
result in a significant increase in the total number of trades, rising from 5,494 to 7,668
over the 17-year period (see Table 4). These results highlight the enhanced risk-return
profile achieved by the strategic modifications, albeit with an increase in trading activity.

As a final refinement to our model, we have implemented a sizing methodology that

dynamically adjusts the traded exposure based on daily market volatility. Instead of
maintaining constant full notional exposure, this method targets a daily market volatility
of 2% (σtarget = 2%). Practically, if the recent daily volatility of SPY is 4%, we would

Table 2: Summary statistics of intraday momentum strategy with a) stop loss at opposite band, b)
current band with VWAP, c) SPY Buy&Hold. Commission set at $0.0035 as per Interactive Brokers’
entry-level rate. We highlight in bold coefficients that are statistically significant at 5% level or below.

Total Sharpe Hit

Strategy Stop Size IRR Vol MDD Alpha Beta
Return Ratio Ratio

Momentum Opp.Band 100% 178% 6.2% 10.9% 0.61 54% 21% 7.1% -0.05
Momentum Curr.Band + VWAP 100% 380% 9.7% 7.7% 1.24 43% 12% 9.9% -0.03
SPY (Buy&Hold) 100% 227% 7.2% 20.2% 0.45 54% 56%

13
Electronic copy available at: https://ssrn.com/abstract=4824172
 Table 3: Summary statistics of intraday momentum strategy with a) stop loss at opposite band, b)
current band with VWAP, c) as in (b) with the additional dynamically adjusted share size based on
daily market volatility, and d) SPY Buy&Hold Commission set at $0.0035 as per Interactive Brokers’
entry-level rate. We highlight in bold coefficients that are statistically significant at 5% level or below.

Total Sharpe Hit

Strategy Stop Size IRR Vol MDD Alpha Beta
Return Ratio Ratio

Momentum Opp.Band 100% 178% 6.2% 10.9% 0.61 54% 21% 7.1% -0.05
Momentum Curr.Band + VWAP 100% 380% 9.7% 7.7% 1.24 43% 12% 9.9% -0.03
Momentum Curr.Band + VWAP Dyn. 1,985% 19.6% 14.3% 1.33 43% 25% 19.6% -0.07
SPY (Buy&Hold) 100% 227% 7.2% 20.2% 0.45 54% 56%

trade with half of the capital; conversely, if it is 1%, we would utilize a leverage of 2.
Mathematically, the number of shares traded on day t is computed as

 
AU Mt−1 × min(4, σtarget /σSP Y,t )
Sharest = ,
Opent,9:30
where
v
u
u1 X 14 14
1 X
σSP Y,t =t (rett−i − µSP Y,t )2 , µSP Y,t = (rett−i )
13 i=1 14 i=1

This adjustment results in a more stable risk profile for the active portfolio, making it
less dependent on overall market volatility regimes. In line with mainstream brokerage
policies, the leverage for the active strategy is capped at 4x.

Figure 7 displays the equity line for the modified portfolio, while Table 3 details the main
performance statistics. The active intraday momentum portfolio achieves a total return
of 1,985%, an annualized return of 19.6%, with a volatility of 14.3%. The Sharpe Ratio
improves to 1.33, indicative of a more stable risk profile. The maximum drawdown is
contained at 25%, and the daily Hit Ratio stands at approximately 43%. The portfolio’s
best trade realizes a return of 9.1%, significantly outperforming the worst trade return
of -2.9% (see Table 4).

To assess the outperformance of the active strategy compared to a passive exposure in

SPY, we employ the following classic regression analysis:

14
Electronic copy available at: https://ssrn.com/abstract=4824172
Figure 7: Comparison of equity curves of intraday momentum strategy with a) stop loss at opposite
band, b) current band with VWAP, c) as in (b) with the additional dynamically adjusted share size based
on daily market volatility, and d) SPY buy-hold strategy. Commission set at $0.0035 as per Interactive
Brokers’ entry-level rate.

RetMom,t = α + β × RetSPY,t + ϵt .

The regression analysis reveals that the annualized alpha of the active strategy is 19.6%
and it is highly statistically significant. This indicates that the strategy outperforms the
baseline market passive exposure by a substantial margin on a consistent basis. Fur-
thermore, the beta of the strategy is slightly below 0, which suggests there is a negative

Table 4: Summary statistics at trade-level of intraday momentum strategy with a) stop loss at opposite
band, b) current band with VWAP, c) as in (b) with the additional dynamically adjusted share size based
on daily market volatility, and d) SPY Buy&Hold strategy Commission set at $0.0035 as per Interactive
Brokers’ entry-level rate.

Avg. Trade Hit Avg. PnL Max Loss Max Gain

Strategy Stop Size Trades
x Day Ratio x Share Trade Trade

-8.6% 6.0%
Intra Mom. Opp.Band 100% 5,494 1.3 53% 0.10
(13-Nov-2008) (13-Oct-2008)
-4.0% 4.6%
Intra Mom. Curr.Band + VWAP 100% 7,668 1.8 37% 0.09
(18-Mar-2020) (15-Oct-2008)
-2.9% 9.1%
Intra Mom. Curr.Band + VWAP Dyn. 7,668 1.8 37% 0.09
(31-Jul-2013) (10-Oct-2018)

15
Electronic copy available at: https://ssrn.com/abstract=4824172
correlation between the active strategy’s returns and the overall market movements. This
lack of correlation underscores the strategy’s effectiveness in diversifying risk and achiev-
ing returns independent of market trends.

4 Further Investigations
In this section, we outline the more in-depth statistical analysis we have conducted in or-
der to assess how the profitability of the intraday momentum strategy changes in different
market scenarios. Unless otherwise specified, all tests are conducted using the intraday
momentum strategy that employs a trailing stop based on the current band and VWAP,
and adjusts investment size according to recent market volatility.

4.1 Market Volatility Regimes

Even though our active intraday momentum strategy adjusts daily exposure based on
recent market volatility, it is pertinent to investigate how the expected profitability of
the strategy varies across different market volatility regimes. For this analysis, we utilize
the VIX Index value at the opening of each trading session to categorize the days and
subsequently assess the strategy’s Sharpe Ratio specifically for days when the VIX is
above certain thresholds.

Figure 8 displays the results using a simple bar chart. Consistent with findings by Rosa
[7], the risk-adjusted returns tend to improve with increasing levels of market volatil-
ity. For instance, when the VIX is above 6, the Sharpe Ratio is approximately 1.50;
remarkably, when the VIX exceeds 40, the Sharpe Ratio escalates to an impressive 3.50.
However, for even higher levels of VIX, the Sharpe Ratio begins to decrease, likely due to
the fewer observations available and higher dispersion among results on extremely volatile
days.

This pattern suggests that a more sophisticated implementation of the strategy could
benefit from dynamically adjusting exposure based on current market volatility. Such an
adaptive approach would potentially enhance the strategy’s performance by optimizing

16
Electronic copy available at: https://ssrn.com/abstract=4824172
 Figure 8: Resulting Sharpe Ratio for different level of VIX at market Open.

risk exposure in response to fluctuating market conditions.

4.2 Daily Patterns

Well-known active traders often rely on technical analysis to identify market setups that
could signal an improvement in the expected profitability of a trading strategy. For
instance, the presence of a daily hammer might indicate an upcoming enhancement in
the effectiveness of an intraday momentum strategy. Considering the virtually infinite
number of patterns that could be analyzed, we chose to focus on 8 common patterns
widely recognized and utilized by experienced traders:

• NR4 (Narrow Range 4 Days): Identifies days where the range (high-low) is the
smallest of the last 4 days.

• NR7 (Narrow Range 7 Days): Identifies days where the range is the smallest
of the last 7 days.

• ID (Inside Days): Identifies days when the day’s high is below the previous day’s
high, and the low is above the previous day’s low.

• OD (Outside Days): Identifies days when the high is above the previous day’s
high and the low is below the previous day’s low.

• Triangle: Identifies days when the high is below the highs of the previous 2 days,
and the low is above the lows of the previous 2 days.

• Trend Days: Identifies days when the Open is below the 15th percentile of the
daily range, and the Close is above the 85th percentile, suggesting a strong bullish

17
Electronic copy available at: https://ssrn.com/abstract=4824172
Table 5: Summary comparison of daily patterns and statistics achieved on our intraday momentum
strategy applied on each pattern. We highlight in bold Avg.PnL that are statistically significant at 10%
level or below.

Avg.PnL Hit Sharpe

Daily Pattern Observations t-stat
(in bps) Ratio Ratio
Unconditional 2620 12 5.34 43% 1.7
NR4 660 22 5.14 45% 3.2
NR7 367 16 3.07 43% 2.5
ID 260 5 0.85 41% 0.8
OD 260 7 1.05 43% 1.0
Triangle 664 14 3.19 43% 2.0
Trend 215 -2 -0.24 40% -0.3
Big Tail 85 19 1.17 47% 2.0
Strong/Weak Closure 422 14 2.04 43% 1.6

trend (the reverse applies for bearish trends). The current range must also exceed
the average range of the last 14 days.

• Big Tail: Identified on days with hammers or reverse hammers; where the Open
and Close are within the top 75th percentile of the daily range (or the reverse for
reverse hammers).

• Strong/Weak Closure: Identifies days where the closing price is above the 90th
percentile (strong closure) or below the 10th percentile (weak closure) of the daily
range.

Table 5 presents our findings, which include the number of days each trading pattern
occurs, the average return per day expressed in basis points, the computed t-statistic,
the Hit Ratio and the Sharpe Ratio. The first row provides aggregate results that are
unconditional of specific daily patterns.

The interpretation of the t-statistics is crucial; a value between -1.65 and 1.65 suggests
that the expected return is not statistically different from zero at a 10% confidence level.
This indicates insufficient evidence to reject the null hypothesis of zero average return
at this level of significance. In practice, if looking for reliable edges, the higher the t-
statistics, the better (very negative t-statistics suggest that traders should flip the trading
signal).

18
Electronic copy available at: https://ssrn.com/abstract=4824172
Among the patterns we analyze, the NR4 pattern emerges as the most promising, exhibit-
ing an average return of 22 basis points with a t-statistic of 5.14 and a Hit Ratio of 45%.
This pattern demonstrates statistically significant profitability and warrants considera-
tion for increased investment focus. The Triangle pattern also shows notable profitability
with an average daily return of 14 basis points and a t-statistic of 3.19.

Conversely, the strategy of riding intraday trends after a trend day does not show a
statistically significant edge, suggesting it may not be effective as a stand-alone trading
strategy. The Big Tail pattern, despite an attractive return profile, suffers from a lack of
statistical reliability due to a limited number of observations (85 occurrences).

As reported in the first row, employing a trading strategy that allows for daily trading
irrespective of specific setups, proves to be more reliable and statistically significant.
However, in order to exploit higher expected returns, active day traders might consider
adjusting bet sizes after observing NR4, NR7, and Triangle patterns.

4.3 Day-of-the-Week Effect

The Day-of-the-Week Effect is a focal point for statistical arbitrageurs who leverage cal-
endar seasonality to identify market inefficiencies. While this approach runs the risk of
data-mining, the observed patterns could have substantial economic underpinnings tied
to specific days. For instance, options expirations are traditionally more pronounced
on Wednesdays and Fridays, which are historically the days most options expire in our
dataset. With the emergence of zero-days-to-expiration (0-dte) options, the effects of
options expirations are expected to be more evenly distributed across all weekdays.

In Table 5, we present the number of traded days, the average daily returns, t-statistics,
Hit Ratios and Sharpe Ratio for each trading day. The data highlights that Wednesday,
Thursday, and Friday stand out as the most profitable days with statistically significant
results. Intriguingly, Monday shows no statistical significance, contrasting with our ear-
lier findings of higher Monday profitability in the 5-minute Opening Range Breakout
(ORB), as discussed in our previous publications [4, 20]. This discrepancy could be at-

19
Electronic copy available at: https://ssrn.com/abstract=4824172
Table 6: Summary statistics grouped by day-of-the-week for the momentum intraday strategy. We
highlight in bold Avg.PnL that are statistically significant at 10% level or below.

Avg.PnL Hit Sharpe

Day Observations t-stat
(in bps) Ratio Ratio
Monday 472 9 1.84 43% 1.4
Tuesday 543 9 2.00 42% 1.4
Wednesday 530 18 3.42 45% 2.4
Thursday 588 12 2.39 43% 1.6
Friday 487 13 2.39 44% 1.7

tributed to the weekend effect, where significant market movements on Monday often
occur within the first 30 minutes of trading as traders rush to rebalance portfolios post-
weekend. Contrary to the 5-minute ORB, which aims to capture moves starting from
9:35, the intraday momentum strategy we outline here takes its first position at 10:00,
potentially missing a significant portion of the early move. Additionally, the noted high
profitability on Wednesday may be linked to significant intraday trends often triggered
on Federal Open Market Committee (FOMC) days.

4.4 Volatility Multiplier

A central element of the strategy we discuss in this paper is the definition of the Noise
Area, essential for identifying potential tradable trends. As we initially define it, a mar-
ket movement is recognized as a possible trend when it exceeds the average movement
recorded during the same intraday time interval over the previous 14 days. We now intro-
duce a parameter we term as the Volatility Multiplier (VM), which adjusts the width of
the Noise Area, making the model either more conservative (for values above 1) or more
aggressive (for values below 1). A VM below 1 reduces the required movement in SPY to
signal a potential trend, whereas a multiplier above 1 demands a larger movement before
initiating any exposure.

UpperBoundt,HH:M M = Opent,9:30 × (1 + VM × σt,9:30−HH:M M )

LowerBoundt,HH:M M = Opent,9:30 × (1 − VM × σt,9:30−HH:M M )

20
Electronic copy available at: https://ssrn.com/abstract=4824172
Figure 9: The effect of Volatility Multiplier on the Sharpe Ratio and Total Return of the intraday
momentum strategy

We have analyzed the profitability of the intraday momentum strategy across different
settings of the VM and the results are presented in Figure 9. The bottom bar chart in
Figure 9 illustrates that the total return negatively correlates with the VM. This sug-
gests that a too conservative approach (i.e. a very high VM) may result in less exploitable
trading opportunities. Conversely, the top bar chart in Figure 9 provides a more com-
prehensive view by considering not only average returns but also the volatility of the
strategy’s returns. This chart indicates that the best risk-adjusted returns are achieved
with a VM of approximately 1.5, highlighting a subjective trade-off between the total
return and risk-adjusted returns.

Given the complexity of the decision between higher returns and better risk-adjusted
performance, we leave it to each trader to determine the most appropriate implemen-
tation. To maintain simplicity and statistical intuition, the strategies presented in this
paper employ a VM of 1, even though this may not be the most efficient setting based
on retrospective analysis.

21
Electronic copy available at: https://ssrn.com/abstract=4824172
4.5 Gamma Imbalance

As Baltussen et al [8] and Barbon and Buraschi [21] note, the delta hedging activities
carried out by financial institutions to manage the risk in their options portfolios can
result in self-fulfilling trends in intraday price movements. This phenomenon becomes
particularly pronounced when the overall gamma of dealers is negative, necessitating ad-
justments to the delta in the same direction as the market trend. Typically, it is assumed
that dealers sell put options to investors seeking to hedge against market downturns,
while simultaneously purchasing call options from large institutional investors engaged in
call-overwriting strategies. Consequently, if the market hovers around the strike prices of
predominantly purchased calls, dealers are compelled to counteract intraday movements
to rebalance their portfolio’s delta to zero—this involves selling on intraday rallies and
buying on dips. Conversely, if the market approaches the strike prices of predominantly
sold puts, dealers must enhance intraday movements to align their portfolio delta to zero,
thus amplifying the intraday trend.

Although it is beyond the scope of this paper to provide a precise estimate of gamma im-
balance, we hypothesize that utilizing a short-term Relative Strength Index (RSI) might
allow us to gauge whether the market is near the bulk of call strikes or put strikes. To
explore this, we propose running a regression analysis between the daily returns of our
active momentum portfolio (dependent variable) and the 5-day RSI of SPY (explana-
tory variable) as of the previous day’s close. We anticipate that a higher RSI indicates
greater dealer gamma, suggesting a higher likelihood of intraday mean reversion and,
consequently, lower returns for our intraday momentum strategy. Conversely, a lower
RSI suggests reduced gamma, implying more pronounced intraday trends and potentially
higher profitability for the momentum portfolio. By running the following regression3 :

RetMom,t = α + β × RSISPY,t−1 + ϵt .

we obtain a value of beta of -3.25 with a p-value of 0.001. The results indicate that the
slope coefficient of the regression is negative and significantly different from zero. This
3
To enhance the readability of the regression results, we express RetMom,t in basis points. For instance,
a return of 100 basis points is equivalent to a return of 1%.

22
Electronic copy available at: https://ssrn.com/abstract=4824172
finding suggests a substantial negative linear relationship between the RSI level and the
profitability of the intraday momentum strategy. More specifically, a higher RSI, indi-
cating that the market has recently moved higher, correlates with lower expected returns
from the strategy. Conversely, a lower RSI, suggesting a recent market decline, correlates
with increased expected returns.

This pattern aligns with the hypothesis that when the market has been rising, there is
a higher likelihood of dealers having a long gamma position, which could cause intraday
trends to be short-lived due to their delta hedging practices. Conversely, a sharp decline
in market levels in recent days may lead dealers to shift to a negative gamma position,
thus initiating a trend-chasing mechanism in their delta hedging strategies.

4.6 The Impact of Commissions (and Slippage)

When dealing with low-frequency trading strategies that involve trading liquid instru-
ments less than once per month, the role of transaction costs and slippage is negligible.
However, when evaluating the profitability of intraday trading strategies, it is very im-
portant to conduct an in-depth analysis of how commissions and slippage could impact a
trading edge. Estimating commissions is straightforward, as most brokers provide trans-
parent details about the cost per share charged for each transaction. What is usually
more difficult to estimate is the slippage paid for each transaction.

As we anticipate in the previous section, in the base strategy, we use the entry-level
commission charged by Interactive Brokers: $0.0035 per share. Typically, the commis-
sion charged by the broker depends on the monthly trading volume by the client. For
example, at Interactive Brokers, the commission decreases to $0.0020 per share when the
traded volume exceeds 300,000 shares per month. In our backtest, we estimate that on
approximately 20% of the days, we benefit from lower commission fees due to a 1-month
trailing volume exceeding 300,000 shares.

As for slippage, instead of using a random number, we have conducted a real-market

experiment on a few trading days in April 2024. Using an Interactive Brokers account

23
Electronic copy available at: https://ssrn.com/abstract=4824172
 Figure 10: Impact of commissions on total return of intraday momentum strategy.

funded with $100,000, we instruct an algorithm to send market orders on SPY for $100,000
just a few milliseconds before the start of minute t. The algorithm randomly takes long
or short positions. The goal of the study is to evaluate the magnitude of the slippage
relative to the Open price of minute t. Mathematically, we evaluate the distribution of
the following variable:

Slippage = Side × (ExecutedP rice − Open9:30 ) (1)

After executing over 1,000 trades, involving more than $50 million in notional value, the
average slippage recorded is approximately $0.001, with a median value even lower at
$0.0005. We recognize that these figures might rise with an increase in portfolio size,
yet we remain confident that employing advanced trading techniques can facilitate large
orders without significantly impacting the market. For simplicity and consistency with
our real-trading analysis, we have chosen to apply a standard slippage rate of $0.001 in
our backtest.

To further assess the sensitivity of our results to varying commission levels and associated
slippage, we rerun the backtest using a spectrum of commission values. Figure 10 illus-
trates how the total return of the intraday momentum strategy fluctuates across different
commission levels, offering a more nuanced understanding of how costs impact profitabil-
ity. Given the available diversity in commission structures and order methodologies, we
recommend that traders evaluate the profitability of the strategy based on their specific
circumstances.

24
Electronic copy available at: https://ssrn.com/abstract=4824172
5 Conclusion
In conclusion, this paper delves into the realm of intraday momentum trading strategies,
building upon the foundational work laid by Jegadeesh and Titman’s momentum effect.
Through empirical analysis and rigorous backtesting, we demonstrate the profitability
and viability of implementing intraday momentum strategies within the context of the
SPY ETF, tracking the S&P 500 index.

Our findings reveal that a simple intraday momentum strategy, initiated upon indica-
tions of abnormal demand/supply imbalance in intraday price action and incorporating
dynamic trailing stops, yields substantial returns over the period from 2007 to early 2024.
The intraday momentum portfolio achieves a remarkable total return of 1,985%, with an
annualized return of 19.6%, and a Sharpe Ratio of 1.33, underscoring its robustness and
potential as a trading approach.

Moreover, we have conducted extensive statistical tests to assess the impact of different
market volatility regimes and dealer gamma imbalances on the strategy’s profitability.
Additionally, our analysis of day-of-the-week effects and performance against well-known
technical patterns provide further insights into the behavior of intraday momentum strate-
gies under various market conditions.

This study contributes to the ongoing discourse on market anomalies and the practical
applications of momentum trading, particularly in intraday contexts. By extending Je-
gadeesh and Titman’s momentum framework to intraday trading, we provide valuable
insights for traders operating in fast-paced intraday markets. Our research underscores
the persistent nature of momentum effects within shorter time frames and highlights op-
portunities for profitable trading strategies in today’s dynamic trading landscape.

In summary, the results offer compelling evidence supporting the efficacy of intraday
momentum strategies, further enriching our understanding of market dynamics and of-
fering actionable insights for market participants seeking to capitalize on short-term price
movements.

25
Electronic copy available at: https://ssrn.com/abstract=4824172
Author Biography

Andrew Aziz is a Canadian trader, investor, and official

Forbes Council member. He has ranked as one of the top
100 best-selling authors in ”Business and Finance” for 7 con-
secutive years from 2016 to 2023. Aziz’s book on finance has
been published in 13 different languages. Originally from Iran,
Andrew moved to Canada in 2008 to pursue a PhD in chem-
ical engineering, initiating a distinguished career in academia
and industry. As a research scientist, Andrew made significant
contributions to the field, authoring 13 papers and securing 3
US patents. Following a successful stint in research in chemical
engineering and clean technology, he transitioned to the world
of trading. Currently Andrew is a trader and proprietary fund
manager at Peak Capital Trading in Vancouver, BC Canada.

Carlo Zarattini, originally from Italy, currently resides in

Lugano, Switzerland. After completing his mathematics de-
gree in Padova, he pursued a dual master’s in quantitative
finance at Imperial College London and USI Lugano. He for-
merly served as a quantitative analyst at BlackRock, where
he developed volatility and trend-following trading strategies.
Carlo later established Concretum Research, assisting institu-
tional clients with both high and medium-frequency quantita-
tive strategies in stocks, futures, and options. Additionally, he
founded R-Candles.com, the first backtester for discretionary
technical traders.

Andrea Barbon, born in Venice, currently resides in Zurich,

Switzerland. He holds a Master degree in pure mathemat-
ics from the University of Amsterdam and a PhD in finance
from the University of Lugano. He is Assistant Professor of
Financial Technology at the FSI Center of the University of
St.Gallen, Switzerland and at the Swiss Finance Institute. His
research interests include asset pricing, monetary policy, fin-
tech, blockchain, and machine learning. He is also head of AI
at Concretum Research, and lead developer for the R-Candles
web application.

26
Electronic copy available at: https://ssrn.com/abstract=4824172
References
[1] N. Jegadeesh and S. Titman. Returns to buying winners and selling losers: Impli-
cations for stock market efficiency. J Finance, 48(1):65–91, 1993.

[2] A. Aziz. Advanced Techniques in Day Trading: A Practical Guide to High Probability
Strategies and Methods. AMS Publishing Group, 2018.

[3] A. Aziz. How to Day Trade for a Living: A Beginner’s Guide to Trading Tools and
Tactics, Money Management, Discipline and Trading Psychology. AMS Publishing
Group, 2015.

[4] C. Zarattini and A. Aziz. Can day trading really be profitable? evidence of sustain-
able long-term profits from opening range breakout (orb) day trading strategy vs.
benchmark in the us stock market. SSRN Electronic Journal, 2023.

[5] C. Zarattini and A. Aziz. Volume weighted average price (vwap) the holy grail for
day trading systems. SSRN Electronic Journal, 2023.

[6] Li Gao, Yan Han, S. Zhengzi Li, and Guofu Zhou. Market intraday momentum. J
financ econ, 129(2):394–414, 2018.

[7] C. Rosa. Understanding intraday momentum strategies. Journal of Futures Markets,

42(12):2218–2234, 2022.

[8] G. Baltussen, Z. Da, S. Lammers, and M. Martens. Hedging demand and market
intraday momentum. Journal of Financial Economics, 142(1):377–403, 2021.

[9] Y. Zhang, F. Ma, and B. Zhu. Intraday momentum and stock return predictability:
Evidence from china. Economic Modelling, 76:319–329, 2019.

[10] M. Jin, F. Kearney, Y. Li, and Y.C. Yang. Intraday time-series momentum: Evidence
from china. Journal of Futures Markets, 40(4):632–650, 2020.

[11] Z. Li, A. Sakkas, and A. Urquhart. Intraday time series momentum: Global evidence
and links to market characteristics. Journal of Financial Markets, 57:100619, 2022.

[12] A. Aaziznia and A. Aziz. A Beginner’s Guide to Investing and Trading in the Modern
Stock Market. 2020.

[13] Tobias J Moskowitz, Yao Hua Ooi, and Lasse Heje Pedersen. Time series momentum.
Journal of Financial Economics, 104(2):228–250, 2012.

[14] Andrea Frazzini. The disposition effect and underreaction to news. The Journal of
Finance, 61(4):2017–2046, 2006.

[15] Stefano DellaVigna and Joshua M Pollet. Investor inattention and friday earnings
announcements. The Journal of Finance, 64(2):709–749, 2009.

[16] Andrew B Abel, Janice C Eberly, and Stavros Panageas. Optimal inattention to
the stock market with information costs and transactions costs. Econometrica,
81(4):1455–1481, 2013.

27
Electronic copy available at: https://ssrn.com/abstract=4824172
[17] Darrell Duffie. Presidential address: Asset price dynamics with slow-moving capital.
The Journal of Finance, 65(4):1237–1267, 2010.

[18] T. Crabel. Day trading with short term price patterns and opening range breakout.
1990.

[19] P. J. Kaufman. Trading Systems and Methods. Wiley, 5 edition, 2020.

[20] C. Zarattini, A. Barbon, and A. Aziz. A profitable day trading strategy for the u.s.
equity market. SSRN Electronic Journal, February 2024.

[21] A. Barbon and A. Buraschi. Gamma fragility. SSRN Electronic Journal, November
2020.

28
Electronic copy available at: https://ssrn.com/abstract=4824172
 A Further Tables

Table A1: Summary statistics of intraday momentum strategy with a) stop loss at opposite band, b) current band with VWAP, c) as in (b) with the
additional dynamically adjusted share size based on daily market volatility, and d) SPY buy-hold strategy Commission set at $0.0035 as per Interactive
Brokers’ entry-level rate. We highlight in bold coefficients that are statistically significant at 5% level or below.

Total Sharpe Hit

Strategy Stop Size IRR Vol MDD Alpha Beta Skewness Worst Day Best Day
Return Ratio Ratio

Momentum Opp.Band 100% 178% 6.2% 10.9% 0.61 54% 21% 7.1% -0.05 -1.24 -10.3% 6.1%
Momentum Curr.Band + VWAP 100% 380% 9.7% 7.7% 1.24 43% 12% 9.9% -0.03 1.29 -4.8% 5.1%
Momentum Curr.Band + VWAP Dyn. 1,985% 19.6% 14.3% 1.33 43% 25% 19.6% -0.07 1.22 -4.5% 9.1%
SPY (Buy&Hold) 100% 227% 7.2% 20.2% 0.45 54% 56% 0.08 -10.9 14.5

29
Electronic copy available at: https://ssrn.com/abstract=4824172