TruTrend

• Pham Vo Quynh Nhu - 21125052
• Ho Viet Bao Long - 21125071
• Nguyen Phuc Bao Uyen - 21125146

⚠️ Note: Make sure you're inside the Group 03 folder (the root directory) before continuing.

Abstract

This study introduces a rule-based trading strategy for VN30F1M, using technical indicators to define precise entry and exit points for long and short positions. Long entries rely on short-term EMAs crossing above long-term EMAs, high RSI, and strong volume, while short entries occur under opposite conditions. Exits are governed by strict rules, including stop-loss limits, EMA cross reversals, RSI shifts, and an ATR-based trailing stop. The strategy aims to enhance trade consistency and risk management in a Vietnam derivatives market.

Introduction

The derivatives market in Vietnam officially launched on August 10, 2017, with the VN30 Index Futures as its first traded product. The market was introduced to offer investors additional tools for risk hedging, speculation, and portfolio management. During the period from 2018 to 2022, the VN30 Index Futures contract demonstrated strong performance, with an average annual increase of 38.65% [1]. This notable growth highlights the significant opportunities available in the market, especially for individual investors.

However, individual investors often tend to be influenced by various factors such as emotions, human error, and slow decision-making and ordering, which can lead to missed opportunities or failure to execute orders at target prices since the investors are unable to monitor the market continuously [2]. These obstacles can be addressed through the use of algorithmic trading, which not only helps overcome the issues mentioned above but also saves time for investors by automating decision-making and order execution [2].

In this project, we propose an algorithm to automate the trading process in Vietnam derivative market. The algorithm are made of from the idea of technical indicator rules, with a combinations of Exponential Moving Averages (EMA), Relative Strength Index (RSI), and ATR (Average True Range), and Volume of the market. The OLHC is collected as candle of 5-minutes of VN30F1M. Generally, we open a position if a clear trend is identified (LONG or SHORT) and it is confirmed by a sufficiently high volume, indicating strong market participation and increasing the likelihood that the trend is genuine and sustainable rather than a temporary fluctuation. Otherwise, close position acted if the loss is larger than maximum loss boundering, ending of the trend, or trailing signal.

We tested our hypothesis in two phases: an in-sample test using VN30F1M data from January 3, 2023 to December 21, 2023, and an out-of-sample test using data from January 2, 2024 to December 19, 2024, with an optimization process conducted in between. The initial capital was set at 40,000,000 VND, and the strategy was restricted to holding only one contract at a time. In the in-sample test, the strategy achieved a Sharpe ratio of 1.0846, a maximum drawdown (MDD) of -16.49%, and a net profit of 14,332,000 VND, indicating that the hypothesis holds promise for further optimization. For the optimization process, we first defined a list of possible parameter values and then performed random sampling from the full set of parameter combinations, testing each configuration on the 2023 dataset. We selected the top five parameter sets with the highest Sharpe ratios for further evaluation in the out-of-sample test. The final model was chosen as the one among the top five that yielded the highest Sharpe ratio on the 2024 out-of-sample data. As a result, the selected model achieved a Sharpe ratio of 2.5949 on the in-sample data and 1.6497 on the out-of-sample data.

The remainder of this template is dedicated to describing our implementation process, organized into several key sections. The Background section introduces and explains the main technical indicators used in the algorithm. The Trading Hypothesis section details the proposed algorithm, including how these indicators are combined to determine entry and exit signals. The Data section outlines the data collection process and how the data is utilized in model development and testing. The Implementation section describes the setup and environment used to run the experiments. The In-sample Testing section explains how the initial evaluation was conducted using a default set of parameters. The Optimization section illustrates the process of testing a wide range of parameter combinations and presents the results. The Out-of-sample Testing section evaluates the top-performing parameter sets selected from the optimization phase on unseen data. The Paper Trading section provides a tutorial on deploying the algorithm in a real-time trading simulation. Finally, the Conclusion summarizes our work and key findings.

Background

EMA (Exponential Moving Average)

The Exponential Moving Average (EMA) is a weighted moving average that gives more weight to recent prices [3]. It is calculated as:

$EMA_t = EMA_{t-1} + \alpha \times (P_t - EMA_{t-1})$

Where:

• $EMA_t$ is the EMA at time $t$,
• $EMA_{t-1}$ is the previous EMA value,
• $P_t$ is the current price at time $t$,
• $\alpha$ is the smoothing factor, defined as $\alpha$ = $\frac{2}{N+1}$,
• $N$ is the number of periods.

RSI (Relative Strength Index)

RSI measures momentum by comparing recent gains to recent losses, with values between 0 and 100 [4]. It is calculated as:

$RSI = 100 - \left( \frac{100}{1 + RS} \right)$

Where:

• $RS$ is the average gain of up periods divided by the average loss of down periods.

ATR (Average True Range)

The Average True Range (ATR) measures market volatility [5]. The formula is:

$ATR = \frac{1}{N} \sum_{i=1}^{N} \text{True Range}_i$

Where:

• $N$ is the number of periods,
• $\text{True Range}_i$ is the maximum of:
  • Current high - current low,
  • Absolute value of current high - previous close,
  • Absolute value of current low - previous close.

Trading (Algorithm) Hypothesis

The trading strategy focuses on trend-following and momentum signals to determine entry points. A position is only opened when sufficient historical data is available and all specified conditions are met. For a LONG position: a short-term EMA above the long-term EMA, price above the short EMA, RSI above a set upper threshold, and strong volume. A SHORT position is triggered under the opposite conditions, indicating a bearish trend with high trading activity. Only one position is held at a time, ensuring clear directional exposure.

Preconditions

• The starting balance is 40,000,000 VND

• The portfolio holds only one position at a time

• Only one action is allowed per trade: either Long or Short

Hypothesis

Generally, the decision-making process of the algorithm is driven by the following parameters:

Parameter	Description
EMA_SHORT	Period for calculating the short-term Exponential Moving Average (EMA).
EMA_LONG	Period for calculating the long-term Exponential Moving Average (EMA).
RSI_PERIOD	Period for calculating the Relative Strength Index (RSI).
RSI_LOWER_THRESHOLD	RSI value below which a long position may be considered (oversold zone).
RSI_UPPER_THRESHOLD	RSI value above which a short position may be considered (overbought zone).
ATR_PERIOD	Period for calculating the Average True Range (ATR).
ATR_MULTIPLIER	Factor used to scale the ATR for setting trailing stop-loss levels.
MIN_PROFIT	Minimum required profit to trigger a profit-taking condition.
MAX_LOSS	Maximum allowed loss before triggering a stop-loss.
VOLUME_THRESHOLD	Multiplier applied to average volume to detect significant volume spikes.
VOLUME_WINDOW	Number of bars used to compute rolling average volume.
RSI_EXIT_THRESHOLD	RSI value used to help decide when to exit an active trade.

There are two main components to the hypothesis: Criteria for opening a position and Criteria for closing a position.

Opening a Position

• No position is opened if either:
  • There is insufficient historical data (e.g., not enough bars to calculate RSI or EMA), or
  • None of the entry conditions are satisfied.
• Open a LONG position when all of the following conditions are met:
  • EMA_SHORT is greater than EMA_LONG (indicating an uptrend),
  • CURRENT_PRICE is above EMA_SHORT,
  • RSI exceeds the RSI_UPPER_THRESHOLD,
  • CURRENT_VOLUME is greater than VOLUME_THRESHOLD $\times$ AVERAGE_VOLUME (indicating strong market interest).
• Open a SHORT position when all of the following conditions are met:
  • EMA_SHORT is less than EMA_LONG (indicating a downtrend).
  • CURRENT_PRICE is below EMA_SHORT.
  • RSI falls below the RSI_LOWER_THRESHOLD.
  • CURRENT_VOLUME is greater than VOLUME_THRESHOLD $\times$ AVERAGE_VOLUME.

Close position:

• No position is closed if none of the exit conditions are triggered.
• Close a LONG position when any of the following conditions apply
  • The current loss exceeds the MAX_LOSS threshold.
  • CURRENT_PRICE drops below EMA_LONG and RSI falls below the RSI_EXIT_THRESHOLD.
  • Trailing Stop Condition: If profit exceeds MIN_PROFIT, and CURRENT_PRICE < max(ENTRY_PRICE + MIN_PROFIT,PREVIOUS_CLOSE - ATR × ATR_MULTIPLIER).
• Close SHORT position when any of the following conditions apply:
  • The current loss exceeds the MAX_LOSS threshold.
  • CURRENT_PRICE rises above EMA_LONG and RSI is greater than the RSI_EXIT_THRESHOLD.
  • Trailing Stop Condition: If profit exceeds MIN_PROFIT, and CURRENT_PRICE > min(ENTRY_PRICE - MIN_PROFIT, PREVIOUS_CLOSE + ATR × ATR_MULTIPLIER)

Data

• Data source: Algotrade internship database
• Data type: VN30F1M
• Data period: 2023- present
• Both input and output data will be stored in a specified data path. The default data path is data folder in the src directory of the project. You can change the data path by changing the DATA_PATH variable in the .env file.

Data collection

• The data is collected from Algotrade internship database

Data Processing

• The queried data is extracted and saved as CSV files, with the results stored in the <DATA_PATH> directory.

• The data is then processed into tick data based on a preferred interval (more infomation in the Implementation Data Collection and Preprocessing). In this project, we focus on a 5-minute interval.

The resulting file will be saved under the name: <start_date>_to_<end_date>_by_<tick_interval>.csv For example:

src/data/2023-01-01_to_2023-12-31_by_5T.csv

The CSV file will include the following columns:

• Date: The date of the interval
• Time: The time of the interval in format (YY-MM-DD)
• tickersymbol: That month ticketsymbol
• Open: Opening price
• Close: Closing price
• High: Highest price during the interval
• Low: Lowest price during the interval
• Volume: Trading volume

Implementation

Environment Setup

1. Set up Python Virtual Environment

python3 -m venv venv
python -m venv venv
source venv/bin/activate # for Linux/MacOS
.\venv\Scripts\activate.bat # for Windows command line
.\venv\Scripts\Activate.ps1 # for Windows PowerShell

2. Install the required packages

pip install -r requirements.txt

3. (OPTIONAL) Create .env file in the directory src of the project and fill in the required information. The .env file is used to store environment variables that are used in the project. The following is an example of a .env file:

DB_NAME=<database name>
DB_USER=<database user name>
DB_PASSWORD=<database password>
DB_HOST=<host name or IP address>
DB_PORT=<database port>
DATA_PATH=<path to the data folder of your choice, if not specified, the default is `data`>

The DATA_PATH variable is used to specify the path to the data folder where the input data is stored. The other variables are used to connect to the database.

Data Collection and Preprocessing

Option 1. Using Available Data Files All available data files are located in the src/data/ directory.

Option 2. Run code to collect data This command runs the src/preprocess.py script to extract and preprocess market data between specified dates with a defined resampling interval.

• --start_date: The beginning date of the data range (format: YYYY-MM-DD).

• --end_date: The ending date of the data range (format: YYYY-MM-DD).

• --interval: The time interval for resampling the data using pandas (default: 5T, which means 5-minute intervals).

python src/preprocess.py --start_date 2023-01-01 --end_date 2023-12-31 --interval 5T

This example processes data from January 1, 2023, to December 31, 2023, and resamples it into 5-minute intervals. The process may take approximately 2-3 minutes to complete.

The resulting file will be saved in the <DATA_PATH> directory (default src/data/) under the name: <start_date>_to_<end_date>_by_<tick_interval>.csv The CSV file will include the following columns:

,Date,Time,tickersymbol,Open,Close,High,Low,Volume
1,2023-01-03,09:00:00,VN30F2301,1000.7,1002.0,1002.6,1000.7,186

In-sample Backtesting

In the src/parameters.csv file, key hyperparameters for our hypothesis are set to default values for in-sample backtesting. These include settings for indicators such as EMA, RSI, ATR, and volume thresholds, as well as risk management parameters like stop-loss and take-profit levels.

EMA_Short	EMA_Long	RSI_Period	RSI_lower	RSI_upper	ATR_period	Max_Loss	Min_Profit	ATR_Mult	Volume_Threshold	Volume_window	RSI_exit_threshold
10	30	14	45	55	14	3.0	0.5	1.5	1.2	10	50

For backtesting, we use data from the period 2023-01-01 to 2023-12-31 with a 5T interval as the in-sample period. By default, the data file is saved in src/data if you do not define DATAPATH in src/.env.

• --dataset: specify the dataset file to be used
• --result_dir: specify the directory for saving charts, trade logs, and portfolio values
• --parameters: specify the directory for parameters.

python src/backtesting.py --dataset 2023-01-01_to_2023-12-31_by_5T.csv --result_dir src/result_in_sample --parameters src/parameters_in_sample.csv

In-sample Backtesting Result

After running the command below, we obtained the Sharpe Ratio and Maximum Drawdown:

SHARPE RATIO: 1.0846
MDD: -16.49%

Here is the in-sample backtesting result:

Optimization

To identify a well-performing set of parameters for our hypothesis, we applied a random sampling approach. Initially, we generated the full set of possible parameter combinations using Python's itertools.product, based on predefined ranges. From this comprehensive list, we employed NumPy to randomly sample a subset for evaluation. Each sampled configuration was tested through backtesting, and its performance was assessed using the Sharpe Ratio and Maximum Drawdown (MDD) metrics. The results were then visualized using Python's Matplotlib library to facilitate analysis.

In total, twelve parameters were considered in the optimization process: ema_short, ema_long, rsi_range, rsi_lower_threshold, rsi_upper_threshold, atr_period, max_loss, min_profit, atr_multiplier, volume_threshold, volume_window, and rsi_exit_threshold. The range of possible values for each parameter is specified in the optimization.json file, with the following defaults:

{
  "ema_short_range": [5, 8, 10, 12, 15],
  "ema_long_range": [20, 25, 30, 35, 40],
  "rsi_range": [10, 14, 18],
  "rsi_lower_threshold_range": [25, 30, 35, 40, 45],
  "rsi_upper_threshold_range": [50, 55, 60, 65, 70],
  "atr_period_range": [10, 14, 21],
  "max_loss_range": [1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5, 5.0],
  "min_profit_range": [0.5, 1.0, 1.5, 2.0],
  "atr_mult_range": [1.0, 1.5, 2.0],
  "vol_thresh_range": [0.8, 1.0, 1.2, 1.4],
  "volume_window_range": [5, 10, 15, 20],
  "rsi_exit_threshold_range": [45, 50, 55]
}

To run optimization, run the following command:

• -d: Specify the dataset file to be used
• -p: Path to the JSON file containing the possible values for each parameter to be optimized.
• -n: Number of models randomly selected from the list of parameter combinations.
• -o: Location to save the optimization results

python src/optimization.py -d src/data/2023-01-01_to_2023-12-31_by_5T.csv -p src/optimization.json -m 2500 -o src/OPTIMIZATION_RANDOM

To draw the graphs to illustrate the results of optimization process, run the following script:

• -v: Path to summary.csv file created by the optimization process.
• ⚠️ Note: Click the exit button (red button in the top-left corner) of each figure to view the next one. Do not terminate the program. The figures will appear one after another.

python src/optimization.py -v src/OPTIMIZATION_RANDOM/summary.csv

The set of configurations was evaluated using the following scoring formula:

$$ Score = Sharpe - \alpha \cdot |MDD| $$

• $\alpha = 0.5$

This scoring function considers the Sharpe ratio as the primary evaluation metric, while the impacts of Maximum Drawdown is adjusted using the weights $\alpha$. Run the following command to compute the scores based on the results:

python src/optimization.py -s src/OPTIMIZATION_RANDOM/summary.csv -f src/OPTIMIZATION_RESULTS/scoreModel.csv

Optimization Result

The optimization process took approximately 4.5 hours in total, during which 2,500 parameter combinations were evaluated. The following graphs illustrate the results of the optimization process using VN30F1M data from January 3, 2023 to December 21, 2023.

The figure below illustrates the relationship between the Sharpe Ratio and MDD for the 2,500 parameter combinations considered. The highest Sharpe Ratio observed was 2.782, while the lowest was -0.4861. In contrast, the MDD ranged from a maximum of -0.0556 to a minimum of -0.3143.

The following figure illustrates the distribution of net profits across all tested configurations. The majority of the profits fell within the range of 10,000,000 VND to 20,000,000 VND, indicating a consistent level of profitability among many parameter sets.

The figure below shows the relationship between the Sharpe ratio and the net profit of the tested configurations. The results suggest a linear correlation, indicating that higher Sharpe ratios tend to be associated with greater net profits.

The figure below shows the distribution of the scores. Most models achieved scores in the range of 1.0 to 1.5. The highest score reached around 2.74, while the lowest was approximately -0.64.

Out-of-sample Backtesting

After optimization, we identified the top 5 parameter sets with the highest scores. The results are saved in src/parameters_out_sample.csv, and the top 5 are listed below.

SampleIndex	EMA_Short	EMA_Long	RSI_Period	RSI_lower	RSI_upper	ATR_period	Max_Loss	Min_Profit	ATR_Mult	Volume_Threshold	Volume_window	RSI_exit_threshold	Sharpe	mdd	net_profit	score
271	10	25	18	25	70	14	2.5	1	1	1.2	20	45	2.782016973	-0.07720320466	29702000	2.74
1634	8	40	14	35	70	21	5	1	1	0.8	10	45	2.717278496	-0.07548636353	41320000	2.68
974	5	20	18	35	70	14	4.5	1	1	0.8	15	50	2.674389187	-0.06388517122	36240000	2.64
2500	15	25	18	25	70	14	1.5	1	1	0.8	20	50	2.594883786	-0.07613530181	29810000	2.55
2358	10	30	18	25	70	21	2	0.5	1	0.8	15	55	2.573981968	-0.07130522918	30525000	2.54

For out-of-sample testing, we use data from the period 2024-01-01 to 2025-01-01 with a 5T interval as the in-sample period. By default, the data file is saved in src/data if you do not define DATAPATH in src/.env.

python src/backtesting.py --dataset 2024-01-01_to_2025-01-01_by_5T.csv --result_dir src/result_out_sample --parameters src/parameters_out_sample.csv

After running the command above, we obtained the Sharpe Ratio and Maximum Drawdown:

Sample Index	Sharpe Ratio	Maximum Drawdown (MDD)
0	0.4091	-19.32%
271	0.5741	-11.77%
1634	-0.0494	-25.98%
974	0.3945	-17.39%
2500	1.6497	-8.00%
2358	1.4416	-9.66%

The highest parameter for the outsample is

SampleIndex	EMA_Short	EMA_Long	RSI_Period	RSI_lower	RSI_upper	ATR_period	Max_Loss	Min_Profit	ATR_Mult	Volume_Threshold	Volume_window	RSI_exit_threshold
2500	15	25	18	25	70	14	1.5	1	1	0.8	20	50

Paper Trading

All the code in paper trading is in the paper_trading folder

cd src/paper_trading
python paper_trading.py

We have configured the paper trading environment using incoming messages from Kafka in combination with historical data retrieved from the database. However, the current backtesting implementation does not support dynamic parameter adjustment and is operating with a fixed set of parameters.

EMA_FAST	EMA_SLOW	RSI_PERIOD	ATR_PERIOD	VOL_WINDOW	VOL_THRES	RSI_UPPER	RSI_LOWER	MAX_LOSS	MIN_PROFIT	ATR_MULT	RSI_EXIT
10	30	18	21	15	0.8	70	25	2.0	0.5	1.0	55

The message received from Kafka is preprocessed to align with the format of the above data.

Conclusion

This trading strategy combines trend and momentum indicators—namely EMA, RSI, and volume—to identify high-probability market entry points. The project provides a foundational overview of the algorithmic trading workflow. Overall, the results were promising, demonstrating relatively strong Sharpe ratios and acceptable drawdowns in both in-sample and out-of-sample evaluations. Notably, the final model achieved solid performance on both datasets, with profit rates of 74.525% and 43.735%, respectively, based on an initial capital of 40,000,000 VND.

However, the optimization was conducted on only a subset of parameter combinations, which may have resulted in overlooking potentially better-performing configurations. Additionally, the paper trading phase lasted only three days, providing insufficient data to thoroughly evaluate the model’s performance. The current hypotheses are strictly based on technical analysis, which limits the model’s ability to account for market events and news. Therefore, future work should consider expanding the parameter search space to explore more configurations, extending the paper trading period for more comprehensive monitoring, and incorporating modern techniques such as Machine Learning or Deep Learning to better capture market-driven events and anomalies.

Reference

[1] Vietnam News Agency, “Six years of derivative stock market: VN30 index futures contract up 27.46% annually,” VietnamPlus, Aug. 14, 2023. [Online]. Available: https://en.vietnamplus.vn/six-years-of-derivative-stock-market-vn30-index-futures-contract-up-2746-annually-post266231.vnp. [Accessed: Apr. 23, 2025].​

[2] Algotrade, “Thực tiễn giao dịch thuật toán tại thị trường chứng khoán Việt Nam,” Algotrade Knowledge Hub, [Online]. Available: https://hub.algotrade.vn/knowledge-hub/thuc-tien-giao-dich-thuat-toan-tai-thi-truong-chung-khoan-viet-nam/. [Accessed: Apr. 23, 2025].​

[3] J. Chen, “Exponential Moving Average (EMA),” Investopedia, May 3, 2023. [Online]. Available: https://www.investopedia.com/terms/e/ema.asp. [Accessed: Apr. 23, 2025].

[4] J. Chen, “Relative Strength Index (RSI),” Investopedia, May 3, 2023. [Online]. Available: https://www.investopedia.com/terms/r/rsi.asp. [Accessed: Apr. 23, 2025].

[5] A. HAYES, “Average True Range (ATR),” Investopedia, May 4, 2023. [Online]. Available: https://www.investopedia.com/terms/a/atr.asp. [Accessed: Apr. 23, 2025].