Abstract

The core hypothesis of this strategy is that combining technical indicators (ATR, RSI, SMA) and momentum signal in VN30F1M data and VN30 data (since VN30 and VN30F1M is highly correlated) with dynamic position sizing can produce consistent, profitable trading signals. By scaling into positions based on volatility and signal strength, and applying robust risk controls (stop-losses, take-profits, partial exits, and trailing stops), the strategy aims to capture market opportunities while limiting downside risk.

Introduction

In today's dynamic financial markets, systematic trading strategies require robust frameworks that adapt to changing conditions while effectively managing risk. This approach provides a comprehensive and systematic framework that bridges theoretical trading models with real-world applications, offering traders a disciplined method to navigate volatile markets while aiming for consistent performance.

Strategy's Key Idea

• The investment capital allocated for trading derivatives in the Vietnamese market is 1.5 billion VND (45 contracts at a time)
• VN30F1M in these year is kind of a sideway market so reversion strategy with a reversal momentum indicator is my choice to aim for a strategy with big take profit

Long position conditions:

{
    def check_long_position_conditions(self, row, acceleration_threshold, quantity_multiply, sma_gap, short_acceleration_threshold, rsi_threshold):
        conditions = [
            row['Acceleration'] > acceleration_threshold,
            row['VN30 Acceleration'] > 0,
            row['volume'] > row['Average Quantity'] * quantity_multiply,
            row['Price/SMA'] < 1 - sma_gap,
            row['Short Acceleration'] > short_acceleration_threshold,
            row['RSI'] < 50 - rsi_threshold
        ] 
}

$$ \text{Acceleration} = \mathtt{matchedprice}_t - \mathtt{matchedprice}_{t - momentumlookback} $$

• 'Acceleration': Defined as the difference between the matched price at time 't' and the matched price at the previous momentum lookback interval. Positive acceleration indicates a genuine price reversal.

• 'Short Acceleration': Represents acceleration measured over the most recent intervals, verifying immediate past price momentum.

• 'VN30 Acceleration': VN30F1M is highly correlated with VN30 data so I also use VN30 data as a confirmation of the reversal signal

• 'Volume': The current trading volume per minute must exceed the historical average volume 'Average Quantity' multiplied by a specified optimization factor 'quantity_multiply'. Increased volume combined with price reversal signals supports entering a long position.

• 'Price/SMA Ratio': The current price relative to its Simple Moving Average (SMA) must be below a threshold (1 - sma_gap), signaling undervaluation.

• 'Relative Strength Index (RSI)': An RSI value lower than the threshold (50 - rsi_threshold) indicates an oversold condition, thus reinforcing the long position signal. An RSI below 20 strongly suggests an oversold market.

Since the strategy is quite complex to matched all the signal, just 4 in 6 condition is needed to entry a position

Short Position Conditions:

Short positions are taken under conditions inverse to those for long positions, with a higher take-profit threshold. Observationally, downward market trends typically show stronger momentum compared to upward trends, justifying a higher profit expectation from short positions.

Position Sizing and Contract Management:

With a capital of 1.5 billion VND, the strategy can simultaneously trade up to 45 contracts. Thus, determining the strength of the trading signal is crucial for optimal contract allocation.

Signal Strength Scoring:

• Signal strength is calculated as: Acceleration/Acceleration Threshold

• Higher values of 'a' suggest a stronger signal, prompting a larger number of contracts to be traded.

• The base number of contracts 'base_contract' is calculated by flooring the value of 'a' and multiplying by 10.

Volatility Adjustment:

• During periods of high volatility (measured by the standard deviation of matched prices over the last 50 intervals, with each interval representing one minute), the number of contracts traded is reduced to 50% of the 'base_contract'.

• In low-volatility periods, contracts traded are increased to 150% of the 'base_contract'.

• Normal volatility periods warrant trading the exact 'base_contract' amount.

Adjusted Entry Price:

When multiple orders are placed at different price levels, the entry price and total contracts are recalculated into a single average entry price to simplify strategy tracking and profit-and-loss (PnL) calculations:

$$ \text{Adjusted Entry Price} = \frac{Contracts_{t-1}.P_{t-1} + Contracts_t.P_t}{Contracts_{t-1} + Contracts_t} $$

Feature

• Research the 1-minute candle scalping to beat the fee and spread (0.47 in total)
• Validate Test Case: Momentum Signal, Reversion Signal and Backtesting
• Optimize parameters
• Evaluate backtesting and optimization
• Paper trade

Data

• Historical market data of this project from 01/2023 to 04/2024 including OHLC (Open, High, Low, Close) prices, trading volumes (from Algotrade database), and the VN30 index values (from SSI fast connect data API). More detail of SSI API can be found here
• Data is sourced from reliable financial providers and stored as CSV format in data folder.

Installation

Create environment:

python -m venv myenv

Activate environment:

• Windows (CMD):

myenv\Scripts\activate.bat

• Windows (PowerShell):

myenv\Scripts\activate.ps1

• macOS / Linux:

source myenv/bin/activate

After activation, your terminal prompt should change to indicate you're in a virtual environment, for example:

(myenv) user@computer:~$

Clone the Repository:

git clone https://github.com/algotrade-research/scalping-strategy.git

• Requirement: pip
• Install the dependencies by:

pip install -r requirements.txt

• Install ssi fast connect

cd data/fc-data
pip install ssi-fc-data

Implementation

Tick based data is really noise and hard to develop the larger take profit strategy so I convert to 1 minute candle data for less noise and enhance more technical analysis.

Environment Setup and Replication Steps

In-sample Backtesting

The data from 01/01/2023 to 31/12/2023

Parameters

• Strategy parameters (e.g., SMA window length, momentum lookback, acceleration thresholds, etc.)
• Risk management settings (e.g., stop-loss, take-profit thresholds)
• Initial asset value (e.g., 10,000)

In-sample Backtesting Result

The in-sample backtesting results are summarized in a performance table and visualized with a cumulative PNL chart.

To see the results, run this command with the main.py file and choose option in in the terminal

python main.py

Optimization

• Number of Trials: 10000
• Sampler Seed: 42
• Study Direction: Maximize cumulative PNL

It can take 10-15 second per trial, which can make it 5-6 hours for an entire process. Becareful

Run the optimization process by

python optimize.py

Optimization Process / Methods / Library

• Library: Optuna
• Method: Tree-structured Parzen Estimator (TPE) sampler
• Process: Multiple trials are conducted to evaluate various parameter combinations, selecting the best based on the objective function (maximizing cumulative PNL).

Parameters to Optimize

• SMA window length
• SMA gap
• Momentum lookback
• Acceleration thresholds (long and short)
• Take-profit threshold
• Cut-loss threshold
• Quantity window
• Quantity multiplier
• Short extra profit
• RSI window
• RSI threshold

Optimization Result

The results is stored in optimization/sma.db The best parameters is stored in optimization/best_params.json file

{
    "sma_window_length": 25,
    "sma_gap": 0.02017081163304353,
    "momentum_lookback": 7,
    "acceleration_threshold": 0.27265410618834895,
    "short_acceleration_threshold": 0.08577163230799632,
    "take_profit_threshold": 4.9983573167104245,
    "cut_loss_threshold": 1.955171845695386,
    "quantity_window": 20,
    "quantity_multiply": 2,
    "short_extra_profit": 1.9352589805341556,
    "rsi_window": 38,
    "rsi_threshold": 39
}

Cumulative PnL

Metric	In-sample
Sharpe Ratio	0.109
Maximum Drawdown	10.65%
Win Rate	40.96%
Total Long Trades	5959
Total Short Trades	6355

Out-of-sample Backtesting

Cumulative PnL

Metric	Out-sample
Sharpe Ratio	0.132
Maximum Drawdown	8.09%
Win Rate	44.17%
Total Long Trades	1840
Total Short Trades	1455

To see the results, run this command with the main.py file and choose option out in the terminal

python main.py

Conclusion

In conclusion, this scalping strategy tackles the challenge of high transaction costs (0.47%), showing strong results in in-sample testing but limited success out-of-sample. While not consistently profitable across all conditions, it still achieved a positive PnL and, more importantly, provided valuable insights. Through this process, I gained a deeper understanding of strategy design, risk management, and the complexities of trading in high-fee environments—paving the way for future improvements.

	Sharpe Ratio	MMMD	Win Rate	Total Long Trades	Total Short Trades
In-Sample	0.109	10.65%	40.96%	5959	6355
Out-Sample	0.132	8.09%	44.17%	1840	1455

References

This project did not have References