This repository presents a complete and profitable algorithmic trading strategy for the USDCHF currency pair on a 5-minute timeframe. The system is implemented in Python using the Backtrader framework and features an advanced, dynamic risk management system based on the Average True Range (ATR).
The strategy was backtested over 5 years of historical data, achieving a 1.35 Profit Factor with a 35.4% return on initial capital.
The core of this project's backtesting methodology is the "Dual Cerebro" approach, where LONG and SHORT strategies are run as two independent portfolios. This prevents conflicting signals and allows each directional bias to be optimized and evaluated on its own merits. The final "Combined" performance is the aggregation of these two independent results.
- Purple Line (Combined Portfolio): Represents the total equity curve, showing a final P&L of +$35,399.
- Green Line (LONG Only Portfolio): Shows the performance if only buy signals were traded, resulting in a P&L of +$18,343.
- Red Line (SHORT Only Portfolio): Shows the performance if only sell signals were traded, resulting in a P&L of +$17,056.
This chart demonstrates that both the LONG and SHORT logic contribute positively and consistently to the overall profitability of the system.
- 📈 Advanced Pullback Entry System: The strategy does not enter on the initial signal. It waits for a confirmation pullback and enters on a breakout, significantly improving the entry quality and risk/reward ratio.
- 🛡️ Dynamic ATR Risk Management: Stop Loss and Take Profit levels are not fixed. They are calculated dynamically for every trade based on the market's current volatility (ATR), adapting the risk profile to live conditions.
- 📐 EMA Angle Momentum Filter: A unique filter that measures the slope of the confirmation EMA. It ensures trades are only taken during periods of strong, decisive market momentum, filtering out flat or choppy markets.
- 🔄 Dual Cerebro Backtesting: By testing LONG and SHORT logic independently, the strategy avoids signal interference and provides a clearer picture of each component's performance.
- 🕰️ Time-of-Day Filtering: Trading is restricted to the most liquid market hours (07:00 - 17:00 UTC) to focus on the highest probability setups.
- 💰 Risk-Based Position Sizing: Automatically calculates trade size to risk a fixed 1% of the account on every position, maintaining consistent risk exposure.
The strategy is a multi-layered, trend-following system. An entry is only triggered when a sequence of conditions are met.
- Signal Detection: A fast EMA (1) crosses over a basket of slower EMAs (14, 18, 24).
- Pullback Wait: The system waits for a small, controlled move against the trend (e.g., 1-2 red candles for a LONG signal).
- Breakout Entry: A market order is triggered only if the price breaks the high/low of the pullback, confirming the trend's continuation.
This chart shows the strategy in action:
- Candlestick Chart: The 5-minute price action for USDCHF.
- EMA Lines: The multiple EMAs used for signal generation.
- Green Triangle (Buy Signal): Marks the exact bar where a LONG trade was executed.
- Dashed Lines (SL/TP Observer): These are the most important feature. The dashed red line is the initial Stop Loss, and the dashed green line is the Take Profit target. Both are calculated using ATR multiples at the moment of entry.
For a detailed view of the complete repository structure and file descriptions, see: Repository Structure Guide
The entire logic is contained within src/strategy/sunrise_osiris.py. Here is a brief guide to its structure:
- Global Configuration (Lines 1-150): This top section contains all user-editable parameters, such as date ranges, starting cash, ATR thresholds, and filter settings. This makes tuning the strategy easy without touching the core logic.
SunriseOsirisClassparams(Lines 153-270): The officialBacktraderparameters dictionary. These are the values that can be optimized during a parameter sweep.__init__(self)(Lines 605-700): The strategy constructor where all indicators (EMAs, ATR) and state variables (e.g., for the pullback machine) are initialized.next(self)(Lines 875-1045): The heart of the strategy. This method is called for every bar of data. It contains the main logic for checking entry/exit conditions and managing the trade lifecycle.- Pullback State Machine (
_handle_pullback_entry): A set of helper functions that manage the 3-phase entry logic (Signal -> Wait -> Breakout). - Risk Management (
_calculate_forex_position_size): A function dedicated to calculating the correct position size based on the stop-loss distance and the 1% account risk rule. - Order Notifications (
notify_order,notify_trade): These methods handle the feedback from the broker (simulated or real), placing the OCA (One-Cancels-All) Stop Loss and Take Profit orders after an entry is confirmed.
Follow these instructions to set up and run the backtest on your local machine.
- Python 3.8 or newer.
- Git version control.
Open your terminal and clone the repository:
git clone https://github.com/[Your-Username]/backtrader-atr-stop-loss-usdchf.git
cd backtrader-atr-stop-loss-usdchfIt is highly recommended to use a virtual environment to manage dependencies.
# Create the virtual environment
python -m venv venv
# Activate it
# On Windows:
venv\Scripts\activate
# On macOS/Linux:
source venv/bin/activateInstall the necessary Python libraries using the provided requirements.txt file.
pip install -r requirements.txtPlace your 5-minute USDCHF data file into the /data directory. The project expects the file to be named USDCHF_5m_5Yea.csv.
Execute the strategy script from the root directory of the project.
python src/strategy/sunrise_osiris.pyThe script will run the full Dual Cerebro backtest, print the performance summary to the console, generate detailed trade reports in the /temp_reports folder, and display the final performance chart.
Contributions are welcome! Whether you want to report a bug, suggest an enhancement, or add a new feature, please feel free to do so.
- Report a Bug: Open an Issue and describe the problem in detail.
- Suggest an Enhancement: Open an Issue to discuss your idea.
- Submit a Pull Request: Fork the repository, make your changes, and submit a Pull Request with a clear description of your work.
This project is licensed under the MIT License. See the LICENSE file for details.
This project is for educational and research purposes ONLY. It is not financial advice. Algorithmic trading involves substantial risk of loss and is not suitable for every investor. Past performance is not indicative of future results.