Trading Strategy Optimizer

This project provides a Python script for backtesting and heuristic optimization of trading strategies based on technical indicators, being applied to spot market time series.

As main advantages, the project provides:

• simulated annealing, genetic and grid search algorithms to perform a global search for the best strategies.
• creation of spreadsheets and figures with performance results from backtesting and optimization procedure.
• open-source code, allowing flexibility in adjusting the search space and implementing technical indicators.

🗂 Project Structure

The project is organized around a modular pipeline, where each class has a responsibility:

• Loader handles configuration files for tickers and market settings.
• Indicator applies technical indicators (MA crossover, Bollinger Bands, MACD) and generates trading signals.
• Backtester evaluates strategies on historical data and computes performance metrics.
• Optimizer searches the indicator parameter space using heuristic optimization.
• Strategies scores and ranks candidate strategies based on configurable objective functions.

The project has the following structure:

trading_strategy_optimizer/ 
│  
├── trading_strategy_optimizer.py 
├── trading_strategy_optimizer_app.py 
|  
├── core/   
│   ├── __init__.py  
│   ├── loader.py  
│   ├── indicator.py  
│   ├── backtester.py  
│   ├── optimizer.py  
│   ├── strategies.py  
│   ├── exporter.py  
│   └── visualizer.py  
|  
├── core_app/   
│   ├── __init__.py  
│   ├── gui.py  
│   ├── redirector.py  
│  
├── config/  
│   ├── config.json  
│   └── tickers.json  
│  
├── data/  
│   └── results/ 
|       ├── best_results.xlsx 
│       ├── strategies.csv  
│       ├── backtests.png   
│       └── logs.txt   
│  
├── images/
├── requirements.txt  
├── README.md  
└── LICENSE  

📈 Available Strategies

The project currently supports the following indicators:

• Moving Average crossover using:
  • SMA (Simple Moving Average);
  • EMA (Exponential Moving Average);
  • WMA (Weighted Moving Average).
• Bollinger bands;
• Moving Average Convergence/Divergence (MACD).

⚙️ How to Use

1. Install dependencies:

    pip install pandas
    pip install numpy
    pip install matplotlib
    pip install yfinance
    pip install requests
    pip install openpyxl

2. Configure parameters and tickers

   • In config/config.json add the configuration parameters.
   • In config/tickers.json add the stock symbols to analyze, one per line.

3. Run the script

   • To run the optimization with backtests, execute:

     python trading_strategy_optimizer.py

🧩 Output Examples

• Backtest charts for MACD strategy

  After running trading_strategy_optimizer.py optimization it is generated strategy charts, spreadsheets for each ticker, and a summary with best results. The generated figure (for each combination evaluated) follows the example:

  

  Notice that the asset ends the evaluated periodo at 1.75 times its initial price, so a Buy & hold strategy would yield 175% return. On the other hand, strictly following the MACD 17/22/11 strategy would produce approximately 350% return over the same period, excluding any transactions fees.

• Optimization log and charts for MACD strategy

  After running trading_strategy_optimizer.py it is generated optimization logs and charts, for each ticker, are. The generated files follow the example:

  k = 1: x = {'ind_t': 'MACD', 'ind_p': [5, 26, 9]} | f(x) = 1.0535 | T = 0.95 | alpha = 0.90
  k = 2: x = {'ind_t': 'MACD', 'ind_p': [8, 20, 10]} | f(x) = 1.3766 | T = 0.90 | alpha = 0.81
  k = 3: x = {'ind_t': 'MACD', 'ind_p': [6, 35, 10]} | f(x) = 1.8316 | T = 0.86 | alpha = 0.73
  k = 4: x = {'ind_t': 'MACD', 'ind_p': [5, 33, 5]} | f(x) = 1.0936 | T = 0.81 | alpha = 0.66
  k = 5: x = {'ind_t': 'MACD', 'ind_p': [6, 21, 8]} | f(x) = 0.9848 | T = 0.77 | alpha = 0.59
  k = 6: x = {'ind_t': 'MACD', 'ind_p': [7, 24, 8]} | f(x) = 1.2401 | T = 0.74 | alpha = 0.53
  k = 7: x = {'ind_t': 'MACD', 'ind_p': [8, 41, 9]} | f(x) = 1.4409 | T = 0.70 | alpha = 0.48
  k = 8: x = {'ind_t': 'MACD', 'ind_p': [11, 45, 10]} | f(x) = 2.4746 | T = 0.66 | alpha = 0.43
  k = 9: x = {'ind_t': 'MACD', 'ind_p': [13, 33, 8]} | f(x) = 2.0475 | T = 0.63 | alpha = 0.39
  k = 10: x = {'ind_t': 'MACD', 'ind_p': [11, 32, 10]} | f(x) = 2.2553 | T = 0.60 | alpha = 0.35
  k = 11: x = {'ind_t': 'MACD', 'ind_p': [11, 37, 11]} | f(x) = 2.7945 | T = 0.57 | alpha = 0.31
  k = 12: x = {'ind_t': 'MACD', 'ind_p': [9, 38, 13]} | f(x) = 2.7945 | T = 0.54 | alpha = 0.28
  k = 13: x = {'ind_t': 'MACD', 'ind_p': [8, 42, 14]} | f(x) = 2.8524 | T = 0.51 | alpha = 0.25
  k = 14: x = {'ind_t': 'MACD', 'ind_p': [9, 36, 15]} | f(x) = 2.7918 | T = 0.49 | alpha = 0.23
  k = 15: x = {'ind_t': 'MACD', 'ind_p': [8, 32, 14]} | f(x) = 2.4582 | T = 0.46 | alpha = 0.21
  k = 16: x = {'ind_t': 'MACD', 'ind_p': [7, 33, 14]} | f(x) = 1.8741 | T = 0.44 | alpha = 0.19
  k = 17: x = {'ind_t': 'MACD', 'ind_p': [7, 32, 12]} | f(x) = 1.4098 | T = 0.42 | alpha = 0.17
  k = 18: x = {'ind_t': 'MACD', 'ind_p': [8, 35, 12]} | f(x) = 1.7622 | T = 0.40 | alpha = 0.15
  k = 19: x = {'ind_t': 'MACD', 'ind_p': [8, 35, 13]} | f(x) = 2.4355 | T = 0.38 | alpha = 0.14
  k = 20: x = {'ind_t': 'MACD', 'ind_p': [9, 36, 12]} | f(x) = 2.7753 | T = 0.36 | alpha = 0.12
  k = 21: x = {'ind_t': 'MACD', 'ind_p': [10, 34, 14]} | f(x) = 2.7413 | T = 0.34 | alpha = 0.11
  k = 22: x = {'ind_t': 'MACD', 'ind_p': [9, 34, 14]} | f(x) = 2.7945 | T = 0.32 | alpha = 0.10
  k = 23: x = {'ind_t': 'MACD', 'ind_p': [10, 35, 15]} | f(x) = 2.6018 | T = 0.31 | alpha = 0.09
  k = 24: x = {'ind_t': 'MACD', 'ind_p': [11, 36, 14]} | f(x) = 2.0969 | T = 0.29 | alpha = 0.08
  k = 25: x = {'ind_t': 'MACD', 'ind_p': [13, 33, 15]} | f(x) = 1.9881 | T = 0.28 | alpha = 0.07
  k = 26: x = {'ind_t': 'MACD', 'ind_p': [9, 32, 14]} | f(x) = 3.0010 | T = 0.26 | alpha = 0.06
  k = 27: x = {'ind_t': 'MACD', 'ind_p': [10, 32, 15]} | f(x) = 2.9623 | T = 0.25 | alpha = 0.06
  k = 28: x = {'ind_t': 'MACD', 'ind_p': [11, 29, 15]} | f(x) = 2.8012 | T = 0.24 | alpha = 0.05
  k = 29: x = {'ind_t': 'MACD', 'ind_p': [12, 28, 15]} | f(x) = 2.8012 | T = 0.23 | alpha = 0.05
  k = 30: x = {'ind_t': 'MACD', 'ind_p': [12, 29, 14]} | f(x) = 2.8012 | T = 0.21 | alpha = 0.04
  k = 31: x = {'ind_t': 'MACD', 'ind_p': [11, 25, 12]} | f(x) = 2.4843 | T = 0.20 | alpha = 0.04
  k = 32: x = {'ind_t': 'MACD', 'ind_p': [12, 24, 14]} | f(x) = 2.8184 | T = 0.19 | alpha = 0.03
  k = 33: x = {'ind_t': 'MACD', 'ind_p': [12, 24, 15]} | f(x) = 2.9717 | T = 0.18 | alpha = 0.03
  k = 34: x = {'ind_t': 'MACD', 'ind_p': [11, 27, 14]} | f(x) = 2.9190 | T = 0.17 | alpha = 0.03
  k = 35: x = {'ind_t': 'MACD', 'ind_p': [12, 28, 15]} | f(x) = 2.8012 | T = 0.17 | alpha = 0.03
  k = 36: x = {'ind_t': 'MACD', 'ind_p': [12, 27, 14]} | f(x) = 2.8276 | T = 0.16 | alpha = 0.02
  k = 37: x = {'ind_t': 'MACD', 'ind_p': [14, 25, 12]} | f(x) = 3.0677 | T = 0.15 | alpha = 0.02
  k = 38: x = {'ind_t': 'MACD', 'ind_p': [14, 25, 12]} | f(x) = 3.0677 | T = 0.14 | alpha = 0.02
  k = 39: x = {'ind_t': 'MACD', 'ind_p': [14, 25, 13]} | f(x) = 2.8276 | T = 0.14 | alpha = 0.02
  k = 40: x = {'ind_t': 'MACD', 'ind_p': [14, 26, 13]} | f(x) = 2.8276 | T = 0.13 | alpha = 0.01
  k = 41: x = {'ind_t': 'MACD', 'ind_p': [14, 26, 12]} | f(x) = 2.8577 | T = 0.12 | alpha = 0.01
  k = 42: x = {'ind_t': 'MACD', 'ind_p': [14, 26, 13]} | f(x) = 2.8276 | T = 0.12 | alpha = 0.01
  k = 43: x = {'ind_t': 'MACD', 'ind_p': [13, 25, 13]} | f(x) = 3.0677 | T = 0.11 | alpha = 0.01
  k = 44: x = {'ind_t': 'MACD', 'ind_p': [14, 25, 12]} | f(x) = 3.0677 | T = 0.10 | alpha = 0.01
  k = 45: x = {'ind_t': 'MACD', 'ind_p': [14, 25, 12]} | f(x) = 3.0677 | T = 0.10 | alpha = 0.01
  k = 46: x = {'ind_t': 'MACD', 'ind_p': [14, 25, 12]} | f(x) = 3.0677 | T = 0.09 | alpha = 0.01
  k = 47: x = {'ind_t': 'MACD', 'ind_p': [14, 25, 12]} | f(x) = 3.0677 | T = 0.09 | alpha = 0.01
  k = 48: x = {'ind_t': 'MACD', 'ind_p': [14, 25, 12]} | f(x) = 3.0677 | T = 0.09 | alpha = 0.01
  k = 49: x = {'ind_t': 'MACD', 'ind_p': [14, 25, 12]} | f(x) = 3.0677 | T = 0.08 | alpha = 0.01
  k = 50: x = {'ind_t': 'MACD', 'ind_p': [13, 24, 13]} | f(x) = 3.0919 | T = 0.08 | alpha = 0.01

  

  Notice how Grid Search performs an exhaustive search in the solution space (1728 evaluations), which requires significant time and computational effort. On the other hand, optimization with Simulated Annealing approaches the global maximum (Score = 3.20) after a few iterations.

📌 Notes

• Contributions are welcome! Open an issue or submit a pull request.
• Future improvements and new features may be added, including:
  • improve objective function with new weights and presets; ✅
  • improve optimization techniques;
  • file saving in cloud.

🤝 Support

This repository is independently maintained, in free time. If you find the code useful and wish to support its continued development, consider donating:

• PayPal

Your support helps keep the project alive and evolving, by adding new optimization methods, improvements and detailed documentation.