GitHub - zerotech-studio/zack: Backtesting engine in Zig

6 min read Original article ↗

📈 Zack: A Simple Backtesting Engine in Zig 📉

Welcome to Zack! This project is a lightweight yet powerful backtesting engine for trading strategies, written entirely in Zig ⚡. It allows you to test your trading ideas against historical market data to see how they might have performed.

🤔 What is it?

Zack simulates the process of trading based on a predefined strategy using historical OHLCV (Open, High, Low, Close, Volume) data. It processes data bar-by-bar, generates trading signals, simulates order execution, manages a virtual portfolio, and reports the performance.

✨ Why Zig?

Zig offers several advantages for this kind of application:

  • Performance: Zig compiles to fast, efficient machine code, crucial for processing potentially large datasets quickly.
  • Memory Control: Manual memory management allows for fine-tuned optimization and avoids hidden overhead.
  • Simplicity: Zig's focus on simplicity and explicitness makes the codebase easier to understand and maintain (no hidden control flow!).

⚙ī¸ How it Works: The Engine Flow

The backtesting process is driven by an event loop within the BacktestEngine. Here's a breakdown of the core components and their interactions:

  1. Initialization:

    • The main function loads configuration (config/config.json, config/<strategy_name>.json) and CSV data (data/<data_file>.csv) using AppContext.
    • It then initializes the BacktestEngine, which in turn sets up all other components.

  2. The Event Loop (BacktestEngine.run): The engine iterates through the historical data bar by bar. For each current_bar:

    • Data Handling (DataHandler): Provides the current_bar (parsed from the CSV data). It uses Bar.parse to convert CSV rows into structured Bar objects.
    • Portfolio Update (Portfolio): The portfolio calculates its current market value based on the current_bar.close price and any open Position. It records the total equity at this point in time (EquityPoint).
    • Lookahead: The engine fetches the next_bar from the DataHandler. This is crucial for simulating execution delays.
    • Strategy Signal (BuyAndHoldStrategy): The current strategy (BuyAndHoldStrategy in this case) receives the current_bar data and the portfolio's state (e.g., has_position). It decides if a trading signal (Signal) should be generated based on its rules (e.g., bar.open >= buyAt). 
      // Inside strategy.generateSignal
if (!has_position and bar.open >= @as(f64, @floatFromInt(self.config.buyAt))) {
    return Signal{ .type = .Long }; // Generate Buy signal
}
    • Order Generation (Portfolio): If a Signal is received, the Portfolio determines the details of the Order (e.g., MarketBuy, quantity). It might use the current_bar's price for approximate sizing. 
      // Inside portfolio.handleSignal
const quantity_to_buy = cash_to_use / current_bar.close;
return Order{ .type = .MarketBuy, .quantity = quantity_to_buy };
    • Execution Simulation (ExecutionHandler): The Order is sent to the ExecutionHandler. Crucially, it uses the next_bar.open price to simulate the fill, modeling the delay between deciding to trade and the order actually executing in the next period. It also calculates commission. 
      // Inside execution_handler.executeOrder
const fill_price = next_bar.open; // Fill at NEXT bar's open
const commission = COMMISSION_PER_TRADE;
return Fill{ /* ...details... */ };
    • Portfolio Update (Portfolio): The resulting Fill event is sent back to the Portfolio, which updates its current_cash, position, and current_holdings_value. 
      // Inside portfolio.handleFill
self.current_cash -= cost;
self.position = Position{ .entry_price = fill.fill_price, /*...*/ };
    • Loop: The process repeats with the next_bar becoming the current_bar.

  3. Results: After processing all bars, the BacktestEngine.logResults function prints a summary of the performance.

đŸŽ¯ Current Strategy: Buy and Hold

The engine currently implements a simple "Buy and Hold" strategy (src/engine/strategy.zig).

  • Logic: It generates a single "Buy" (Long) signal when the open price of a bar crosses above a predefined threshold (buyAt), but only if the portfolio does not already hold a position. It never generates a sell signal; the position is held until the end of the backtest.
  • Configuration: The buyAt threshold is set in the strategy's configuration file (e.g., config/buy-and-hold.json):

🛠ī¸ Configuration

The main simulation parameters are set in config/config.json:

{
  "budget": 10000,         // Initial capital for the simulation
  "strategy": "buy-and-hold.json", // Which strategy config file to load from config/
  "data": "btc.csv"        // Which data file to load from data/
}

📊 Data Format

The engine expects OHLCV data in CSV format in the data/ directory:

timestamp,open,high,low,close,volume
2024-01-01T00:00:00Z,42000.00,42100.00,41900.00,42050.00,100.50
2024-01-01T01:00:00Z,42050.00,42200.00,42000.00,42150.00,120.75
...
  • timestamp: ISO 8601 format (currently treated as a string).
  • open, high, low, close, volume: Floating-point numbers.

📁 Project Structure

.
├── build.zig        # Zig build script
├── config/
│   ├── config.json         # Main configuration
│   └── buy-and-hold.json   # Strategy-specific parameters
├── data/
│   └── btc.csv             # Sample OHLCV data
├── src/
│   ├── main.zig            # Application entry point
│   ├── csv/                # CSV parser utility
│   │   └── csv-parser.zig
│   ├── engine/             # Core backtesting engine components
│   │   ├── common.zig          # Shared structs (Bar, Signal, Order, Fill, Position)
│   │   ├── data_handler.zig    # Loads and provides Bars
│   │   ├── strategy.zig        # Strategy logic (BuyAndHoldStrategy)
│   │   ├── portfolio.zig       # Manages cash, position, equity
│   │   ├── execution_handler.zig # Simulates order fills
│   │   └── backtest_engine.zig # Orchestrates the simulation loop
│   └── utils/              # Utility functions
│       ├── load-config.zig   # JSON config loading
│       └── logger.zig        # Simple logging utility
└── README.md       # This file

🚀 How to Run

  1. Ensure you have Zig installed (see ziglang.org).

  2. Clone the repository.

  3. Run the simulation using the Zig build system:

    Alternatively, run the main file directly:

📝 Example Output

Running the engine with the default configuration and sample btc.csv data produces output similar to this:

ℹī¸ [INFO] ⚙ī¸ Configuration Loaded:
ℹī¸ [INFO]   Budget:   10000
ℹī¸ [INFO]   Strategy: buy-and-hold.json
ℹī¸ [INFO]   Data File:btc.csv
ℹī¸ [INFO] 📈 Strategy Settings:
ℹī¸ [INFO]   Buy At Threshold: 1000

--- Starting Backtest Run ---
PORTFOLIO: Received LONG signal, generating MarketBuy order for ~0.23547619047619048 units.
EXECUTION: Executing MarketBuy order for 0.23547619047619048 units @ 42050 (Commission: 1)
PORTFOLIO: Handled MarketBuy fill. Cash: 9.99999999999909, Position Qty: 0.23547619047619048, Entry: 42050
--- Backtest Run Finished ---

ℹī¸ [INFO]
📊 Backtest Results:
ℹī¸ [INFO]   Initial Capital: 10000.00
ℹī¸ [INFO]   Final Equity:    10443.75
ℹī¸ [INFO]   Total Return:    4.44%
ℹī¸ [INFO]   Ending Position: 0.2355 units @ entry 42050.00
ℹī¸ [INFO]   (More detailed performance metrics TBD)

Application finished successfully.

(Note: Exact float values might differ slightly)

Key Observations from Output:

  • The Long signal is generated based on the first bar (open=42000 >= buyAt=1000).
  • The MarketBuy order is executed at the open price of the second bar (42050), as expected due to the one-bar delay simulation.
  • The final equity reflects the initial capital minus the buy cost plus the value of the holding at the final bar's close price.

🔮 Future Work

  • Implement more sophisticated performance metrics (Sharpe Ratio, Max Drawdown, etc.).
  • Implement more strategies.
  • Implement technical indicators.
  • Add comprehensive unit tests for all engine components.

Contributions and suggestions are welcome!