Background

The instant I discovered algorithmic trading, I knew I would need a tool to measure the performance of various trading strategies. I decided to design my own backtester-- one that would allow me to quickly test parameterized algorithms over years of historical data and thousands of tickers.

Project

So I created a backtesting application in C++. It runs entirely on the command line, and the tedious computations are done on multiple threads to maximize efficiency. I used data from Alpha Vantage and used Doxygen for documentation.

Considering the difficulty of obtaining historical data, the application has been designed with limitations:

• The only market data available comes from the equities market
• All data comes in daily tradebars (open, high, low, close, volume, timestamp)
• Assets come from only a few US exchanges
• The backtester does not take into account fees or slippage

The backtester allows you to customize any strategy-- you can modify the initial equity, start date, end date, and other parameters. You can open/close limit/market orders that short/long any ticker for any duration.

To begin, simply create a class that inherits from Strategy:

class ExampleStrategy : public Strategy {
    public:
        ExampleStrategy();
        void onInitialize();
        void onDay();
        void onOrderFilled(const shared_ptr<Order>& order, const shared_ptr<Position>& position, double price, int shares);
};

Once you have created your class, add it to the driver in driver.cpp:

Driver::Driver() {
    strategies.push_back(make_shared<ExampleStrategy>());
}

All boilerplate is defined in Strategy-- here, you can pull market data, create orders, view positions, and much more.

For more detailed instruction, view the repository.