Backtesting.py

Backtesting.py suits a focused question about entry and exit rules for one instrument. Supply a pandas table with Open, High, Low and Close columns, plus optional Volume. It is not designed for stock selection, arbitrage across assets or whole-portfolio rebalancing.

Define indicators in a Strategy’s init() method and the per-bar decisions in next(). Construct a Backtest with the data, strategy, starting cash and commission assumption. Calling run() returns summary metrics, the equity curve and detailed trades; plot() creates an interactive view. Installation uses pip install backtesting, without registering for a hosted service.

Strategy parameters can be compared through optimize(). Supply the allowed values and a constraint that rejects unsuitable combinations, such as a fast moving average period that is not shorter than the slow period. The objective can be a selected result metric, rather than only final portfolio value.

The simulation makes decisions once each price bar is available. It cannot infer the path inside a candle; finer timing requires finer input data. Market orders normally use the next bar’s open, with a separate option to use the current close. These choices affect the meaning of the resulting trades. The project’s AGPL licence also needs attention when modifying the framework for a network service.