Requirements
Backtesting.py requires Python 3.9 or later and the following packages, which are installed automatically:
| Package | Minimum version | Purpose |
|---|
numpy | 1.17.0 | Numerical array operations and indicator computation |
pandas | 0.25.1 | DataFrame-based OHLCV data handling and results |
bokeh | 3.0.0 (not 3.0.x or 3.2.x) | Interactive HTML chart output |
Install with pip
The primary installation method is pip:
This installs backtesting.py and all required dependencies into your active Python environment.
Virtual environment (recommended)
Using a virtual environment keeps your project dependencies isolated:
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install backtesting
Google Colab
In a Google Colab notebook, prefix the install command with !:
Then restart the runtime when prompted, or call importlib.reload on the relevant modules.
Verify the installation
Confirm the package is installed and importable:
import backtesting
print(backtesting.__version__)
from backtesting import Backtest, Strategy
from backtesting.test import GOOG
class BuyAndHold(Strategy):
def init(self):
pass
def next(self):
if not self.position:
self.buy()
bt = Backtest(GOOG, BuyAndHold)
stats = bt.run()
print(stats['Return [%]'])
Optional dependencies
Jupyter
Running backtests inside a Jupyter notebook is a common workflow. Install Jupyter separately if you do not already have it:
pip install notebook # classic notebook interface
# or
pip install jupyterlab # JupyterLab interface
In a Jupyter notebook, bt.plot() renders the interactive Bokeh chart inline in the cell output. In a standalone Python script, it opens the chart in your default browser.
bt.optimize() method supports Bayesian optimisation via sambo and displays a progress bar via tqdm. These are optional but recommended if you plan to use parameter optimisation:
pip install sambo tqdm ipywidgets
Indicator libraries
Backtesting.py is indicator-library-agnostic. Any library that produces NumPy arrays works. Popular choices include:
pip install ta-lib # TA-Lib (requires the C library; see TA-Lib docs)
pip install pandas-ta # pandas-ta
pip install tulipy # Tulipy
pip install scikit-learn # scikit-learn (for ML-based indicators)
Plotting in different environments
| Environment | Behaviour of bt.plot() |
|---|
| Jupyter notebook | Renders the interactive chart inline in the cell. |
| Standalone script | Opens the chart in your default web browser. |
| Google Colab | Renders inline; use bt.plot(open_browser=False) if the browser tab causes issues. |
| Headless / CI | Pass filename="output.html" to write the chart to a file without opening a browser. |
Bokeh versions 3.0.x and 3.2.x are explicitly excluded due to known incompatibilities. If you see Bokeh-related errors after install, run pip install "bokeh>=3.1,!=3.2.*" to force a compatible version.
Controlling Bokeh output mode
By default, backtesting.py auto-detects whether it is running inside a Jupyter notebook and configures Bokeh output accordingly. You can override this with set_bokeh_output:
from backtesting import set_bokeh_output
set_bokeh_output(notebook=True) # force inline notebook output
set_bokeh_output(notebook=False) # force file/browser output
Next steps
Once installed, follow the quickstart guide to run your first backtest in a few lines of code.