Skip to content

froggleston/freqtrade_analysis_notebook

Repository files navigation

freqtrade_analysis_notebook

A Jupyter Notebook to make backtesting, analysis and plotting of freqtrade strategies easier

Prerequisites

Clone this repository onto your machine.

You will need to install jupyter to run this notebook. Please follow any Jupyter installation instructions for your OS or environment.

You will also need extra dependencies for running the notebook. If using an existing freqtrade install (setup.sh script), activate your existing freqtrade venv environment and pip install:

cd /path/to/freqtrade
source .venv/bin/activate

pip install -r /path/to/freqtrade_analysis_notebook/requirements.txt

Docker

This has not been tested in any docker environments, so YMMV, but some instructions that might be useful are in this issue

Installation

Follow one of the two methods below:

Easiest installation

  • Copy all .py files and the RollingBacktestNotebook.ipynb into your base freqtrade folder, not the user_data/notebooks folder.
  • Set the freqtrade_dir variable in the notebook to "."

Easy-ish installation

  • Leave the .py files and RollingBacktestNotebook.ipynb notebook in your cloned directory.
  • Set the freqtrade_dir to the absolute path of your base freqtrade folder, e.g. /home/froggleston/freqtrade
  • If you set your user_data_dir and strategy_path in your config, they need to also be set to the absolute paths to your folders, e.g.
    • "user_data_dir" : "/home/froggleston/freqtrade/user_data"
    • "strategy_path": "/home/froggleston/freqtrade/user_data/strategies"

Running

Via the command line

If you copied the notebook and helper files to your freqtrade folder, navigate there:

cd /path/to/freqtrade

Or if you left the notebook and helper files in the cloned git folder, navigate there:

cd /path/to/freqtrade_analysis_notebook

Then run the jupyter server:

jupyter notebook --port 8889 --ip 0.0.0.0 --no-browser --NotebookApp.allow_origin='*'

The output from the command should contain three links to the Jupyter server. Pick any of these to open up a new Jupyter file browser tab in your preferred browser.

Via an IDE

If using an IDE like vscode, install an available jupyter extension and open the freqtrade folder. Then open the ipynb file and run the cells as normal.

Usage

Use the toolbar at the top of the plot to change behaviour or select/deselect data series.

Pan x only, pan x and y, or drag zoom to move around. The plot will automatically readjust to fit the candles in the current view.

Clicking Reset will zoom back out to the whole configured plot_tr timerange.

Click on a data point to view the indicator values in the main plot for that x index. When selected a dashed line will appear through all subplots making it easier to keep track of the selected date index across large numbers of subplots. Double clicking anywhere in the plot removes the highlight.

Mouseover main plot and subplot data series to see individual values.

Known Issues

bokeh

  • If you scroll wheel in or out too fast it breaks the plot and you have to regenerate it by re-running the cell.
  • Click selection requires clicking on a data point and not anywhere on the plot.
  • Clicking a data series in the legend will only hide the line plot for that series and not the scatter (which is used for click selection).
  • Hover tooltips for the main plot are very difficult to get working and maintain readability, hence the summary table on the right.
  • The summary table on the right cannot be auto-sorted, so to get the indicator list for a click selection point, click the value column twice to sort descending. (After you've done this once however, the table maintains that sort order when selecting other data points).
  • The signal tooltips appear overlaid on the candle tolltips, making the candle data hard to read.
  • No OHLCV data in the summary table.

plotly

  • works but slow
  • not as fancy

About

A Jupyter notebook to analyse freqtrade results

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •