AI Trader Bench Alpha. Currently just downloads the data and nuggets it to visualize it. Build path is now active.

pip install -r requirements.txt

Great installation instructions in https://mrjbq7.github.io/ta-lib/install.html so I won't reinvent the wheel.

On Windows: (tested)

Easiest way of doing this was going to https://www.lfd.uci.edu/~gohlke/pythonlibs/#python-snappy

Download python_snappy‑0.5.4‑cp37‑cp37m‑win_amd64.whl matches to Python 3.7 on a 64 bit system

pip install python_snappy‑0.5.4‑cp37‑cp37m‑win_amd64.whl

On DEB-based Linux: (untested)

sudo apt-get install libsnappy-dev
pip install python-snappy 

On RPM-based Linux: (untested)

sudo yum install libsnappy-devel
pip install python-snappy 

On Mac: (tested with Mike)

brew install snappy
pip install python-snappy

python app.py

Then go to http://localhost:5000/register

Then sign in http://localhost:5000/login

If you wish to use an SQL server instead of SQLite go to http://localhost:5000/setup and restart the app.

https://www.kaggle.com/jorijnsmit/binance-full-history

Binance is a decent exchange with a proven track record. All data going back to starting pairs. Thanks Pierre! Requires Signup

https://www.dukascopy.com/swiss/english/marketwatch/historical/

Dukascopy has always been an excellent source of data for backtesting. 1M bars go back for 3 years which is enough. Thanks Huub! Requires Signup

No idea yet of a good source which has 1M bars with volume. Probably IB is a good source: https://interactivebrokers.github.io/tws-api/historical_data.html

graph LR
    A>Connection] --> B[(Data 1m)]-->C[Samples]
    D([TimeFrame])-->C
    C-->E[Nuggets]
    F([Riches])-->G(Enrichments X)-->E
    H([Dependant y])-->E
    E-->I((Observatorium))
    E-->J((AI))
    I-->K[View]
    I-->L[Correlate]
    I-->M[Feature Selection]