Cryptocurrencies trading

This is a project that I did for myself in 2020. I am showing you this project to demonstrate my capability of working on a big project by myself. However, the code quality is far from good. I invite you to have a look at another project person_detection which has a better code quality.

Structure of the project

This project is structured as follows.

• notebooks contains Jupyter Notebooks. Those are useful to explore the data efficiently.
• shell_scripts contains scripts whose purpose is to automate the deployment and run the source code on a remote machine.
• src contains the source code.
  • feature engineering : all the preprocessing to extract attributes of the data from which the prediction will be made.
  • strategies : contains algorithms for automatic trading.
  • utils contains some common elementary algorithms useful for this project. Among others, the market_rules which simulates the fees applied to the trading actions, and wallet which tracks the amount of currencies (crypto of fiat) left in the user's wallet.
  • download_big_data.py is a script that downloads all the data from the trades made on Kraken.

TODO (General)

• Make features which reflect the market model closer to reality.
  • WiP Build an approximative cumulative histogram from the data.
  • Done : Base algorithms on the selling price to decide whether to buy or not.
  • Analyze what is the exact probabilistic criterion to buy
  • Re-Run data collection after interruption

TODO (Data Science)

• What is the daily volume traded on the market ?
• What kind of actors are playing on this market ?
• Is there a correlation with the values of CAC40, S&P500 or other european indicators ?
• Visualize the data, with dimension reduction methods for instance.
• What is the average volatility ?
• Scrap Twitter data and other social media, like Reddit.
• Scrap data from specialized press.
• Measure daily tendencies.
• Modelize what would give multiple deals at once, e.g. EUR -> XBT -> ETH -> XBT -> EUR
• See how markets communicate. Maybe Kraken is not the only interesting market.
• See the strategy of big actors : the big actors have probably interesting information.