Real Time ML trading BOT
Developing this to learn more about applying ML to real-time trading strategies. In future I plan to add other sources of data such as:
- Sentiment (Twitter/articles)
- On-chain analytics
- Specific wallet movements (whales moving crypto around)
Currently just gathering test/train/validation data for creating an algorithm.
Idea:
- LOB tells us in the short term what investors are thinking:
- Is there a large sell wall close to the market price?
- Have a lot of bids been placed in the last 30 seconds?
- Were lots of cancellations just made on a particular side?
- On short enough time frames can this be used to quickly enter and exit trades?
Learning Phase:
- Gathers level 2 LOB data using the Coinbase Pro API
- Generalises LOB curves to a set of discrete 'feature points'
- Future market price will be to train the ML model based on the feature set