Brain Machine Learning Signals on Cryptocurrencies
Overview
Brain has developed a machine learning framework to create middle-frequency trading signals based on systematic investment strategies designed for integration into client operations. This approach combines rigorous quantitative modelling with flexible parameterization, enabling consistent performance across various cryptocurrencies.
Key Objectives
The key objectives of the strategies are the following:
Key Objectives
The key objectives of the strategies are the following:
- Return maximization: exploitation of inefficiencies in the cryptocurrency market through optimized parametric strategies that capture time-based inefficiencies, combining momentum and mean- reversion trends.
- Volatility reduction: diversification by combining weakly correlated strategies.
- Ease of implementation: middle-frequency trading (e.g. hourly), also suitable for initial manual or paper- trading setups before automation.
- Scalability: adaptable to larger trading volumes and multiple asset classes.
- Robustness: effective under varying market conditions, with strong overfitting control via proprietary validation.
The dataset is updated at intraday frequency (e.g., hourly), with several years of historical data available for testing depending on the asset. Data is delivered daily as CSV files published in an S3 bucket. Hourly strategies on liquid pairs such as ETHUSDT and BTCUSDT have shown strong results, with a Sharpe ratio well above one as highlited in the related factsheet. A dashboard to monitor the live signals is available, and dedicated login credentials can be provided upon request.