Continious Price Prediction for Market Making in Cryptocurrencies

The emergence of cryptocurrencies has introduced new dynamics in financial markets, necessitating the development of advanced strategies for market making. This work aims to address the problem of constructing an optimal algorithmic trading strategy for market making in cryptocurrencies. Using historical data from cryptocurrency exchanges, we generate statistically significant features based on Ornstein-Uhlenbeck equations that model the financial process of price formation, including trend, mean reversion tendency, volatility trends, and the coefficient in Brownian motion. Additional features are derived from liquidity information in the order book, such as volumes at each level and order book imbalance. These features are fed into a machine learning model to predict new positions for the best bid and ask offers, enabling the placement of market maker orders in the order book. The subsequent backtest of the strategy aims to maximize PnL and explores alternative approaches to creating optimal market making strategies. This method represents a comprehensive attempt to apply statistical methods and machine learning technologies to increase profitability and improve the efficiency of market making in the volatile cryptocurrency market.