Reinforcement learning in trading

In this master thesis we examine the application of reinforcement learning to the cryptocurrency trading task. This method is chosen because its framework fits perfectly to the problem at hand. Instead of predicting prices, the model tries to predict the best action among a buy, hold and sell options which should lead to the highest expected return. The primary goal of this work is to investigate whether on a rising market the reinforcement learning model can outperform the simple “Buy & Hold” strategy. The work is oriented on a practical application of the presented methods, and working implementation of the tested models is provided in detail.