Predictive Market Making via Machine Learning

Abstract Market making (MM) is an important means of providing liquidity to the stock markets. Recent research suggests that reinforcement learning (RL) can improve MM significantly in terms returns. In latest work on RL-based MM, reward a function equity returns, calculated based its current price, and inventory agent. As result, agent’s return maximised provided. If price movement known this information optimally utilised, there potential be further improved. Important questions are, how predict movement, utilise such prediction? paper, we introduce concept predictive market marking (PMM) present our method for PMM, which comprises agent deep neural network (DNN)-based predictor. A key component PMM consolidated equation (CPE), amalgamates equity’s predicted prices into used generate ask bid quotes reflect both future movement. Our evaluated against state-of-the-art (RL-based MM) traditional method, using ten stocks three exchange traded funds (ETFs). Out-of-sample backtesting showed outperformed two benchmark methods.