Profitability of high-frequency market-making under market instability and manipulative behaviors

Abstract

This study investigates the relationship between market-making performance and risk factors in high-frequency trading. Using a deep neural network to approximate the optimal market-making strategy, we generate synthetic datasets capturing expected rewards under varying market conditions, with focus on stability shaped by market and limit order dynamics Our analysis yields two key insights. First, market-making profitability is positively linked to instability in market order flows, especially when long-term correlations between buy and sell orders are strong: a 10% increase in buy-sell clustering raises the expected profit by 12.2% on average. Second, instability in limit order flows arising from manipulative trading substantially reduces profitability: a 10% increase in the frequency lowers the expected profit by 4% on average. Finally, we discuss the regulatory implications of these findings quantifying the harm caused by manipulative orders.