A Comparison of Different Automated Market-Maker Strategies

Financial markets such as stock exchanges and electronic prediction markets frequently use the services of an entity called the market-maker to ensure that the market’s traders can make their transactions. Recently, several strategies that can be used by market-makers to control market trading prices have been proposed by various researchers. A detailed comparison of these market maker strategies using real trading data extracted from financial markets is essential to understanding the relative merits and requirements of the different market-maker strategies. We address this aspect of market-maker strategies by empirically comparing different strategies with data obtained from the NASDAQ market. Our results show that a reinforcement learning-based strategy performs well in maintaining low spread as well as in obtaining high utilities, whereas other strategies only succeed in either maintaining low spread or outperforming others in utilities. 1