Arbitrage bots in experimental asset markets

Trading algorithms are an integral component of modern asset markets. In twin experimental markets for long-lived correlated assets we examine the impact of alternative types of arbitrage-seeking algorithms. These arbitrage robot traders vary in their latency and whether they make or take market liquidity. All arbitrage robot traders we examine generate greater conformity to the law-of-one-price across the twin markets. However, only the liquidity providing arbitrage robot trader moves prices into closer alignment with fundamental values. The reduced mispricing comes with varying social costs; arbitrage robot traders’ gains reduce the earnings of human traders. We identify factors which drive differences in human trader performance and find that the presence of an arbitrage robot trader has no disproportionate effect with respect to these factors on subjects’ earnings.