Master's Thesis : Low latency arbitrage detection system

Abstract

Arbitrage is a trading strategy which is used to detect opportunities to earn money by exploiting the fact that given assets have different values in different places. In this work, we design a system capable of detecting arbitrage opportunities in exchange platforms by annotating simple cycles in a directed multigraph. The system that is designed is resilient and distributed. As a result, it can be scaled in order to annotate more simple cycles while keeping a stable and quick response time. Tests performed using a single node demonstrated that the system is able to maintain an average response time of 0.579880 milliseconds while monitoring 8192 simple cycles simultaneously. The expected duration of the arbitrage opportunities found by the running system over a period of 4 days is 1.18649 seconds.