SIGNATURE WORK
CONFERENCE & EXHIBITION 2022

Trading through automated market makers (AMMs) on Ethereum is gaining a significant tendency in Decentralized finance (DeFi). However, the mechanism of slippage tolerance set by trades in AMM offers opportunities for predatory trading bots to gain extra profits by conducting invisible sandwich attacks. Considering that few studies have been conducted to investigate the trading behavior within the attack from a micro perspective, our study proposed an agent-based model to simulate the sandwich game to evaluate how different players would perform under the predate, especially when trading Uniswap. We adopt Heimbach et al.’s (2022) model, which could automatically adjust traders’ slippage tolerance to minimize costs instead of using the initially fixed slippage set in the AMM. Moreover, we further revise the model by adding other kinds of behavioral players considering bounded rationality. We demonstrate the impact of bounded rationality on trading behaviors concerning the trading loss. This study would provide a micro foundation for the macro context, specifically the DeFi token transaction network structure, and economic performance evolution. The novel application scenarios and proposed modeling methods would greatly inspire future research on agent-based modeling for the cryptocurrency market for more complex environments.