Order Book Inspired Automated Market Making
Tuan Tran,Duc A. Tran,Tam Nguyen
DOI: https://doi.org/10.1109/access.2024.3372402
IF: 3.9
2024-03-15
IEEE Access
Abstract:Decentralized exchanges are becoming a competitive necessity for Web3 users. However, they cannot beat centralized exchanges in terms of user experience. Due to expensive gas fees, the blockchain cannot implement the classic order book model which is the pillar for traditional finance exchanges. Instead, most decentralized exchanges operate on the Automated Market Maker (AMM) model using a pre-defined mathematical pricing rule. AMM is more efficient to run on the blockchain but the biggest tradeoff is impermanent loss for liquidity providers and price slippage for traders. This remains the most noticeable drawback of today's AMM. In this paper, we make the following contributions. First, we observe that if AMM is virtualized as an order book, its "order-book" shape is awkwardly different from that of a real-world order book. We argue that this is conceptually connected to the above weakness. We are thus motivated to design an AMM, the first of its kind, that mimics the price impact behaviors of real-world order books. Second, the proposed AMM, thanks to this property, significantly outperforms the state-of-the-art AMM in impermanent loss. Interestingly, our AMM can even result in impermanent gain. We are also better for large orders where price slippage is a concern. Third, another feature is that, while today's AMM typically requires a fixed inventory ratio for the liquidity pool, the new AMM allows this ratio to vary, giving liquidity providers flexible options for joining or exiting the pool. All these advantages are offered without losing desirable properties of an AMM regarding split-order exploitation, arbitrage risks, and liquidity continuity. Our findings are validated by theoretical analysis with mathematical proofs and, also, experimental evaluation which was comprehensively conducted using two real-world datasets and a synthetic dataset representing different market scenarios. Our research is the first in the literature on AMM design that factors in statistical properties from the order book model.
computer science, information systems,telecommunications,engineering, electrical & electronic