Bit by Bit: How to realistically simulate a Crypto-Exchange#
2. Affiliation:#
The first author’s affiliation is University of Southampton.
3. Keywords:#
Financial Exchange, Continuous Double Auction, Market Simulator, Multi-Agent System, Cryptocurrency
4. Urls:#
Paper link: https://dl.acm.org/doi/10.1145/3490354.3494380, Github: None
5. Summary:#
(1): The background of this article is to create a market simulator that can replicate market dynamics accurately.
(2): Previous methods have been limited to well-established equities, FX, and futures markets, and they struggle to replicate some of the key market properties. This paper proposes an approach to simulate a cryptocurrency exchange, where market properties are both hand-tuned and empirically optimized using data collected from the Bitcoin/USD market.
(3): The proposed research methodology is a multi-agent simulation, which effectively represents a cryptocurrency exchange by populating a high-fidelity market simulator with a set of low intelligence agents to represent distinct types of market participants. Fine-tuning the ratios and hyperparameters of these agents against empirical findings, the authors present a simulation with the ability to closely mimic expected market behavior.
(4): The performance of the proposed method achieves the ability to closely mimic expected market behavior, making it a valuable tool for testing, competing, and comparing strategies.
6. Conclusion:#
(1): The significance of this piece of work is to develop a methodology for simulating a cryptocurrency exchange that can closely mimic expected market behavior, providing a valuable tool for testing, competing, and comparing strategies.
(2): Innovation point: The proposed method stands out in its ability to simulate cryptocurrency exchanges with high fidelity, capturing key market properties that are difficult to replicate with previous approaches. Performance: The simulation demonstrates the ability to closely mimic expected market behavior. Workload: The methodology requires fine-tuning the ratios and hyperparameters of low intelligence agents, which may be time-consuming.