Trading in CEXs and DEXs with Priority Fees and Stochastic Delays

Battle Count: 106

This paper is highly relevant to quantitative trading, particularly in the DeFi/crypto space. It provides the first continuous-time optimal control framework for CEX-DEX trading that accounts for both priority fees and stochastic execution delays. The results directly inform how traders should manage latency risk, select priority fees, and time their order submissions across centralized and decentralized venues. The 18.2% outperformance of optimal fee selection over random selection demonstrates significant practical value. The framework is directly applicable to algorithmic execution strategies in crypto markets.

The paper involves highly advanced mathematics: viscosity solutions of HJB quasi-variational inequalities, dynamic programming principles for mixed control problems, stochastic differential equations with jumps, and Poisson process theory. Numerical implementation requires solving systems of PDEs with non-local operators (the intervention operator M and execution operator J), handling multiple pending order states (combinatorial explosion with K pending orders and N fee levels), and implementing finite difference or finite element schemes for the HJB-QVI. The state space dimensionality grows with the number of pending orders and priority fee configurations. The theoretical framework is sophisticated, and practical implementation would require significant expertise in both stochastic control and numerical PDE methods.

The paper provides detailed model parameters calibrated to ETH-USDC market data (Section 4.2), including initial prices, volatilities, mean-reversion speed, pool depth, and fee/delay vectors. However, no code repository or raw data is provided. The numerical schemes are described as 'standard' without specifying exact discretization methods. The mathematical proofs are complete and self-contained, enabling theoretical reproduction. The viscosity solution framework and HJB-QVI characterization are fully specified.