RAmmStein: Regime Adaptation in Mean-reverting Markets with Stein thresholds - Optimal Impulse Control in Concentrated AMMs

Battle Count: 92

Highly relevant to quantitative trading in DeFi/AMM contexts. The impulse control formulation, OU mean-reversion modeling, and regime-aware decision-making are directly applicable to market making and liquidity provision strategies. The paper bridges stochastic optimal control theory with practical DeFi operations. However, it is specifically focused on AMM liquidity management rather than traditional order-book trading, limiting broader applicability to conventional quant strategies.

Moderate complexity. The DDQN architecture is standard (2 hidden layers, 128-64 neurons). The main complexity lies in: (1) building a realistic AMM environment simulator with fee accrual, gas costs, and swap mechanics; (2) implementing rolling OLS for OU parameter estimation; (3) designing the 8-dimensional state representation; (4) calibrating the reward function with appropriate scaling. The theoretical HJB-QVI derivation is mathematically involved but the actual implementation is a standard DDQN training loop.

The paper provides detailed hyperparameters (Table I), environment parameters (Table II), state representation, action space, reward function, and network architecture. However, no code repository is linked. The dataset (Coinbase WebSocket ETH-USD trades) is publicly accessible but the exact preprocessing pipeline and volume scaling factor (alpha=0.10) would need to be replicated. The OU parameter estimation via rolling OLS is well-specified.