Net Impact of Illiquidity and Persistence Parameters on Optimal Hedging Positions

Characterize the net effect of interactions among the illiquidity parameters α and β appearing in the buy/sell cost functions F^a and F^b and the impact persistence parameters λ_a and λ_b governing the processes A_t and B_t on the optimal hedging positions X_t produced by the deep reinforcement learning hedging approach, particularly near maturity under pin risk.

Background

In the pin risk analysis with impact persistence, the authors observe non-monotonic behavior: the DRL policy with intermediate persistence sometimes takes larger positions than both the no-persistence and permanent-impact cases, indicating complex parameter interactions.

They explicitly state that the combined influence of market impact parameters on optimal positions is unclear a priori, highlighting the need for further characterization of these interactions within the deep hedging framework.