RL-Exec: Impact-Aware Reinforcement Learning for Opportunistic Optimal Liquidation, Outperforms TWAP and a Book-Liquidity VWAP on BTC-USD Replays

In a recent study, RL-Exec, a reinforcement learning agent, was developed to optimize liquidation strategies in BTC-USD limit-order books. The evaluation period spanned from January to February 2020, during which RL-Exec was rigorously tested against established methods such as TWAP and a VWAP-like baseline. The results showed that RL-Exec significantly outperformed these traditional strategies, with performance gaps increasing over time—by 2-3 basis points at 30 minutes, 7-8 basis points at 60 minutes, and 23 basis points at 120 minutes. This improvement in liquidation efficiency is vital for traders seeking to minimize costs and maximize returns, ultimately contributing to enhanced market stability.