RL for execution algos should be trained with careful attention to:
Select an answer to reveal the explanation.
Short Explanation and Infographic
RL execution needs realistic impact sims, constrained actions, offline eval — not live-from-day-one.
Full explanation below image
Full Explanation
Sim-to-real gaps are acute in market microstructure RL.