Figure 2
A line graph showing Sharpe ratio convergence over timesteps for different agents. The x-axis represents timesteps ranging from 20,000 to 140,000, and the y-axis represents the Sharpe ratio ranging from -0.2 to 1. The graph includes multiple lines representing different agents: A 2 C, D D P G, P P O, R E C U R R E N T P P O, S A C, T D 3, and T Q C. P P O achieves the highest and most consistent improvement, followed by T D 3 and R E C U R R E N T P P O, whereas S A C and T Q C remain comparatively flat. All values are approximated.Checkpoint-based Sharpe ratio convergence for representative training runs. PPO (green) achieves the highest and most consistent improvement, followed by TD3 (cyan) and RecurrentPPO (light purple), whereas SAC (light orange) and TQC (pink) remain comparatively flat