Reinforcement Learning target performance estimation on K2D13 500-1-late