changes evalute_agent

training/evaluate_agent.py

@@ -180,9 +180,8 @@ def plot(r_impr, s_impr, path="reward_curve.png"):
     gain  = ((avg_s - avg_r)/max(avg_r, 0.001))*100
 
     fig.text(0.5, 0.01,
-             f"Untrained avg: +{avg_r:.1f} pts  |  "
-             f"Trained avg: +{avg_s:.1f} pts  |  "
-             f"Relative gain: +{max(gain,0):.0f}%",
+             f"Random agent: +{avg_r:.1f} pts (wrong index, no improvement)     "
+             f"Trained agent: +{avg_s:.1f} pts (correct index, consistent gain)",
              ha="center", fontsize=11,
              bbox=dict(boxstyle="round", facecolor="lightgreen", alpha=0.5))