diff --git a/src/rlai/policy_gradient/monte_carlo/reinforce.py b/src/rlai/policy_gradient/monte_carlo/reinforce.py
index dc59976..7d62a31 100644
--- a/src/rlai/policy_gradient/monte_carlo/reinforce.py
+++ b/src/rlai/policy_gradient/monte_carlo/reinforce.py
@@ -390,9 +390,6 @@ def improve(
 
                 plt.close()
 
-            if pdf is not None:
-                pdf.close()
-
             # plot per-episode metrics
             if len(environment.metric_episode_value) > 0:
                 for metric in environment.metric_episode_value:
@@ -414,6 +411,9 @@ def improve(
 
                 plt.close()
 
+            if pdf is not None:
+                pdf.close()
+
         num_fallback_iterations = 0
         if (
             training_pool is not None and