diff --git a/mcts/node.py b/mcts/node.py
index 3efae2e..fce7a21 100644
--- a/mcts/node.py
+++ b/mcts/node.py
@@ -472,8 +472,11 @@ def get_analysis_from_status_list(self, mode: str,
         """
         out = ""
         if mode == "cgos":
+            # 探索結果が未反映 (node_visits == 0) の場合はゼロ除算を避け、
+            # ニューラルネットワークの評価値で代用する
             cgos_dict = {
-                "winrate" : float(self.node_value_sum) / self.node_visits,
+                "winrate" : float(self.node_value_sum) / self.node_visits \
+                    if self.node_visits > 0 else float(self.raw_value),
                 "visits" : self.node_visits,
                 "moves" : []
             }
diff --git a/mcts/tree.py b/mcts/tree.py
index f10ce6c..862b85e 100644
--- a/mcts/tree.py
+++ b/mcts/tree.py
@@ -162,6 +162,10 @@ def search(self, board: GoBoard, color: Stone, time_manager: TimeManager, \
                         break
 
         if len(analysis_query) > 0 and interval == 0:
+            # 未反映のミニバッチ (最大 batch_size - 1 プレイアウト) を
+            # 反映してから解析結果を出力する
+            if len(self.batch_queue.node_index) > 0:
+                self.process_mini_batch(board)
             root = self.node[self.current_root]
             mode = analysis_query.get("mode", "lz")
             sys.stdout.write(root.get_analysis(board, mode, self.get_pv_lists))