diff --git a/freqtrade/templates/freqaiprimer.py b/freqtrade/templates/freqaiprimer.py
index 0d40868f..9fa8eba8 100644
--- a/freqtrade/templates/freqaiprimer.py
+++ b/freqtrade/templates/freqaiprimer.py
@@ -1588,12 +1588,47 @@ class FreqaiPrimer(IStrategy):
                             # 确保概率在 [0, 1] 范围内（分类器输出可能有浮点误差）
                             entry_prob = max(0.0, min(1.0, entry_prob))
                             entry_threshold = self.ml_entry_signal_threshold.value
-                            if entry_prob < entry_threshold:
-                                rejected_conditions.append(f"ML审核官: entry_signal 概率 {entry_prob:.2f} < 阈值 {entry_threshold:.2f}")
-                                self.strategy_log(f"[{pair}] {rejected_conditions[-1]}")
-                                ml_rejected = True
+                            
+                            # 获取 exit_signal 值
+                            exit_prob = None
+                            if '&s-exit_signal' in df.columns:
+                                exit_prob = float(last_row['&s-exit_signal'])
+                            elif '&-exit_signal_prob' in df.columns:
+                                exit_prob = float(last_row['&-exit_signal_prob'])
+                            elif '&-s-exit_signal_prob' in df.columns:
+                                exit_prob = float(last_row['&-s-exit_signal_prob'])
+                            elif '&-exit_signal' in df.columns:
+                                val = last_row['&-exit_signal']
+                                if isinstance(val, (int, float)):
+                                    exit_prob = float(val)
+                                else:
+                                    # 文本标签时，简单映射为 0/1
+                                    exit_prob = 1.0 if str(val).lower() in ['exit', 'sell', '1'] else 0.0
+                            
+                            # 应用新规则：entry_signal > exit_signal * 1.33 且 > 阈值
+                            signal_ratio_threshold = 1.33
+                            if exit_prob is not None:
+                                # 确保 exit_prob 在 [0, 1] 范围内
+                                exit_prob = max(0.0, min(1.0, exit_prob))
+                                
+                                if entry_prob < entry_threshold:
+                                    rejected_conditions.append(f"ML审核官: entry_signal 概率 {entry_prob:.2f} < 阈值 {entry_threshold:.2f}")
+                                    self.strategy_log(f"[{pair}] {rejected_conditions[-1]}")
+                                    ml_rejected = True
+                                elif entry_prob <= exit_prob * signal_ratio_threshold:
+                                    rejected_conditions.append(f"ML审核官: entry_signal ({entry_prob:.2f}) 未超过 exit_signal ({exit_prob:.2f}) 的 {signal_ratio_threshold:.2f} 倍")
+                                    self.strategy_log(f"[{pair}] {rejected_conditions[-1]}")
+                                    ml_rejected = True
+                                else:
+                                    self.strategy_log(f"[{pair}] ML 审核官允许入场: entry_signal 概率 {entry_prob:.2f} >= 阈值 {entry_threshold:.2f} 且超过 exit_signal ({exit_prob:.2f}) 的 {signal_ratio_threshold:.2f} 倍")
                             else:
-                                self.strategy_log(f"[{pair}] ML 审核官允许入场: entry_signal 概率 {entry_prob:.2f} >= 阈值 {entry_threshold:.2f}")
+                                # 如果没有 exit_signal，只检查阈值
+                                if entry_prob < entry_threshold:
+                                    rejected_conditions.append(f"ML审核官: entry_signal 概率 {entry_prob:.2f} < 阈值 {entry_threshold:.2f}")
+                                    self.strategy_log(f"[{pair}] {rejected_conditions[-1]}")
+                                    ml_rejected = True
+                                else:
+                                    self.strategy_log(f"[{pair}] ML 审核官允许入场: entry_signal 概率 {entry_prob:.2f} >= 阈值 {entry_threshold:.2f} (无 exit_signal 数据)")
                     else:
                         self.strategy_log(f"[{pair}] 无数据分析数据，跳过ML审核")
                 else:
@@ -1661,10 +1696,10 @@ class FreqaiPrimer(IStrategy):
         try:
             # 获取当前的ML预测数据
             df, _ = self.dp.get_analyzed_dataframe(pair, self.timeframe)
-            if len(df) > 0 and len(df) > 0:
+            if len(df) > 0:
                 last_row = df.iloc[-1]
                 
-                # 获取当前入场信号概率（反映当前市场入场条件）
+                # 获取当前入场信号概率
                 current_entry_prob = None
                 if '&s-entry_signal' in df.columns:
                     current_entry_prob = float(last_row['&s-entry_signal'])
@@ -1673,6 +1708,15 @@ class FreqaiPrimer(IStrategy):
                 elif '&-s-entry_signal_prob' in df.columns:
                     current_entry_prob = float(last_row['&-s-entry_signal_prob'])
                 
+                # 获取当前出场信号概率
+                current_exit_prob = None
+                if '&s-exit_signal' in df.columns:
+                    current_exit_prob = float(last_row['&s-exit_signal'])
+                elif '&-exit_signal_prob' in df.columns:
+                    current_exit_prob = float(last_row['&-exit_signal_prob'])
+                elif '&-s-exit_signal_prob' in df.columns:
+                    current_exit_prob = float(last_row['&-s-exit_signal_prob'])
+                
                 # 获取波动率预测
                 predicted_volatility = None
                 if '&-target_volatility' in df.columns:
@@ -1683,11 +1727,21 @@ class FreqaiPrimer(IStrategy):
                 if '&-target_trend' in df.columns:
                     trend_prediction = float(last_row['&-target_trend'])
                 
-                # 基于ML预测的智能决策（移除了对不可靠的current_profit的依赖）
+                # 应用新规则：当exit_signal > entry_signal × 1.33时退出
+                signal_ratio_threshold = 1.33
+                if current_entry_prob is not None and current_exit_prob is not None:
+                    # 确保概率在 [0, 1] 范围内
+                    current_entry_prob = max(0.0, min(1.0, current_entry_prob))
+                    current_exit_prob = max(0.0, min(1.0, current_exit_prob))
+                    
+                    if current_exit_prob > current_entry_prob * signal_ratio_threshold:
+                        self.strategy_log(f"[{pair}] exit_signal ({current_exit_prob:.2f}) 超过 entry_signal ({current_entry_prob:.2f}) 的 {signal_ratio_threshold:.2f} 倍，支持出场")
+                        return True
+                
+                # 原有ML预测逻辑作为补充
                 if current_entry_prob is not None and predicted_volatility is not None:
                     # 如果当前入场信号概率很高，说明市场条件仍然良好，可以继续持有
                     if current_entry_prob > 0.6:
-                        # 高入场概率，建议继续持有
                         self.strategy_log(f"[{pair}] 当前入场信号概率高({current_entry_prob:.2f})，建议继续持有")
                         return False  # 阻止出场