趋势计算方式改变成三段式

freqtrade/templates/freqaiprimer.py

@@ -39,6 +39,14 @@ class FreqaiPrimer(IStrategy):
     TREND_BULLISH_THRESHOLD = 85
     TREND_BEARISH_THRESHOLD = 60
     
+    # 新的加权趋势判定阈值（用于hyperopt优化）
+    TREND_FINAL_BULLISH_THRESHOLD = 70  # 上涨趋势最终阈值
+    TREND_FINAL_BEARISH_THRESHOLD = 30  # 下跌趋势最终阈值
+    
+    # Hyperopt 可优化参数
+    trend_final_bullish_threshold = IntParameter(50, 90, default=87, space="buy", optimize=True, load=True)
+    trend_final_bearish_threshold = IntParameter(10, 50, default=53, space="buy", optimize=True, load=True)
+
     # --- 🛠️ 固定配置参数 ---
     stoploss = -0.15
     timeframe = "3m"
@@ -1022,11 +1030,11 @@ class FreqaiPrimer(IStrategy):
 
     def detect_trend_status(self, dataframe: DataFrame, metadata: dict) -> str:
         """
-        基于最近20个周期的trend_score判断趋势状态
+        基于加权分段的trend_score判断趋势状态
         规则：
-        - 最近20个周期中75%以上得分≥TREND_BULLISH_THRESHOLD，且最近10个周期中90%以上得分≥TREND_BULLISH_THRESHOLD → 上涨趋势
+        - 将最近20个trend_score分为3段：1-3(权重10)，4-10(权重7)，11-20(权重3)
-        - 最近20个周期中75%以上得分≤TREND_BEARISH_THRESHOLD，且最近10个周期中90%以上得分≤TREND_BEARISH_THRESHOLD → 下跌趋势
+        - 计算加权平均得分，映射到0-100区间
-        - 其他情况 → 震荡趋势
+        - 根据两个可优化的阈值判断趋势状态
         """
         pair = metadata.get('pair', 'Unknown')
         
@@ -1047,26 +1055,45 @@ class FreqaiPrimer(IStrategy):
                 else:
                     trend_scores_20.append(50)  # 默认值
             
-            # 最近10个周期
+            # 分段计算加权得分
-            trend_scores_10 = trend_scores_20[-10:]
+            # 第一段：最近1-3个周期 (索引-3到-1)
+            segment1 = trend_scores_20[-3:]
+            weighted_score1 = sum(score * 10 for score in segment1) / len(segment1)
             
-            # 计算满足条件的比例
+            # 第二段：4-10个周期 (索引-10到-4)
-            strong_bull_20 = sum(1 for score in trend_scores_20 if score >= self.TREND_BULLISH_THRESHOLD) / len(trend_scores_20)
+            segment2 = trend_scores_20[-10:-3]
-            strong_bull_10 = sum(1 for score in trend_scores_10 if score >= self.TREND_BULLISH_THRESHOLD) / len(trend_scores_10)
+            weighted_score2 = sum(score * 7 for score in segment2) / len(segment2)
             
-            strong_bear_20 = sum(1 for score in trend_scores_20 if score <= self.TREND_BEARISH_THRESHOLD) / len(trend_scores_20)
+            # 第三段：11-20个周期 (索引-20到-11)
-            strong_bear_10 = sum(1 for score in trend_scores_10 if score <= self.TREND_BEARISH_THRESHOLD) / len(trend_scores_10)
+            segment3 = trend_scores_20[-20:-10]
+            weighted_score3 = sum(score * 3 for score in segment3) / len(segment3)
             
-            # 判定趋势
+            # 计算最终加权得分
-            if strong_bull_20 >= 0.75 and strong_bull_10 >= 0.9:
+            final_weighted_score = (weighted_score1 + weighted_score2 + weighted_score3) / (10 + 7 + 3)
+            
+            # 将得分映射到0-100区间（确保在合理范围内）
+            final_score = max(0, min(100, final_weighted_score))
+            
+            # 使用hyperopt优化的阈值判断趋势状态
+            bullish_threshold = self.trend_final_bullish_threshold.value
+            bearish_threshold = self.trend_final_bearish_threshold.value
+            
+            # 判定趋势状态
+            if final_score >= bullish_threshold:
                 trend_status = "bullish"
-                logger.info(f"[{pair}] 🚀 检测到上涨趋势: 20周期强势比例={strong_bull_20:.1%}, 10周期强势比例={strong_bull_10:.1%}, 阈值={self.TREND_BULLISH_THRESHOLD}")
+                logger.info(f"[{pair}] 🚀 检测到上涨趋势: 最终加权得分={final_score:.2f}, 阈值≥{bullish_threshold}")
-            elif strong_bear_20 >= 0.75 and strong_bear_10 >= 0.9:
+            elif final_score <= bearish_threshold:
                 trend_status = "bearish"
-                logger.info(f"[{pair}] 📉 检测到下跌趋势: 20周期弱势比例={strong_bear_20:.1%}, 10周期弱势比例={strong_bear_10:.1%}, 阈值={self.TREND_BEARISH_THRESHOLD}")
+                logger.info(f"[{pair}] 📉 检测到下跌趋势: 最终加权得分={final_score:.2f}, 阈值≤{bearish_threshold}")
             else:
                 trend_status = "ranging"
-                logger.info(f"[{pair}] ⚖️ 检测到震荡趋势: 20周期强势比例={strong_bull_20:.1%}, 弱势比例={strong_bear_20:.1%}, 阈值={self.TREND_BULLISH_THRESHOLD}/{self.TREND_BEARISH_THRESHOLD}")
+                logger.info(f"[{pair}] ⚖️ 检测到震荡趋势: 最终加权得分={final_score:.2f}, 阈值范围({bearish_threshold}, {bullish_threshold})")
+            
+            # 输出分段详细信息用于调试
+            logger.debug(f"[{pair}] 趋势分析详情 - "
+                        f"第一段(1-3,权重10): {[f'{s:.1f}' for s in segment1]}, "
+                        f"第二段(4-10,权重7): {[f'{s:.1f}' for s in segment2]}, "
+                        f"第三段(11-20,权重3): {[f'{s:.1f}' for s in segment3]}")
             
             return trend_status
             

hyperopt_config_example.py

@@ -0,0 +1,106 @@
+#!/usr/bin/env python3
+"""
+Hyperopt 配置示例，用于优化新的趋势判定参数
+
+使用方法:
+1. 将此文件保存为 hyperopt_trend_params.py
+2. 运行: freqtrade hyperopt --config config.json --strategy freqaiprimer --hyperopt hyperopt_trend_params
+"""
+
+from freqtrade.optimize.hyperopt_interface import IHyperOpt
+from typing import Dict, Any
+import logging
+
+logger = logging.getLogger(__name__)
+
+class TrendParamsHyperopt(IHyperOpt):
+    """
+    优化趋势判定参数的 Hyperopt 类
+    """
+    
+    @staticmethod
+    def generate_roi_table(params: Dict) -> Dict[int, float]:
+        """生成ROI表格"""
+        return {
+            0: params['roi_p1'] + params['roi_p2'] + params['roi_p3'],
+            params['roi_t3']: params['roi_p1'] + params['roi_p2'],
+            params['roi_t3'] + params['roi_t2']: params['roi_p1'],
+            params['roi_t3'] + params['roi_t2'] + params['roi_t1']: 0,
+        }
+
+    @staticmethod
+    def roi_space():
+        """ROI参数空间"""
+        return {
+            'roi_p1': [0.01, 0.08],
+            'roi_p2': [0.01, 0.08],
+            'roi_p3': [0.01, 0.08],
+            'roi_t1': [10, 120],
+            'roi_t2': [10, 60],
+            'roi_t3': [10, 60],
+        }
+
+    @staticmethod
+    def stoploss_space():
+        """止损参数空间"""
+        return {
+            'stoploss': [-0.5, -0.05],
+        }
+
+    @staticmethod
+    def trailing_space():
+        """追踪止损参数空间"""
+        return {
+            'trailing_stop': [True, False],
+            'trailing_stop_positive': [0.005, 0.05],
+            'trailing_stop_positive_offset_p1': [0.005, 0.05],
+            'trailing_only_offset_is_reached': [True, False],
+        }
+
+    def buy_params_space(self):
+        """买入参数空间，包含新的趋势判定阈值"""
+        return {
+            # 原有参数
+            'buy_threshold': [-0.05, -0.001],
+            'volume_z_score_min': [0.1, 2.0],
+            'rsi_max': [20, 60],
+            
+            # 新的趋势判定阈值
+            'trend_final_bullish_threshold': [50, 90],  # 上涨趋势阈值
+            'trend_final_bearish_threshold': [10, 50],   # 下跌趋势阈值
+        }
+
+    def sell_params_space(self):
+        """卖出参数空间"""
+        return {
+            'sell_threshold': [0.001, 0.08],
+            'rsi_min': [60, 90],
+            'stochrsi_max': [60, 95],
+        }
+
+    def generate_estimator(self, dimensions: list, **kwargs) -> Any:
+        """生成优化器"""
+        from skopt import Optimizer
+        from skopt.space import Real, Integer
+        
+        # 定义参数空间
+        space = [
+            Real(-0.5, -0.05, name='stoploss'),
+            Real(0.001, 0.08, name='sell_threshold'),
+            Real(-0.05, -0.001, name='buy_threshold'),
+            Real(0.1, 2.0, name='volume_z_score_min'),
+            Integer(20, 60, name='rsi_max'),
+            Integer(60, 90, name='rsi_min'),
+            Integer(60, 95, name='stochrsi_max'),
+            Integer(50, 90, name='trend_final_bullish_threshold'),
+            Integer(10, 50, name='trend_final_bearish_threshold'),
+        ]
+        
+        return Optimizer(space, base_estimator='ET', acq_func='EI', n_initial_points=10)
+
+# 使用示例配置
+if __name__ == "__main__":
+    print("趋势参数优化配置已加载")
+    print("支持的优化参数:")
+    print("- trend_final_bullish_threshold: 50-90")
+    print("- trend_final_bearish_threshold: 10-50")