Alvinchen_v35

"""
ADX/DI Strategy without FreqAI - v34.20 (Multi-Timeframe Analysis)
===============================================================

基于 v34_19 的 sentiment filter 优化 + 多时间框架分析。

v34.20 关键改进:
1. 使用 @informative 装饰器实现多时间框架:
   - 1h: EMA20, EMA50, ADX 趋势确认
   - 4h: EMA50, ADX 方向过滤
2. balance_check_threshold: 0.75
3. sentiment_allow_contrarian_shorts: True

MTF逻辑:
- 做多: 15m信号 + 1h EMA20>EMA50 + 4h ADX>25
- 做空: 15m信号 + 1h EMA20<EMA50 + 4h ADX>25

Hyperopt 优化参数:
- adx_threshold: 19.788
- volume_mult: 1.221
"""

import logging
import numpy as np
import pandas as pd
from pandas import DataFrame
from typing import Optional, Dict, Any, Union
from collections import OrderedDict
from datetime import datetime, timedelta
from freqtrade.strategy import IStrategy, DecimalParameter, IntParameter, informative
from freqtrade.persistence import Trade
from freqtrade.enums import RunMode
import talib.abstract as ta
import freqtrade.vendor.qtpylib.indicators as qtpylib
from functools import reduce

# 情绪分析和链上数据模块
try:
    from free_sentiment_api import (
        get_sentiment_for_trading,
        fetch_fear_greed_index,
        fetch_reddit_sentiment,
        fetch_elfa_token_news
    )
    FREE_SENTIMENT_AVAILABLE = True
except ImportError:
    FREE_SENTIMENT_AVAILABLE = False

try:
    from free_onchain_data import (
        get_funding_rate,
        get_long_short_ratio,
        get_market_sentiment
    )
    FREE_ONCHAIN_AVAILABLE = True
except ImportError:
    FREE_ONCHAIN_AVAILABLE = False

# 订单簿分析模块
try:
    from orderbook_analyzer import (
        analyze_orderbook,
        get_obi,
        OrderBookAnalyzer
    )
    ORDERBOOK_AVAILABLE = True
except ImportError:
    ORDERBOOK_AVAILABLE = False

logger = logging.getLogger(__name__)


class Alvinchen_v35(IStrategy):
    INTERFACE_VERSION = 3
    """
    ADX/DI Strategy without FreqAI - v35 (Multi-Timeframe)

    Key features:
    - No FreqAI (pure technical analysis)
    - Optimized long entry conditions (ADX>55, DI_diff>20)
    - Multi-Timeframe confirmation: 1h + 15m
    """

    INTERFACE_VERSION = 3
    can_short = True
    use_exit_signal = True  # 启用退出信号
    exit_profit_only = False
    exit_profit_offset = 0.0  # 退出信号的最小利润阈值

    timeframe = '15m'
    startup_candle_count = 1000  # 需要足够多以计算1d EMA200

    @property
    def informative_pairs(self):
        """指定需要获取多时间框架数据的交易对"""
        # 注意: 不再这里定义1d，因为某些交易对可能没有1d数据
        # 改为在populate_indicators_1d中处理
        whitelist = self.config.get('exchange', {}).get('pair_whitelist', [])
        return [(pair, '1h') for pair in whitelist]

    # ==================== FIXED OPTIMIZED PARAMETERS (Hyperopt Epoch 40) ====================
    # Entry parameters - hyperopt 优化后的固定值
    adx_threshold = 19.788      # ADX入场阈值
    volume_mult = 1.221         # 成交量倍数

    # Exit parameters
    exit_rsi_overbought = 75    # RSI超买阈值
    exit_rsi_oversold = 35      # RSI超卖阈值

    # Stoploss parameters - 宽松初始止损，动态收紧
    long_stoploss = 0.50        # 多单初始止损 50%
    short_stoploss = 0.50       # 空单初始止损 50%

    # ==================== 币种分级配置 ====================
    # 主流币种: 流动性好、波动相对平稳、订单簿厚
    major_coins = ['BTC/USDT', 'ETH/USDT', 'BNB/USDT', 'XRP/USDT', 'SOL/USDT']
    # 二流币种: 流动性较差、波动大、容易被操纵
    minor_coins = ['DOGE/USDT', 'ADA/USDT', 'NEAR/USDT', 'LINK/USDT', 'AVAX/USDT',
                   'DOT/USDT', 'LTC/USDT', 'ATOM/USDT', 'ARB/USDT', 'OP/USDT']

    # 币种分级指标阈值
    # 主流币种可以使用更严格的指标
    major_adx_threshold = 30      # 主流币 ADX 阈值
    major_di_threshold = 5       # 主流币 DI_diff 阈值
    major_volume_ratio = 1.2     # 主流币成交量倍数
    # 二流币种需要更明确的趋势才入场
    minor_adx_threshold = 35      # 二流币 ADX 阈值 (更严格)
    minor_di_threshold = 8       # 二流币 DI_diff 阈值 (更严格)
    minor_volume_ratio = 1.5     # 二流币成交量倍数 (更高要求)

    # Profit lock parameters
    profit_lock_threshold = 0.052   # 利润锁定阈值 5.2%
    profit_lock_ratio = 0.761       # 利润锁定比例 76.1%

    # ==================== FIXED PARAMETERS ====================
    enable_sentiment_check = True          # 启用情绪检查
    enable_onchain_filter = True           # 启用链上数据过滤
    enable_news_filter = True              # 启用新闻过滤
    enable_orderbook_analysis = True       # 启用订单簿分析
    sentiment_bullish_threshold = 0.65    # 看涨情绪阈值
    sentiment_bearish_threshold = 0.35    # 看跌情绪阈值
    sentiment_extreme_fear = 25           # 极度恐慌阈值
    sentiment_extreme_greed = 75          # 极度贪婪阈值
    strict_funding_filter = False          # 严格资金费率过滤
    strict_lsr_filter = False             # 严格多空比过滤
    sentiment_cache_ttl = 300             # 情绪缓存时间(秒)
    news_negative_threshold = 0.7         # 新闻负面情绪阈值

    # v34.19 新增: 情绪过滤器修复
    sentiment_allow_contrarian_shorts = True  # 允许在极度恐慌时做空（强趋势时）
    sentiment_contrarian_adx_threshold = 30.0  # 允许逆势做空的最低ADX阈值
    balance_check_threshold = 0.75        # 多空平衡检查阈值 (原为0.5，现改为0.75)
    news_check_cooldown = 3600            # 新闻检查冷却时间(秒)

    # 订单簿分析参数
    obi_threshold = 0.3                    # OBI阈值 (超过此值视为有效信号)
    obi_strong_threshold = 0.5             # 强OBI阈值
    spread_threshold = 0.1                 # 价差阈值 (%)
    large_order_mult = 3.0                 # 大单倍数阈值
    orderbook_cache_ttl = 5                # 订单簿缓存时间(秒)

    # ========== 动态杠杆配置 ==========
    # v18 优化: 禁用杠杆 - 测试表明无杠杆表现更好
    max_leverage = 1                       # 最大杠杆倍数
    min_leverage = 1                       # 最小杠杆倍数
    leverage_enabled = False               # 禁用杠杆

    # 杠杆置信度阈值
    leverage_threshold_medium = 0.70       # 中等置信度阈值
    leverage_threshold_high = 0.80         # 高置信度阈值 (2x杠杆)
    leverage_threshold_very_high = 0.90    # 极高置信度阈值 (3x杠杆)
    leverage_threshold_extreme = 0.95      # 极端置信度阈值

    # ========== V31: 放宽止损配置 ==========
    # 止损由 hyperopt 参数控制 (见上方 long_stoploss, short_stoploss)
    stoploss = -0.50  # 默认止损50%（用于多单）
    use_custom_stoploss = True  # 启用自定义止损

    # 移动止损参数
    trailing_stop = False  # 禁用内置移动止损
    trailing_stop_positive = 0.02
    trailing_stop_positive_offset = 0.03
    trailing_only_offset_is_reached = True

    # 利润锁定参数由 hyperopt 控制 (见上方 profit_lock_threshold, profit_lock_ratio)

    # ========== V31: 补仓策略配置 ==========
    position_adjustment_enable = False     # 禁用补仓
    max_entry_position_adjustment = -1     # 禁用补仓
    add_position_threshold = -0.05        # 亏损5%时触发补仓
    add_position_ratio = 0.5               # 补仓比例为初始仓位的50%

    # ========== 止盈(ROI)配置 ==========
    # 超快速退出 - 更激进锁定利润
    minimal_roi = {
        "0": 0.03,    # 3%利润立即退出
        "15": 0.02,   # 15分钟后2%退出
        "60": 0.015,  # 1小时后1.5%退出
        "240": 0.01,  # 4小时后1%退出
        "720": 0.005, # 12小时后0.5%退出
        "1440": 0     # 24小时后保本退出
    }

    # 回测模式标志 - 回测时跳过实时数据检查
    _backtest_mode = False
    _dry_run_mode = False

    def __init__(self, config: dict) -> None:
        super().__init__(config)
        self.freqai_enabled = config.get('freqai', {}).get('enabled', False)

        # 检测运行模式
        self._dry_run_mode = config.get('dry_run', False)
        # 回测模式检测: 当 dry_run=True 且没有实时交易时为回测
        # 在回测时 confirm_trade_entry 会被调用但无法访问实时API

        logger.info(f"Trend Following FreqAI Strategy - enabled: {self.freqai_enabled}")

        # 情绪数据缓存
        self._sentiment_cache = {}
        self._onchain_cache = {}
        self._news_cache = {}
        self._last_news_check = {}
        self._orderbook_cache = {}
        # 杠杆缓存 - 用于custom_stoploss
        self._leverage_cache = {}
        logger.info(f"情绪分析模块: {'可用' if FREE_SENTIMENT_AVAILABLE else '不可用'}")
        logger.info(f"链上数据模块: {'可用' if FREE_ONCHAIN_AVAILABLE else '不可用'}")
        logger.info(f"订单簿分析模块: {'可用' if ORDERBOOK_AVAILABLE else '不可用'}")

    def feature_engineering_expand_all(
        self, dataframe: DataFrame, period: int, metadata: dict, **kwargs
    ) -> DataFrame:
        """FreqAI features with high-accuracy indicators

        Research-backed indicators (win rates):
        - ROC: 93% win rate (best overall)
        - WMA: 83% win rate
        - HMA: 77% win rate (reduces lag)
        - MFI: volume-based momentum
        - Short RSI (2-6): better than RSI-14
        """
        # ========== Core Momentum Indicators ==========
        dataframe[f"%-rsi-period_{period}"] = ta.RSI(dataframe, timeperiod=period)
        # Short RSI for mean reversion (research shows 2-6 periods work best)
        dataframe[f"%-rsi_short-period_{period}"] = ta.RSI(dataframe, timeperiod=max(2, period//7))

        # ROC - Price Rate of Change (93% win rate indicator)
        dataframe[f"%-roc-period_{period}"] = ta.ROC(dataframe, timeperiod=period)

        # MOM - Momentum
        dataframe[f"%-mom-period_{period}"] = ta.MOM(dataframe, timeperiod=period)

        # ========== Trend Indicators ==========
        dataframe[f"%-ema-period_{period}"] = ta.EMA(dataframe['close'], timeperiod=period)

        # WMA - Weighted Moving Average (83% win rate)
        dataframe[f"%-wma-period_{period}"] = ta.WMA(dataframe['close'], timeperiod=period)

        # HMA - Hull Moving Average (77% win rate, reduces lag)
        hma_period = max(2, period)
        dataframe[f"%-hma-period_{period}"] = self._hma(dataframe['close'], hma_period)

        # SMA for comparison
        dataframe[f"%-sma-period_{period}"] = ta.SMA(dataframe['close'], timeperiod=period)

        # ========== Volatility Indicators ==========
        dataframe[f"%-atr-period_{period}"] = ta.ATR(dataframe, timeperiod=period)
        dataframe[f"%-atr_pct-period_{period}"] = dataframe[f"%-atr-period_{period}"] / dataframe['close']

        # ========== Volume Indicators ==========
        dataframe[f"%-volume_mean-period_{period}"] = dataframe['volume'].rolling(period).mean()
        dataframe[f"%-volume_ratio-period_{period}"] = dataframe['volume'] / (dataframe[f"%-volume_mean-period_{period}"] + 1e-8)

        # MFI - Money Flow Index (volume-weighted RSI)
        dataframe[f"%-mfi-period_{period}"] = ta.MFI(dataframe, timeperiod=period)

        # ========== Trend Strength ==========
        dataframe[f"%-adx-period_{period}"] = ta.ADX(dataframe, timeperiod=period)
        dataframe[f"%-plus_di-period_{period}"] = ta.PLUS_DI(dataframe, timeperiod=period)
        dataframe[f"%-minus_di-period_{period}"] = ta.MINUS_DI(dataframe, timeperiod=period)

        # ========== Price Action ==========
        dataframe[f"%-returns-period_{period}"] = dataframe['close'].pct_change(period)

        # ========== Composite Indicators (Combining multiple for accuracy) ==========
        # ROC-MOM combination (momentum confirmation)
        dataframe[f"%-roc_mom_combo-period_{period}"] = (
            (dataframe[f"%-roc-period_{period}"] + dataframe[f"%-mom-period_{period}"]) / 2
        )

        # HMA-EMA spread (trend with reduced lag)
        dataframe[f"%-hma_ema_spread-period_{period}"] = (
            dataframe[f"%-hma-period_{period}"] - dataframe[f"%-ema-period_{period}"]
        ) / dataframe['close']

        # WMA-SMA spread (weighted vs simple comparison)
        dataframe[f"%-wma_sma_spread-period_{period}"] = (
            dataframe[f"%-wma-period_{period}"] - dataframe[f"%-sma-period_{period}"]
        ) / dataframe['close']

        # RSI-MFI combination (price + volume momentum)
        dataframe[f"%-rsi_mfi_avg-period_{period}"] = (
            dataframe[f"%-rsi-period_{period}"] + dataframe[f"%-mfi-period_{period}"]
        ) / 2

        # Volume-weighted price momentum
        dataframe[f"%-vwap_proxy-period_{period}"] = (
            (dataframe['close'] * dataframe['volume']).rolling(period).sum() /
            (dataframe['volume'].rolling(period).sum() + 1e-8)
        )

        # Price position relative to HMA
        dataframe[f"%-price_hma_ratio-period_{period}"] = (
            dataframe['close'] / dataframe[f"%-hma-period_{period}"] - 1
        )

        # ATR-normalized momentum
        dataframe[f"%-atr_normalized_mom-period_{period}"] = (
            dataframe[f"%-mom-period_{period}"] / (dataframe[f"%-atr-period_{period}"] + 1e-8)
        )

        return dataframe

    def _hma(self, series: pd.Series, period: int) -> pd.Series:
        """Calculate Hull Moving Average (77% win rate indicator)

        HMA formula: WMA(2*WMA(n/2) - WMA(n), sqrt(n))
        Reduces lag while maintaining smoothness
        """
        half_period = int(period / 2)
        sqrt_period = int(np.sqrt(period))

        wma_half = ta.WMA(series, timeperiod=half_period)
        wma_full = ta.WMA(series, timeperiod=period)

        raw_hma = 2 * wma_half - wma_full
        hma = ta.WMA(raw_hma, timeperiod=sqrt_period)

        return hma

    def _ttm_squeeze(self, dataframe: DataFrame, period: int = 20, mult: float = 1.5) -> pd.Series:
        """
        TTM Squeeze - 波动率压缩检测 (70%+胜率)

        原理: Bollinger Bands在Keltner Channels内 = 波动压缩 = 即将突破
        当squeeze结束(波动扩张)时入场
        """
        # Bollinger Bands
        bb_basis = dataframe['close'].rolling(period).mean()
        bb_std = dataframe['close'].rolling(period).std()
        bb_upper = bb_basis + 2 * bb_std
        bb_lower = bb_basis - 2 * bb_std

        # Keltner Channels
        kc_basis = dataframe['close'].ewm(span=period).mean()
        atr = ta.ATR(dataframe, timeperiod=period)
        kc_upper = kc_basis + mult * atr
        kc_lower = kc_basis - mult * atr

        # Squeeze: BB在KC内 = 波动压缩
        squeeze = (bb_lower > kc_lower) & (bb_upper < kc_upper)
        return squeeze

    def _wave_trend(self, dataframe: DataFrame, channel_len: int = 10, avg_len: int = 21) -> tuple:
        """
        Wave Trend Oscillator (65%+胜率)

        结合RSI和移动平均的超买超卖指标
        wt1上穿wt2且<-60 = 买入信号
        wt1下穿wt2且>60 = 卖出信号
        """
        hlc3 = (dataframe['high'] + dataframe['low'] + dataframe['close']) / 3
        esa = ta.EMA(hlc3, timeperiod=channel_len)
        d = ta.EMA(abs(hlc3 - esa), timeperiod=channel_len)
        ci = (hlc3 - esa) / (0.015 * d + 1e-8)
        wt1 = ta.EMA(ci, timeperiod=avg_len)
        wt2 = ta.SMA(wt1, timeperiod=4)
        return wt1, wt2

    def _supertrend(self, dataframe: DataFrame, period: int = 10, multiplier: float = 3.0) -> tuple:
        """
        Supertrend - 超级趋势指标 (50-60%胜率)

        基于ATR的趋势线，简单有效的趋势跟随
        direction: 1 = 多头, -1 = 空头
        """
        atr = ta.ATR(dataframe, timeperiod=period)
        hl2 = (dataframe['high'] + dataframe['low']) / 2

        upper_band = hl2 + multiplier * atr
        lower_band = hl2 - multiplier * atr

        # Initialize
        supertrend = pd.Series(0.0, index=dataframe.index)
        direction = pd.Series(1, index=dataframe.index)

        for i in range(1, len(dataframe)):
            if dataframe['close'].iloc[i] > upper_band.iloc[i-1]:
                direction.iloc[i] = 1
            elif dataframe['close'].iloc[i] < lower_band.iloc[i-1]:
                direction.iloc[i] = -1
            else:
                direction.iloc[i] = direction.iloc[i-1]

            if direction.iloc[i] == 1:
                supertrend.iloc[i] = lower_band.iloc[i]
            else:
                supertrend.iloc[i] = upper_band.iloc[i]

        return supertrend, direction

    def _ichimoku(self, dataframe: DataFrame) -> tuple:
        """
        Ichimoku Cloud - 一目均衡表 (55-75%胜率)

        多维度趋势、支撑阻力、动量分析
        - 价格在云上方 = 多头趋势
        - TK金叉 = 入场信号
        """
        tenkan = (dataframe['high'].rolling(9).max() + dataframe['low'].rolling(9).min()) / 2
        kijun = (dataframe['high'].rolling(26).max() + dataframe['low'].rolling(26).min()) / 2
        senkou_a = ((tenkan + kijun) / 2).shift(26)
        senkou_b = ((dataframe['high'].rolling(52).max() + dataframe['low'].rolling(52).min()) / 2).shift(26)
        chikou = dataframe['close'].shift(-26)

        return tenkan, kijun, senkou_a, senkou_b, chikou

    def _detect_order_blocks(self, dataframe: DataFrame) -> tuple:
        """
        Order Block Detection - 订单块检测 (60%+胜率)

        SMC概念: 机构买卖区域
        - 看涨OB: 下跌趋势中最后一根上涨K线
        - 看跌OB: 上涨趋势中最后一根下跌K线
        """
        # Bullish OB: 最后一根上涨K线后开始下跌
        bullish_ob = (
            (dataframe['close'] > dataframe['open']) &  # 当前是上涨K线
            (dataframe['close'].shift(-1) < dataframe['open'].shift(-1))  # 下一根开始下跌
        )

        # Bearish OB: 最后一根下跌K线后开始上涨
        bearish_ob = (
            (dataframe['close'] < dataframe['open']) &  # 当前是下跌K线
            (dataframe['close'].shift(-1) > dataframe['open'].shift(-1))  # 下一根开始上涨
        )

        return bullish_ob, bearish_ob

    def feature_engineering_expand_basic(
        self, dataframe: DataFrame, metadata: dict, **kwargs
    ) -> DataFrame:
        """Basic features with high-accuracy indicators

        Research-backed additions:
        - Short RSI (2-6): Better for mean reversion
        - CCI: 50% win rate, good for cyclical trends
        - Aroon: 47% win rate, trend strength detection
        - Stochastic: 43% win rate, oscillator
        - VWAP proxy: 93% win rate
        """
        # ========== Time Features ==========
        dataframe["%-hour"] = dataframe["date"].dt.hour
        dataframe["%-day_of_week"] = dataframe["date"].dt.dayofweek

        # ========== RSI Suite (Short periods work better) ==========
        dataframe["%-rsi"] = ta.RSI(dataframe, timeperiod=14)
        dataframe["%-rsi_fast"] = ta.RSI(dataframe, timeperiod=7)
        dataframe["%-rsi_2"] = ta.RSI(dataframe, timeperiod=2)  # Very short, 91% win rate
        dataframe["%-rsi_3"] = ta.RSI(dataframe, timeperiod=3)
        dataframe["%-rsi_5"] = ta.RSI(dataframe, timeperiod=5)

        # ========== ROC Suite (93% win rate indicator) ==========
        dataframe["%-roc"] = ta.ROC(dataframe, timeperiod=10)
        dataframe["%-roc_fast"] = ta.ROC(dataframe, timeperiod=5)
        dataframe["%-roc_2"] = ta.ROC(dataframe, timeperiod=2)

        # ========== MACD ==========
        macd_result = ta.MACD(dataframe)
        if isinstance(macd_result, dict):
            dataframe["%-macd"] = macd_result['macd']
            dataframe["%-macdsignal"] = macd_result['macdsignal']
            dataframe["%-macdhist"] = macd_result['macdhist']
        else:
            macd, macdsignal, macdhist = ta.MACD(dataframe['close'])
            dataframe["%-macd"] = macd
            dataframe["%-macdsignal"] = macdsignal
            dataframe["%-macdhist"] = macdhist

        # ========== Bollinger Bands ==========
        bb = qtpylib.bollinger_bands(dataframe['close'], window=20, stds=2)
        dataframe["%-bb_lower"] = bb['lower']
        dataframe["%-bb_upper"] = bb['upper']
        dataframe["%-bb_pct"] = (dataframe['close'] - bb['lower']) / (bb['upper'] - bb['lower'] + 1e-8)
        dataframe["%-bb_width"] = (bb['upper'] - bb['lower']) / (bb['mid'] + 1e-8)

        # ========== Moving Averages ==========
        # EMAs
        dataframe["%-ema_20"] = ta.EMA(dataframe['close'], timeperiod=20)
        dataframe["%-ema_50"] = ta.EMA(dataframe['close'], timeperiod=50)
        dataframe["%-ema_100"] = ta.EMA(dataframe['close'], timeperiod=100)
        dataframe["%-ema_200"] = ta.EMA(dataframe['close'], timeperiod=200)

        # WMAs (83% win rate)
        dataframe["%-wma_20"] = ta.WMA(dataframe['close'], timeperiod=20)
        dataframe["%-wma_50"] = ta.WMA(dataframe['close'], timeperiod=50)

        # HMAs (77% win rate, reduces lag)
        dataframe["%-hma_20"] = self._hma(dataframe['close'], 20)
        dataframe["%-hma_50"] = self._hma(dataframe['close'], 50)

        # ========== ADX ==========
        dataframe["%-adx"] = ta.ADX(dataframe, timeperiod=14)
        dataframe["%-plus_di"] = ta.PLUS_DI(dataframe, timeperiod=14)
        dataframe["%-minus_di"] = ta.MINUS_DI(dataframe, timeperiod=14)
        dataframe["%-di_diff"] = dataframe["%-plus_di"] - dataframe["%-minus_di"]

        # ========== ATR ==========
        dataframe["%-atr"] = ta.ATR(dataframe, timeperiod=14)
        dataframe["%-atr_pct"] = dataframe["%-atr"] / dataframe['close']

        # ========== Volume Indicators ==========
        dataframe["%-volume_mean_20"] = dataframe['volume'].rolling(20).mean()
        dataframe["%-volume_ratio"] = dataframe['volume'] / (dataframe["%-volume_mean_20"] + 1e-8)

        # MFI - Money Flow Index (volume-weighted RSI)
        dataframe["%-mfi"] = ta.MFI(dataframe, timeperiod=14)

        # VWAP proxy (93% win rate) - cumulative volume-weighted price
        dataframe["%-vwap_proxy"] = (
            (dataframe['close'] * dataframe['volume']).cumsum() /
            (dataframe['volume'].cumsum() + 1e-8)
        )
        dataframe["%-vwap_deviation"] = dataframe['close'] / dataframe["%-vwap_proxy"] - 1

        # ========== CCI - Commodity Channel Index (50% win rate) ==========
        dataframe["%-cci"] = ta.CCI(dataframe, timeperiod=20)
        dataframe["%-cci_fast"] = ta.CCI(dataframe, timeperiod=10)

        # ========== Stochastic (43% win rate) ==========
        stoch = ta.STOCH(dataframe, fastk_period=14, slowk_period=3, slowd_period=3)
        dataframe["%-stoch_k"] = stoch['slowk']
        dataframe["%-stoch_d"] = stoch['slowd']

        # ========== Momentum ==========
        dataframe["%-momentum_5"] = dataframe['close'] / dataframe['close'].shift(5) - 1
        dataframe["%-momentum_10"] = dataframe['close'] / dataframe['close'].shift(10) - 1

        # ========== Trend Detection ==========
        dataframe["%-uptrend"] = np.where(
            (dataframe["%-ema_20"] > dataframe["%-ema_50"]) &
            (dataframe["%-di_diff"] > 0),
            1, 0
        )
        dataframe["%-downtrend"] = np.where(
            (dataframe["%-ema_20"] < dataframe["%-ema_50"]) &
            (dataframe["%-di_diff"] < 0),
            1, 0
        )

        # ========== Composite Indicators (High Accuracy Combinations) ==========
        # RSI-MFI combination (price + volume momentum consensus)
        dataframe["%-rsi_mfi_consensus"] = np.where(
            (dataframe["%-rsi_2"] < 30) & (dataframe["%-mfi"] < 30), -1,  # Oversold consensus
            np.where(
                (dataframe["%-rsi_2"] > 70) & (dataframe["%-mfi"] > 70), 1,  # Overbought consensus
                0
            )
        )

        # ROC-Momentum combo (93% win rate indicator combined)
        dataframe["%-roc_momentum"] = (
            dataframe["%-roc_2"] + dataframe["%-momentum_5"] * 100
        ) / 2

        # HMA-EMA crossover signal (reduced lag trend)
        dataframe["%-hma_ema_signal"] = np.where(
            dataframe["%-hma_20"] > dataframe["%-ema_20"], 1,
            np.where(dataframe["%-hma_20"] < dataframe["%-ema_20"], -1, 0)
        )

        # Price position relative to VWAP (institutional buying/selling)
        dataframe["%-vwap_position"] = np.where(
            dataframe['close'] > dataframe["%-vwap_proxy"], 1,  # Above VWAP - bullish
            np.where(dataframe['close'] < dataframe["%-vwap_proxy"], -1, 0)  # Below VWAP - bearish
        )

        # Bollinger Band position with RSI (mean reversion signal)
        dataframe["%-bb_rsi_signal"] = np.where(
            (dataframe["%-bb_pct"] < 0.2) & (dataframe["%-rsi_2"] < 30), 1,  # Oversold + low RSI = buy
            np.where(
                (dataframe["%-bb_pct"] > 0.8) & (dataframe["%-rsi_2"] > 70), -1,  # Overbought + high RSI = sell
                0
            )
        )

        # Multi-MA alignment (strong trend signal)
        dataframe["%-ma_alignment"] = np.where(
            (dataframe["%-ema_20"] > dataframe["%-ema_50"]) &
            (dataframe["%-ema_50"] > dataframe["%-ema_100"]) &
            (dataframe["%-ema_100"] > dataframe["%-ema_200"]),
            1,  # Full bullish alignment
            np.where(
                (dataframe["%-ema_20"] < dataframe["%-ema_50"]) &
                (dataframe["%-ema_50"] < dataframe["%-ema_100"]) &
                (dataframe["%-ema_100"] < dataframe["%-ema_200"]),
                -1,  # Full bearish alignment
                0
            )
        )

        # ========== NEW: TTM Squeeze (波动率压缩) ==========
        dataframe["ttm_squeeze"] = self._ttm_squeeze(dataframe)
        dataframe["%-ttm_squeeze"] = dataframe["ttm_squeeze"]  # FreqAI特征

        # ========== NEW: Wave Trend Oscillator ==========
        wt1, wt2 = self._wave_trend(dataframe)
        dataframe["wt1"] = wt1
        dataframe["wt2"] = wt2
        dataframe["%-wt1"] = wt1  # FreqAI特征
        dataframe["%-wt2"] = wt2
        dataframe["wt_signal"] = np.where(
            (wt1 > wt2) & (wt1 < -60), 1,  # 从超卖区域回升
            np.where((wt1 < wt2) & (wt1 > 60), -1, 0)  # 从超买区域回落
        )
        dataframe["%-wt_signal"] = dataframe["wt_signal"]

        # ========== NEW: Supertrend ==========
        supertrend, st_dir = self._supertrend(dataframe)
        dataframe["supertrend"] = supertrend
        dataframe["st_direction"] = st_dir
        dataframe["%-supertrend"] = supertrend  # FreqAI特征
        dataframe["%-st_direction"] = st_dir

        # ========== NEW: Ichimoku Cloud ==========
        tenkan, kijun, senkou_a, senkou_b, chikou = self._ichimoku(dataframe)
        dataframe["tenkan"] = tenkan
        dataframe["kijun"] = kijun
        dataframe["senkou_a"] = senkou_a
        dataframe["senkou_b"] = senkou_b
        dataframe["%-tenkan"] = tenkan  # FreqAI特征
        dataframe["%-kijun"] = kijun
        dataframe["%-senkou_a"] = senkou_a
        dataframe["%-senkou_b"] = senkou_b
        dataframe["tk_cross"] = np.where(tenkan > kijun, 1, -1)  # TK金叉
        dataframe["above_cloud"] = np.where(
            dataframe['close'] > senkou_a, 1,
            np.where(dataframe['close'] < senkou_b, -1, 0)
        )
        dataframe["%-tk_cross"] = dataframe["tk_cross"]
        dataframe["%-above_cloud"] = dataframe["above_cloud"]

        # ========== NEW: Order Blocks ==========
        bullish_ob, bearish_ob = self._detect_order_blocks(dataframe)
        dataframe["bullish_ob"] = bullish_ob.astype(int)
        dataframe["bearish_ob"] = bearish_ob.astype(int)
        dataframe["%-bullish_ob"] = dataframe["bullish_ob"]
        dataframe["%-bearish_ob"] = dataframe["bearish_ob"]

        return dataframe

    def feature_engineering_standard(
        self, dataframe: DataFrame, metadata: dict, **kwargs
    ) -> DataFrame:
        """Standard features"""
        return dataframe

    def set_freqai_targets(self, dataframe: DataFrame, metadata: dict, **kwargs) -> DataFrame:
        """Prediction targets - v34_4: 单一回归目标"""
        label_period = self.config.get('freqai', {}).get('feature_parameters', {}) \
            .get('label_period_candles', 12)

        # 单一回归目标：未来N根K线收盘价滚动均值的变化率
        dataframe["&-s_close_mean"] = (
            dataframe["close"]
            .shift(-label_period)
            .rolling(label_period)
            .mean()
            / dataframe["close"]
            - 1
        )
        return dataframe

    # ========== Multi-Timeframe Indicators (v34.20) ==========
    # 只使用1h作为中期确认 (避免1d数据缺失问题)
    @informative('1h')
    def populate_indicators_1h(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
        """
        1h中期趋势确认指标

        MTF三层模型 (使用1h替代1d避免数据缺失):
        - Layer 1 (1d替代): 通过1h EMA200判断整体趋势 (使用更长的回溯周期)
        - Layer 2 (1h): 中期趋势确认
        - Layer 3 (15m): 精确入场点
        """
        # EMA排列判断趋势方向 (使用更长周期模拟日线效果)
        dataframe['ema_20'] = ta.EMA(dataframe['close'], timeperiod=20)
        dataframe['ema_50'] = ta.EMA(dataframe['close'], timeperiod=50)
        dataframe['ema_200'] = ta.EMA(dataframe['close'], timeperiod=200)

        # ADX趋势强度
        dataframe['adx'] = ta.ADX(dataframe, timeperiod=14)

        # MACD
        macd = ta.MACD(dataframe, fastperiod=12, slowperiod=26, signalperiod=9)
        dataframe['macd'] = macd['macd']
        dataframe['macdsignal'] = macd['macdsignal']

        # RSI
        dataframe['rsi'] = ta.RSI(dataframe['close'], timeperiod=14)

        return dataframe

    def populate_indicators(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
        """Indicators with high-accuracy research-backed indicators"""
        # ========== RSI Suite ==========
        dataframe['rsi'] = ta.RSI(dataframe['close'], timeperiod=14)
        dataframe['rsi_fast'] = ta.RSI(dataframe['close'], timeperiod=7)
        dataframe['rsi_2'] = ta.RSI(dataframe['close'], timeperiod=2)  # Short RSI for mean reversion

        # ========== ROC Suite (93% win rate) ==========
        dataframe['roc'] = ta.ROC(dataframe['close'], timeperiod=10)
        dataframe['roc_2'] = ta.ROC(dataframe['close'], timeperiod=2)

        # ========== MACD ==========
        macd, macdsignal, macdhist = ta.MACD(dataframe['close'])
        dataframe['macd'] = macd
        dataframe['macdsignal'] = macdsignal
        dataframe['macdhist'] = macdhist

        # ========== Bollinger Bands ==========
        bb = qtpylib.bollinger_bands(dataframe['close'], window=20, stds=2)
        dataframe['bb_lower'] = bb['lower']
        dataframe['bb_upper'] = bb['upper']
        dataframe['bb_pct'] = (dataframe['close'] - bb['lower']) / (bb['upper'] - bb['lower'] + 1e-8)
        dataframe['bb_width'] = (bb['upper'] - bb['lower']) / (bb['mid'] + 1e-8)

        # ========== Moving Averages ==========
        dataframe['ema_20'] = ta.EMA(dataframe['close'], timeperiod=20)
        dataframe['ema_50'] = ta.EMA(dataframe['close'], timeperiod=50)
        dataframe['ema_100'] = ta.EMA(dataframe['close'], timeperiod=100)
        dataframe['ema_200'] = ta.EMA(dataframe['close'], timeperiod=200)

        # WMA (83% win rate)
        dataframe['wma_20'] = ta.WMA(dataframe['close'], timeperiod=20)

        # HMA (77% win rate)
        dataframe['hma_20'] = self._hma(dataframe['close'], 20)

        # ========== ADX ==========
        dataframe['adx'] = ta.ADX(dataframe, timeperiod=14)
        dataframe['plus_di'] = ta.PLUS_DI(dataframe, timeperiod=14)
        dataframe['minus_di'] = ta.MINUS_DI(dataframe, timeperiod=14)
        dataframe['di_diff'] = dataframe['plus_di'] - dataframe['minus_di']

        # ========== ATR ==========
        dataframe['atr'] = ta.ATR(dataframe, timeperiod=14)
        dataframe['atr_pct'] = dataframe['atr'] / dataframe['close']

        # ========== Volume ==========
        dataframe['volume_mean_20'] = dataframe['volume'].rolling(20).mean()
        dataframe['volume_ratio'] = dataframe['volume'] / (dataframe['volume_mean_20'] + 1e-8)

        # MFI - Money Flow Index
        dataframe['mfi'] = ta.MFI(dataframe, timeperiod=14)

        # VWAP proxy
        dataframe['vwap_proxy'] = (
            (dataframe['close'] * dataframe['volume']).cumsum() /
            (dataframe['volume'].cumsum() + 1e-8)
        )

        # ========== Momentum ==========
        dataframe['momentum_5'] = dataframe['close'] / dataframe['close'].shift(5) - 1

        # ========== CCI (50% win rate) ==========
        dataframe['cci'] = ta.CCI(dataframe, timeperiod=20)

        # ========== Market Regime Detection ==========
        dataframe['market_regime'] = np.where(
            dataframe['ema_50'] > dataframe['ema_200'], 1,
            np.where(dataframe['ema_50'] < dataframe['ema_200'], -1, 0)
        )

        # ========== Composite Signals ==========
        # RSI-MFI oversold/overbought consensus
        dataframe['rsi_mfi_oversold'] = (dataframe['rsi_2'] < 30) & (dataframe['mfi'] < 30)
        dataframe['rsi_mfi_overbought'] = (dataframe['rsi_2'] > 70) & (dataframe['mfi'] > 70)

        # Price vs VWAP
        dataframe['above_vwap'] = dataframe['close'] > dataframe['vwap_proxy']

        # HMA-EMA crossover
        dataframe['hma_above_ema'] = dataframe['hma_20'] > dataframe['ema_20']

        # ========== TTM Squeeze (波动率压缩) ==========
        dataframe['ttm_squeeze'] = self._ttm_squeeze(dataframe)

        # ========== Wave Trend Oscillator ==========
        wt1, wt2 = self._wave_trend(dataframe)
        dataframe['wt1'] = wt1
        dataframe['wt2'] = wt2
        dataframe['wt_signal'] = np.where(
            (wt1 > wt2) & (wt1 < -60), 1,  # 从超卖区域回升
            np.where((wt1 < wt2) & (wt1 > 60), -1, 0)  # 从超买区域回落
        )

        # FreqAI处理
        if self.freqai_enabled:
            dataframe = self.freqai.start(dataframe, metadata, self)
        else:
            dataframe = self.set_freqai_targets(dataframe, metadata)

        return dataframe

    def populate_entry_trend(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
        """
        Entry conditions - v34.22: 启用多空双方向

        核心原则: "Better to miss than to make mistakes"
        - 只在高度确信时入场
        - 多指标确认才入场
        - 趋势方向必须明确

        MTF三层模型:
        - Layer 1 (1h大周期): EMA50 > EMA200 确认趋势方向
        - Layer 2 (1h中周期): EMA20 > EMA50 确认动量
        - Layer 3 (15m小周期): 多指标确认入场点

        币种分级:
        - 主流币种: 流动性好，可以使用较灵敏的指标
        - 二流币种: 流动性差，需要更明确的信号
        """
        # ========== 获取币种分级和对应阈值 ==========
        pair = metadata.get('pair', '')
        # 提取币种名称 (去掉 /USDT:USDT 后缀)
        coin_base = pair.split('/')[0] if '/' in pair else pair

        # 判断币种分级并选择阈值
        if coin_base in ['BTC', 'ETH', 'BNB', 'XRP', 'SOL']:
            # 主流币种
            is_major = True
            adx_threshold = 30
            di_threshold = 5
            volume_ratio_threshold = 1.2
        else:
            # 二流币种 - 需要更严格的指标
            is_major = False
            adx_threshold = 35
            di_threshold = 8
            volume_ratio_threshold = 1.5

        # ========== 检查1h数据是否存在 ==========
        has_1h_data = 'ema_200_1h' in dataframe.columns

        # ========== Layer 1: 1h大周期趋势判断 ==========
        if has_1h_data:
            # 大周期EMA多头排列 = 上升趋势
            major_bullish = dataframe['ema_50_1h'] > dataframe['ema_200_1h']
            major_bearish = dataframe['ema_50_1h'] < dataframe['ema_200_1h']
            # 大周期趋势强度
            major_strong = dataframe['adx_1h'] > 25
        else:
            major_bullish = dataframe['ema_50'] > dataframe['ema_200']
            major_bearish = dataframe['ema_50'] < dataframe['ema_200']
            major_strong = dataframe['adx'] > 25

        # ========== Layer 2: 15m趋势确认 ==========
        # 15m EMA多头排列
        minor_bullish = dataframe['ema_20'] > dataframe['ema_50']
        minor_bearish = dataframe['ema_20'] < dataframe['ema_50']

        # ========== 15m精确入场指标 (使用分级阈值) ==========
        # 1. ADX趋势强度
        adx_strong = dataframe['adx'] > adx_threshold

        # 2. DI动量差值
        di_bullish = dataframe['di_diff'] > di_threshold
        di_bearish = dataframe['di_diff'] < -di_threshold

        # 3. MACD 动量方向
        macd_bullish = dataframe['macdhist'] > 0
        macd_bearish = dataframe['macdhist'] < 0

        # 4. RSI 动能 - 做空需要明确在下跌过程中
        rsi_bullish = (dataframe['rsi'] > 55) & (dataframe['rsi'] < 70)
        rsi_bearish = (dataframe['rsi'] < 45) & (dataframe['rsi'] > 20)

        # 5. 成交量确认 (使用分级阈值)
        volume_above_avg = dataframe['volume_ratio'] > volume_ratio_threshold

        # 6. 避免在过度反弹时入场
        rsi_not_oversold = dataframe['rsi'] > 25

        # ========== 高准确性入场条件 ==========
        # 做多: 极严格条件 - 只在明确上升趋势中做多
        # v34.21: ADX>55, DI_diff>20, RSI 60-65
        long_entry = (
            major_bullish &             # 大周期多头 (必须)
            minor_bullish &            # 小周期多头 (必须)
            (dataframe['adx'] > 55) &   # ADX > 55 极强上升趋势确认 (原50)
            (dataframe['di_diff'] > 20) &  # DI_diff > 20 极强多头 (原18)
            macd_bullish &             # MACD > 0 确认多头
            (dataframe['rsi'] > 60) &   # RSI > 60 明确上升趋势
            (dataframe['rsi'] < 65) &   # RSI < 65 避免过热
            volume_above_avg &         # 成交量放量确认
            (dataframe['volume'] > 0)
        )

        # 做空: 保持原有严格条件
        short_entry = (
            major_bearish &             # 大周期空头 (必须)
            minor_bearish &            # 小周期空头 (必须)
            (dataframe['adx'] > 35) &   # ADX > 35 强趋势确认
            (dataframe['di_diff'] < -8) &  # DI_diff < -8 明确空头
            macd_bearish &             # MACD < 0 确认空头
            (dataframe['rsi'] < 45) &   # RSI < 45 明确弱势
            (dataframe['rsi'] > 20) &   # RSI > 20 避免超卖反弹
            volume_above_avg &         # 成交量放量确认
            (dataframe['volume'] > 0)
        )

        # ========== 设置入场信号 ==========
        # 启用做多和做空
        dataframe.loc[long_entry, 'enter_long'] = 1
        dataframe.loc[long_entry, 'enter_tag'] = f'mtf_long{"_major" if is_major else "_minor"}'

        dataframe.loc[short_entry, 'enter_short'] = 1
        dataframe.loc[short_entry, 'enter_tag'] = f'mtf_short{"_major" if is_major else "_minor"}'

        return dataframe

    def populate_exit_trend(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
        """
        Exit signals - v34.8: 禁用标准退出信号

        退出逻辑已移至 custom_exit 方法：
        - 只有盈利且信号强烈反转才退出
        - 亏损时不触发退出信号

        这样可以避免假信号导致的过早退出
        """
        # 禁用标准退出信号，使用 custom_exit 处理
        dataframe['exit_long'] = 0
        dataframe['exit_short'] = 0

        return dataframe

    def custom_stake_amount(self, pair: str, current_time: datetime, current_rate: float,
                            proposed_stake: float, min_stake: Optional[float], max_stake: float,
                            leverage: float, entry_tag: Optional[str], side: str,
                            **kwargs) -> float:
        """
        v34_4: 根据FreqAI预测调整仓位

        策略:
        - freqai_boost（FreqAI预测一致）: 仓位增加50%
        - tech_only（仅技术指标）: 正常仓位
        """
        if entry_tag == 'freqai_boost':
            # FreqAI确认的交易，增加仓位50%
            return proposed_stake * 1.5
        else:
            # 正常仓位
            return proposed_stake

    def custom_stoploss(
        self,
        pair: str,
        trade: Trade,
        current_time: datetime,
        current_rate: float,
        current_profit: float,
        after_fill: bool,
        **kwargs
    ) -> Optional[float]:
        """
        v34.21: 动态止损策略

        核心逻辑:
        - 初始止损宽松 (50%)，允许交易呼吸
        - 盈利后动态收紧止损
        - 方向正确时缓慢收紧，方向改变时快速收紧
        """
        # 收紧后的止损值 (只能返回比初始止损更小的值)
        # 多头: 止损应为负值，越负越紧
        # 空头: 止损应为正值，越小越紧
        if trade.is_short:
            # 空头止损: 基于当前价格计算
            # 如果盈利多，可以接受更紧的止损
            if current_profit >= 0.10:
                return 0.05   # 5%止损（紧）
            elif current_profit >= 0.06:
                return 0.08   # 8%止损
            elif current_profit >= 0.03:
                return 0.12   # 12%止损
            else:
                return 0.15   # 15%止损（仍然宽松）
        else:
            # 多头止损: 基于当前价格计算
            if current_profit >= 0.10:
                return -0.05   # 5%止损（紧）
            elif current_profit >= 0.06:
                return -0.08   # 8%止损
            elif current_profit >= 0.03:
                return -0.12   # 12%止损
            else:
                return -0.15   # 15%止损（仍然宽松）

    def adjust_trade_position(
        self,
        trade: Trade,
        current_time: datetime,
        current_rate: float,
        current_profit: float,
        min_stake: float | None,
        max_stake: float,
        current_entry_rate: float,
        current_exit_rate: float,
        current_entry_profit: float,
        current_exit_profit: float,
        **kwargs
    ) -> float | None | tuple[float | None, str | None]:
        """
        V31: 补仓策略

        补仓条件:
        1. 当前有亏损（亏损5%以上触发）
        2. 补仓次数未达上限（最多3次）
        3. 信号仍然有效

        补仓比例: 初始仓位的50%
        """
        # 防止在订单执行期间调整
        if trade.has_open_orders:
            return None

        dataframe, _ = self.dp.get_analyzed_dataframe(trade.pair, self.timeframe)
        if dataframe is None or dataframe.empty:
            return None

        last_candle = dataframe.iloc[-1]

        # 检查是否允许补仓
        if trade.nr_of_successful_entries >= self.max_entry_position_adjustment + 1:
            return None  # 已达补仓上限

        # 检查亏损是否达到补仓阈值
        if current_profit > self.add_position_threshold:
            return None  # 亏损不够，不补仓

        # 检查信号方向是否仍然有效
        if trade.is_short:
            # 空单: 检查做空信号
            signal_valid = (
                last_candle.get('enter_short', 0) == 1 or
                last_candle.get('macdhist', 0) < 0
            )
        else:
            # 多单: 检查做多信号
            signal_valid = (
                last_candle.get('enter_long', 0) == 1 or
                last_candle.get('macdhist', 0) > 0
            )

        if not signal_valid:
            return None  # 信号不再有效，不补仓

        # 计算补仓金额
        current_stake = trade.stake_amount
        add_stake = current_stake * self.add_position_ratio
        add_stake = min(add_stake, max_stake)
        if min_stake and add_stake < min_stake:
            add_stake = min_stake

        logger.info(f"📊 [补仓] {trade.pair} {'做空' if trade.is_short else '做多'} "
                   f"当前亏损: {current_profit:.2%}, 补仓金额: {add_stake:.2f} USDT")

        return add_stake, "dca_increase"

    def custom_exit(
        self,
        pair: str,
        trade: Trade,
        current_time: datetime,
        current_rate: float,
        current_profit: float,
        **kwargs
    ) -> Optional[str]:
        """
        v34.21 自定义退出 - 减少亏损

        退出条件：
        1. 盈利时: 3个反转信号退出
        2. 小幅亏损(-3%以内)时: 4个反转信号（全部指标都反转）退出
        3. 大幅亏损超过-3%: 需要4个信号 + ADX<25（趋势反转确认）

        反转信号：
        - 多单：EMA死叉 + MACD死叉 + DI反转 + RSI超买回落
        - 空单：EMA金叉 + MACD金叉 + DI反转 + RSI超卖回升
        """
        # 获取当前K线数据
        dataframe, _ = self.dp.get_analyzed_dataframe(pair, self.timeframe)
        if dataframe is None or len(dataframe) < 1:
            return None

        last_candle = dataframe.iloc[-1]

        # 多单退出检查
        if not trade.is_short:
            # 计算反转信号数量
            reverse_signals = 0

            # 1. EMA死叉
            if last_candle["ema_20"] < last_candle["ema_50"]:
                reverse_signals += 1

            # 2. MACD死叉
            if last_candle["macdhist"] < 0:
                reverse_signals += 1

            # 3. DI反转（空头增强）
            if last_candle["minus_di"] > last_candle["plus_di"]:
                reverse_signals += 1

            # 4. RSI超买回落 (使用 hyperopt 参数)
            if last_candle["rsi"] > self.exit_rsi_overbought:
                reverse_signals += 1

            # 盈利时3个信号退出
            if current_profit > 0 and reverse_signals >= 3:
                return 'profit_reverse_signal'

            # 小幅亏损(-3%以内)时4个信号退出
            if -0.03 < current_profit <= 0 and reverse_signals >= 4:
                return 'small_loss_exit'

            # 大幅亏损(-3%以上)时4个信号+ADX<25退出
            if current_profit <= -0.03 and reverse_signals >= 4 and last_candle["adx"] < 25:
                return 'big_loss_exit'

        # 空单退出检查
        else:
            reverse_signals = 0

            # 1. EMA金叉
            if last_candle["ema_20"] > last_candle["ema_50"]:
                reverse_signals += 1

            # 2. MACD金叉
            if last_candle["macdhist"] > 0:
                reverse_signals += 1

            # 3. DI反转（多头增强）
            if last_candle["plus_di"] > last_candle["minus_di"]:
                reverse_signals += 1

            # 4. RSI超卖回升 (使用 hyperopt 参数)
            if last_candle["rsi"] < self.exit_rsi_oversold:
                reverse_signals += 1

            # 盈利时3个信号退出
            if current_profit > 0 and reverse_signals >= 3:
                return 'profit_reverse_signal'

            # 小幅亏损(-3%以内)时4个信号退出
            if -0.03 < current_profit <= 0 and reverse_signals >= 4:
                return 'small_loss_exit'

            # 大幅亏损(-3%以上)时4个信号+ADX<25退出
            if current_profit <= -0.03 and reverse_signals >= 4 and last_candle["adx"] < 25:
                return 'big_loss_exit'

        return None

    def leverage(
        self,
        pair: str,
        current_time: datetime,
        current_rate: float,
        proposed_leverage: float,
        max_leverage: float,
        entry_tag: Optional[str],
        side: str,
        **kwargs
    ) -> float:
        """
        智能杠杆策略 v18 - 对高置信度交易使用杠杆

        关键改进:
        1. 对多空双方都使用杠杆 (双向策略)
        2. 杠杆只能是自然数: 1x, 2x, 3x
        3. 根据置信度分配杠杆

        杠杆分配:
        - 置信度 < 0.80: 1x (不使用杠杆)
        - 置信度 0.80-0.90: 2x
        - 置信度 >= 0.90: 3x

        Returns:
            杠杆倍数
        """
        if not self.leverage_enabled:
            return 1

        # 获取当前K线数据
        dataframe, _ = self.dp.get_analyzed_dataframe(pair, self.timeframe)
        if dataframe is None or len(dataframe) < 1:
            return 1

        last_candle = dataframe.iloc[-1]

        # v18: 对多空双方都应用杠杆

        # 计算置信度
        confidence = self._calculate_trade_confidence(pair, side, last_candle)

        # 根据置信度分配杠杆 (返回整数)
        if confidence >= self.leverage_threshold_very_high:  # >= 0.90
            leverage = 3
        elif confidence >= 0.80:
            leverage = 2
        else:
            leverage = 1

        if leverage > 1:
            self._leverage_cache[pair] = leverage
            logger.info(f"杠杆决策: {pair} {side} 置信度={confidence:.3f} 杠杆={leverage}x")

        return leverage

    def _calculate_trade_confidence(self, pair: str, side: str, last_candle: pd.Series) -> float:
        """
        计算交易置信度 (0.0 - 1.0)

        综合考虑:
        1. FreqAI预测强度 (权重 40%)
        2. 技术指标一致性 (权重 35%)
        3. 市场趋势强度 (权重 25%)

        Returns:
            置信度分数 (0.0 - 1.0)
        """
        confidence_scores = []
        weights = []

        # 1. FreqAI预测强度 (权重 40%)
        if self.freqai_enabled and "&-s_close_mean" in last_candle:
            pred_strength = abs(last_candle["&-s_close_mean"])
            # 归一化: 0.5以上为高置信度
            freqai_confidence = min(pred_strength / 0.5, 1.0)
            confidence_scores.append(freqai_confidence)
            weights.append(0.40)
            logger.debug(f"  FreqAI置信度: {freqai_confidence:.3f} (pred={pred_strength:.4f})")

        # 2. 技术指标一致性 (权重 35%)
        tech_confidence = self._calculate_technical_confidence(last_candle, side)
        confidence_scores.append(tech_confidence)
        weights.append(0.35)

        # 3. 市场趋势强度 (权重 25%)
        trend_confidence = self._calculate_trend_confidence(last_candle, side)
        confidence_scores.append(trend_confidence)
        weights.append(0.25)

        # 加权平均
        if confidence_scores and weights:
            total_weight = sum(weights)
            weighted_confidence = sum(s * w for s, w in zip(confidence_scores, weights)) / total_weight
            return min(max(weighted_confidence, 0.0), 1.0)

        return 0.3  # 默认低置信度

    def _calculate_technical_confidence(self, last_candle: pd.Series, side: str) -> float:
        """
        计算技术指标一致性置信度

        检查项:
        - ADX强度 (>30 高置信度)
        - DI方向一致性
        - RSI不极端
        - MACD方向一致
        - 成交量放大
        """
        score = 0.0
        count = 0

        # ADX强度
        if 'adx' in last_candle:
            adx = last_candle['adx']
            if adx > 35:
                score += 1.0
            elif adx > 30:
                score += 0.8
            elif adx > 25:
                score += 0.6
            else:
                score += 0.3
            count += 1

        # DI方向一致性
        if 'di_diff' in last_candle:
            di_diff = last_candle['di_diff']
            if side == 'long' and di_diff > 15:
                score += 1.0
            elif side == 'short' and di_diff < -15:
                score += 1.0
            elif side == 'long' and di_diff > 8:
                score += 0.7
            elif side == 'short' and di_diff < -8:
                score += 0.7
            else:
                score += 0.3
            count += 1

        # RSI合理性 (不极端)
        if 'rsi' in last_candle:
            rsi = last_candle['rsi']
            if side == 'long':
                if 40 <= rsi <= 55:  # 上升趋势初期
                    score += 1.0
                elif 35 <= rsi <= 60:
                    score += 0.7
                else:
                    score += 0.4
            else:  # short
                if 45 <= rsi <= 60:  # 下降趋势初期
                    score += 1.0
                elif 40 <= rsi <= 65:
                    score += 0.7
                else:
                    score += 0.4
            count += 1

        # MACD方向
        if 'macd' in last_candle and 'macdsignal' in last_candle:
            macd_diff = last_candle['macd'] - last_candle['macdsignal']
            if side == 'long' and macd_diff > 0:
                score += 1.0
            elif side == 'short' and macd_diff < 0:
                score += 1.0
            else:
                score += 0.3
            count += 1

        # 成交量放大
        if 'volume_ratio' in last_candle:
            vol_ratio = last_candle['volume_ratio']
            if vol_ratio > 2.0:
                score += 1.0
            elif vol_ratio > 1.5:
                score += 0.8
            elif vol_ratio > 1.2:
                score += 0.6
            else:
                score += 0.3
            count += 1

        return score / count if count > 0 else 0.5

    def _calculate_trend_confidence(self, last_candle: pd.Series, side: str) -> float:
        """
        计算趋势强度置信度

        检查项:
        - EMA排列 (多头/空头排列)
        - 市场regime一致性
        - 动量方向
        """
        score = 0.0
        count = 0

        # EMA排列
        if 'ema_20' in last_candle and 'ema_50' in last_candle and 'ema_100' in last_candle:
            ema_20 = last_candle['ema_20']
            ema_50 = last_candle['ema_50']
            ema_100 = last_candle['ema_100']

            if side == 'long':
                # 多头排列: EMA20 > EMA50 > EMA100
                if ema_20 > ema_50 > ema_100:
                    score += 1.0
                elif ema_20 > ema_50:
                    score += 0.7
                else:
                    score += 0.3
            else:  # short
                # 空头排列: EMA20 < EMA50 < EMA100
                if ema_20 < ema_50 < ema_100:
                    score += 1.0
                elif ema_20 < ema_50:
                    score += 0.7
                else:
                    score += 0.3
            count += 1

        # 市场regime
        if 'market_regime' in last_candle:
            regime = last_candle['market_regime']
            if side == 'long' and regime > 0:
                score += 1.0
            elif side == 'short' and regime < 0:
                score += 1.0
            else:
                score += 0.3
            count += 1

        # 动量方向
        if 'momentum_5' in last_candle:
            momentum = last_candle['momentum_5']
            if side == 'long' and momentum > 0.01:
                score += 1.0
            elif side == 'short' and momentum < -0.01:
                score += 1.0
            elif side == 'long' and momentum > 0:
                score += 0.6
            elif side == 'short' and momentum < 0:
                score += 0.6
            else:
                score += 0.3
            count += 1

        return score / count if count > 0 else 0.5

    def _get_leverage_from_confidence(self, confidence: float) -> int:
        """
        根据置信度返回杠杆倍数

        Args:
            confidence: 置信度 (0.0 - 1.0)

        Returns:
            杠杆倍数 (整数 1-5)
        """
        if confidence >= self.leverage_threshold_extreme:
            return 5   # 极端置信度: 5x杠杆
        elif confidence >= self.leverage_threshold_very_high:
            return 4   # 极高置信度: 4x杠杆
        elif confidence >= self.leverage_threshold_high:
            return 3   # 高置信度: 3x杠杆
        elif confidence >= self.leverage_threshold_medium:
            return 2   # 中置信度: 2x杠杆
        else:
            return 1   # 低置信度: 1x杠杆

    def _get_stoploss_for_leverage(self, leverage_value: float) -> float:
        """
        根据杠杆计算止损比例

        逻辑: 使用平方根衰减，给交易更多波动空间
        - 基础止损 8% (无杠杆时的标的波动)
        - 杠杆越高，止损越小 (但不是线性关系)
        - 使用平方根: stop_loss = base_stoploss / sqrt(leverage)

        对应表:
        | 杠杆 | 止损比例 |
        |------|---------|
        | 1x   | -8.0%   |
        | 2x   | -5.66%  |
        | 3x   | -4.62%  |

        Args:
            leverage_value: 杠杆倍数

        Returns:
            止损比例 (负数)
        """
        import math

        # 确保杠杆至少为1
        leverage_value = max(leverage_value, 1.0)

        # 使用平方根衰减，给交易更多波动空间
        # 这比线性衰减更合理，因为高杠杆需要更大的止损空间
        adjusted_stoploss = self.base_stoploss / math.sqrt(leverage_value)

        # 设置最小止损，避免过于敏感被误触
        adjusted_stoploss = max(adjusted_stoploss, self.min_stoploss_pct)

        return -adjusted_stoploss

    def _is_backtesting(self) -> bool:
        """检测是否在回测模式"""
        # 方法1: 检查 Dataprovider 是否有回测特征
        if hasattr(self.dp, 'runmode'):
            from freqtrade.enums import RunMode
            return self.dp.runmode in (RunMode.BACKTEST, RunMode.HYPEROPT)

        # 方法2: 检查是否能获取实时订单簿 (回测时无法获取)
        # 这个检查在 confirm_trade_entry 中进行，因为这里 dp 可能还未初始化
        return False

    def confirm_trade_entry(
        self,
        pair: str,
        order_type: str,
        amount: float,
        rate: float,
        time_in_force: str,
        current_time: datetime,
        entry_tag: Optional[str],
        side: str,
        **kwargs
    ) -> bool:
        """Entry confirmation with additional filters including sentiment and on-chain data"""

        # ========== 多空平衡检查 ==========
        # 检查当前持仓的多空比例，确保每个方向不超过 max_open_trades / 2
        trades = Trade.get_open_trades()
        if trades:
            long_trades = [t for t in trades if not t.is_short]
            short_trades = [t for t in trades if t.is_short]

            # 计算每个方向的最大订单数 = max_open_trades / 2
            max_per_direction = max(1, self.max_open_trades // 2)
            long_count = len(long_trades)
            short_count = len(short_trades)

            # 如果该方向订单数已达上限，禁止开仓
            if side == 'long' and long_count >= max_per_direction:
                logger.info(f"⚖️ [多空平衡拒绝] {pair} {side}: 多单数量已达上限 ({long_count}/{max_per_direction})")
                return False

            if side == 'short' and short_count >= max_per_direction:
                logger.info(f"⚖️ [多空平衡拒绝] {pair} {side}: 空单数量已达上限 ({short_count}/{max_per_direction})")
                return False

            # 额外检查: 资金平衡 (如果总订单数超过阈值)
            total_count = long_count + short_count
            if total_count >= max_per_direction * 1.5:  # 当总订单超过单个方向上限的1.5倍时检查资金
                long_stake = sum(t.stake_amount for t in long_trades)
                short_stake = sum(t.stake_amount for t in short_trades)
                total_stake = long_stake + short_stake
                threshold = self.balance_check_threshold
                if side == 'long' and (long_stake > total_stake * threshold):
                    logger.info(f"⚖️ [多空平衡拒绝] {pair} {side}: 多单资金占比过高 (资金:{long_stake:.2f}/{total_stake:.2f})")
                    return False
                if side == 'short' and (short_stake > total_stake * threshold):
                    logger.info(f"⚖️ [多空平衡拒绝] {pair} {side}: 空单资金占比过高 (资金:{short_stake:.2f}/{total_stake:.2f})")
                    return False

        # ========== 回测模式检测 ==========
        # 回测时跳过所有实时数据检查，因为无法获取历史API数据
        is_backtest = self._is_backtesting()
        if is_backtest:
            # 回测模式: 只做基础技术指标检查
            dataframe, _ = self.dp.get_analyzed_dataframe(pair, self.timeframe)
            if dataframe is not None and not dataframe.empty:
                last_candle = dataframe.iloc[-1]
                # FreqAI check
                if self.freqai_enabled and "do_predict" in last_candle:
                    if last_candle["do_predict"] != 1:
                        logger.info(f"🔍 [回测拒绝] {pair} {side}: do_predict={last_candle.get('do_predict', 'N/A')}")
                        return False
                # RSI extreme check
                if 'rsi' in last_candle:
                    if side == 'long' and last_candle['rsi'] > 70:
                        logger.info(f"🔍 [回测拒绝] {pair} {side}: RSI过高 {last_candle['rsi']:.1f}")
                        return False
                    if side == 'short' and last_candle['rsi'] < 30:
                        logger.info(f"🔍 [回测拒绝] {pair} {side}: RSI过低 {last_candle['rsi']:.1f}")
                        return False
            return True

        dataframe, _ = self.dp.get_analyzed_dataframe(pair, self.timeframe)

        if dataframe is None or dataframe.empty:
            logger.info(f"🔍 [拒绝] {pair} {side}: 无数据")
            return True

        last_candle = dataframe.iloc[-1]

        # FreqAI check
        if self.freqai_enabled and "do_predict" in last_candle:
            if last_candle["do_predict"] != 1:
                logger.info(f"🚫 [FreqAI拒绝] {pair} {side}: do_predict={last_candle.get('do_predict', 'N/A')}")
                return False

        # Volatility check - reject high volatility
        if 'atr_pct' in last_candle and last_candle['atr_pct'] > 0.04:
            logger.info(f"🚫 [波动率拒绝] {pair} {side}: ATR%={last_candle['atr_pct']:.2%} > 4%")
            return False

        # ADX check - need trend strength
        if 'adx' in last_candle and last_candle['adx'] < 20:
            logger.info(f"🚫 [ADX拒绝] {pair} {side}: ADX={last_candle['adx']:.1f} < 20")
            return False

        # RSI extreme check
        if 'rsi' in last_candle:
            if side == 'long' and last_candle['rsi'] > 70:
                logger.info(f"🚫 [RSI拒绝] {pair} {side}: RSI={last_candle['rsi']:.1f} > 70")
                return False
            if side == 'short' and last_candle['rsi'] < 30:
                logger.info(f"🚫 [RSI拒绝] {pair} {side}: RSI={last_candle['rsi']:.1f} < 30")
                return False

        # Momentum direction check - don't enter against momentum
        if 'momentum_5' in last_candle:
            if side == 'long' and last_candle['momentum_5'] < -0.01:
                logger.info(f"🚫 [动量拒绝] {pair} {side}: momentum={last_candle['momentum_5']:.3f} < -0.01")
                return False
            if side == 'short' and last_candle['momentum_5'] > 0.01:
                logger.info(f"🚫 [动量拒绝] {pair} {side}: momentum={last_candle['momentum_5']:.3f} > 0.01")
                return False

        # DI alignment check - ensure trend direction matches trade
        if 'di_diff' in last_candle:
            if side == 'long' and last_candle['di_diff'] < 0:
                logger.info(f"🚫 [DI拒绝] {pair} {side}: DI_diff={last_candle['di_diff']:.1f} < 0")
                return False
            if side == 'short' and last_candle['di_diff'] > 0:
                logger.info(f"🚫 [DI拒绝] {pair} {side}: DI_diff={last_candle['di_diff']:.1f} > 0")
                return False

        # ========== 情绪分析检查 (仅实盘) ==========
        sentiment_allowed, sentiment_reason = self.check_sentiment_filter(pair, side, last_candle)
        if not sentiment_allowed:
            logger.warning(f"🚫 [情绪过滤] {pair} {side} 被拒绝: {sentiment_reason}")
            return False

        # ========== 链上数据检查 (仅实盘) ==========
        onchain_allowed, onchain_reason = self.check_onchain_signals(pair, side)
        if not onchain_allowed:
            logger.warning(f"🚫 [链上过滤] {pair} {side} 被拒绝: {onchain_reason}")
            return False

        # ========== 新闻过滤检查 (仅实盘) ==========
        news_allowed, news_reason = self.check_news_filter(pair, side)
        if not news_allowed:
            logger.warning(f"🚫 [新闻过滤] {pair} {side} 被拒绝: {news_reason}")
            return False

        # ========== 订单簿分析检查 (仅实盘) ==========
        ob_allowed, ob_reason = self.check_orderbook_signals(pair, side)
        if not ob_allowed:
            logger.warning(f"🚫 [订单簿过滤] {pair} {side} 被拒绝: {ob_reason}")
            return False

        logger.info(f"✅ [确认入场] {pair} {side}: 所有检查通过")
        return True

    def check_sentiment_filter(self, pair: str, side: str, last_candle=None) -> tuple:
        """
        情绪分析过滤 (v34.19 修复版)

        改进: 允许在极度恐慌时做空，如果ADX很高（强趋势）

        Returns:
            (allowed: bool, reason: str)
        """
        if not self.enable_sentiment_check:
            return True, ""

        if not FREE_SENTIMENT_AVAILABLE:
            return True, "情绪分析模块不可用"

        try:
            import time
            # 提取基础币种
            symbol = pair.split('/')[0].replace(':USDT', '').replace('USDT', '')

            # 检查缓存
            cache_key = f"sentiment_{symbol}"
            if cache_key in self._sentiment_cache:
                cached = self._sentiment_cache[cache_key]
                if time.time() - cached['timestamp'] < self.sentiment_cache_ttl:
                    sentiment_data = cached['data']
                else:
                    sentiment_data = None
            else:
                sentiment_data = None

            # 获取情绪数据
            if sentiment_data is None:
                sentiment_data = get_sentiment_for_trading(symbol)
                self._sentiment_cache[cache_key] = {
                    'timestamp': time.time(),
                    'data': sentiment_data
                }

            sentiment = sentiment_data.get('sentiment', 0.5)
            fng = sentiment_data.get('fear_greed', 50)
            signal = sentiment_data.get('signal', 'hold')

            logger.info(f"📊 [情绪检查] {symbol}: 情绪={sentiment:.2f}, F&G={fng}, 信号={signal}")

            # v34.19: 极度恐慌时不再完全禁止做空
            if fng <= self.sentiment_extreme_fear:
                if side == 'short':
                    # 检查是否可以逆势做空
                    if self.sentiment_allow_contrarian_shorts and last_candle is not None:
                        adx = last_candle.get('adx', 0)
                        if adx > self.sentiment_contrarian_adx_threshold:
                            logger.info(f"📈 [情绪信号] 极度恐慌(F&G={fng})但ADX={adx:.1f}强, 允许做空")
                            return True, f"强趋势逆势做空: F&G={fng}, ADX={adx:.1f}"
                    logger.info(f"📈 [情绪信号] 极度恐慌(F&G={fng}), 禁止做空")
                    return False, f"极度恐慌(F&G={fng}), 应考虑做多"
                logger.info(f"✅ [情绪信号] 极度恐慌(F&G={fng}), 允许做多")

            # 极度贪婪时禁止做多 (逆向操作)
            elif fng >= self.sentiment_extreme_greed:
                if side == 'long':
                    logger.info(f"📉 [情绪信号] 极度贪婪(F&G={fng}), 禁止做多")
                    return False, f"极度贪婪(F&G={fng}), 应考虑做空"
                logger.info(f"✅ [情绪信号] 极度贪婪(F&G={fng}), 允许做空")

            # 中性区间 - 根据综合情绪判断
            else:
                if sentiment >= self.sentiment_bullish_threshold and side == 'short':
                    logger.info(f"📉 [情绪信号] 情绪看涨({sentiment:.2f}), 不建议做空")
                    return False, f"情绪看涨({sentiment:.2f})"
                elif sentiment <= self.sentiment_bearish_threshold and side == 'long':
                    logger.info(f"📉 [情绪信号] 情绪看跌({sentiment:.2f}), 不建议做多")
                    return False, f"情绪看跌({sentiment:.2f})"

            return True, f"情绪检查通过: sentiment={sentiment:.2f}, F&G={fng}"

        except Exception as e:
            logger.warning(f"⚠️ [情绪检查] 异常: {e}")
            return True, f"检查异常: {e}"

    def check_onchain_signals(self, pair: str, side: str) -> tuple:
        """
        链上数据检查 (资金费率、多空比等)

        Returns:
            (allowed: bool, reason: str)
        """
        if not self.enable_onchain_filter:
            return True, ""

        if not FREE_ONCHAIN_AVAILABLE:
            return True, "链上数据模块不可用"

        try:
            import time
            # 提取基础币种
            base_coin = pair.split('/')[0].replace(':USDT', '').replace('USDT', '')

            # 检查缓存
            cache_key = f"onchain_{base_coin}"
            if cache_key in self._onchain_cache:
                cached = self._onchain_cache[cache_key]
                if time.time() - cached['timestamp'] < self.sentiment_cache_ttl:
                    sentiment = cached['data']
                else:
                    sentiment = None
            else:
                sentiment = None

            # 获取链上数据
            if sentiment is None:
                sentiment = get_market_sentiment(base_coin)
                self._onchain_cache[cache_key] = {
                    'timestamp': time.time(),
                    'data': sentiment
                }

            overall_signal = sentiment.get('overall_signal', 'neutral')
            confidence = sentiment.get('confidence', 0)

            # 获取资金费率
            funding = sentiment.get('funding_rate', {})
            funding_rate_pct = funding.get('rate_pct', 0)
            funding_signal = funding.get('signal', 'neutral')

            # 获取多空比
            lsr = sentiment.get('long_short_ratio', {})
            lsr_ratio = lsr.get('longShortRatio', 1)

            logger.info(f"📊 [链上数据] {base_coin}: 资金费率={funding_rate_pct:.4f}%, "
                       f"多空比={lsr_ratio:.2f}, 综合信号={overall_signal}")

            # 极端资金费率过滤
            if abs(funding_rate_pct) > 0.1:  # 资金费率超过0.1%
                if side == 'long' and funding_signal == 'short_bias':
                    logger.warning(f"⚠️ [链上过滤] {base_coin} 资金费率极端看多({funding_rate_pct:.4f}%)")
                    if self.strict_funding_filter:
                        return False, f"资金费率极端({funding_rate_pct:.4f}%), 反向信号"
                elif side == 'short' and funding_signal == 'long_bias':
                    logger.warning(f"⚠️ [链上过滤] {base_coin} 资金费率极端看空({funding_rate_pct:.4f}%)")
                    if self.strict_funding_filter:
                        return False, f"资金费率极端({funding_rate_pct:.4f}%), 反向信号"

            # 极端多空比过滤
            if lsr_ratio > 2.0 and side == 'long':
                logger.warning(f"⚠️ [链上过滤] {base_coin} 多空比过高({lsr_ratio:.2f}), 多头拥挤")
                if self.strict_lsr_filter:
                    return False, f"多空比极端({lsr_ratio:.2f}), 多头拥挤"
            elif lsr_ratio < 0.5 and side == 'short':
                logger.warning(f"⚠️ [链上过滤] {base_coin} 多空比过低({lsr_ratio:.2f}), 空头拥挤")
                if self.strict_lsr_filter:
                    return False, f"多空比极端({lsr_ratio:.2f}), 空头拥挤"

            return True, f"链上数据检查通过: {overall_signal}({confidence:.2f})"

        except Exception as e:
            logger.error(f"📊 [链上数据] 检查异常: {e}")
            return True, f"检查异常: {e}"

    def check_news_filter(self, pair: str, side: str) -> tuple:
        """
        新闻过滤检查 - 检查是否有重大负面新闻

        Returns:
            (allowed: bool, reason: str)
        """
        if not self.enable_news_filter:
            return True, ""

        if not FREE_SENTIMENT_AVAILABLE:
            return True, "情绪模块不可用"

        try:
            import time
            # 提取基础币种
            symbol = pair.split('/')[0].replace(':USDT', '').replace('USDT', '').lower()

            # 检查冷却时间
            if symbol in self._last_news_check:
                if time.time() - self._last_news_check[symbol] < self.news_check_cooldown:
                    return True, "新闻检查冷却中"

            # 检查缓存
            cache_key = f"news_{symbol}"
            if cache_key in self._news_cache:
                cached = self._news_cache[cache_key]
                if time.time() - cached['timestamp'] < self.sentiment_cache_ttl:
                    news_data = cached['data']
                else:
                    news_data = None
            else:
                news_data = None

            # 获取新闻数据
            if news_data is None:
                news_data = fetch_elfa_token_news(symbol)
                self._news_cache[cache_key] = {
                    'timestamp': time.time(),
                    'data': news_data
                }
                self._last_news_check[symbol] = time.time()

            news_list = news_data.get('news', [])
            if not news_list:
                return True, "无相关新闻"

            # 检查最近新闻的情绪
            negative_count = 0
            for news in news_list[:5]:  # 检查最近5条新闻
                sentiment = news.get('sentiment', 'neutral')
                if sentiment == 'negative' or sentiment == 'bearish':
                    negative_count += 1

            # 如果负面新闻比例过高，拒绝交易
            negative_ratio = negative_count / min(len(news_list), 5)
            if negative_ratio >= self.news_negative_threshold:
                logger.warning(f"📰 [新闻过滤] {symbol} 负面新闻过多: {negative_count}/{min(len(news_list), 5)}")
                return False, f"负面新闻过多({negative_count}条)"

            logger.info(f"📰 [新闻检查] {symbol}: {len(news_list)}条新闻, 负面比例={negative_ratio:.0%}")
            return True, f"新闻检查通过: {len(news_list)}条新闻"

        except Exception as e:
            logger.warning(f"⚠️ [新闻检查] 异常: {e}")
            return True, f"检查异常: {e}"

    def check_orderbook_signals(self, pair: str, side: str) -> tuple:
        """
        订单簿分析检查 - 基于订单簿撮合机制分析市场深度和价格压力

        订单簿撮合三大原则：
        1. 价格优先：高买价、低卖价的订单优先成交
        2. 时间优先：同一价格档位下，先提交的订单优先成交
        3. 成交价取中间值：结合买价、卖价和前一笔成交价的中间值

        Returns:
            (allowed: bool, reason: str)
        """
        if not self.enable_orderbook_analysis:
            return True, ""

        if not ORDERBOOK_AVAILABLE:
            return True, "订单簿分析模块不可用"

        try:
            import time
            # 转换交易对格式 (如 BTC/USDT:USDT -> BTCUSDT)
            symbol = pair.replace('/', '').replace(':USDT', '')

            # 检查缓存
            cache_key = f"orderbook_{symbol}"
            if cache_key in self._orderbook_cache:
                cached = self._orderbook_cache[cache_key]
                if time.time() - cached['timestamp'] < self.orderbook_cache_ttl:
                    ob_data = cached['data']
                else:
                    ob_data = None
            else:
                ob_data = None

            # 获取订单簿数据 (使用Bitget API)
            if ob_data is None:
                ob_data = analyze_orderbook(symbol, "bitget")
                self._orderbook_cache[cache_key] = {
                    'timestamp': time.time(),
                    'data': ob_data
                }

            if 'error' in ob_data:
                return True, f"订单簿获取失败: {ob_data.get('error')}"

            obi = ob_data.get('obi', 0)
            weighted_obi = ob_data.get('weighted_obi', 0)
            spread = ob_data.get('spread', {})
            depth = ob_data.get('depth', {})
            pressure = ob_data.get('pressure', {})
            large_orders = ob_data.get('large_orders', {})
            overall_signal = ob_data.get('overall_signal', 'hold')
            confidence = ob_data.get('confidence', 0)

            logger.info(
                f"📊 [订单簿检查] {symbol}: OBI={obi:.3f}, 加权OBI={weighted_obi:.3f}, "
                f"价差={spread.get('spread_pct', 0):.4f}%, 信号={overall_signal}"
            )

            # 1. OBI检查 - 订单簿失衡
            # 做多时需要正向OBI支持，做空时需要负向OBI支持
            if abs(obi) >= self.obi_strong_threshold:
                if side == 'long' and obi < -self.obi_strong_threshold:
                    return False, f"OBI强烈看空({obi:.3f})"
                if side == 'short' and obi > self.obi_strong_threshold:
                    return False, f"OBI强烈看多({obi:.3f})"

            # 2. 加权OBI检查 - 更重视最优价位的订单
            if abs(weighted_obi) >= self.obi_strong_threshold:
                if side == 'long' and weighted_obi < -self.obi_strong_threshold:
                    return False, f"加权OBI强烈看空({weighted_obi:.3f})"
                if side == 'short' and weighted_obi > self.obi_strong_threshold:
                    return False, f"加权OBI强烈看多({weighted_obi:.3f})"

            # 3. 价差检查 - 价差过大时拒绝交易 (流动性不足)
            spread_pct = spread.get('spread_pct', 0)
            if spread_pct > self.spread_threshold:
                return False, f"价差过大({spread_pct:.4f}%), 流动性不足"

            # 4. 深度失衡检查
            depth_imbalance = depth.get('depth_imbalance', 0)
            if abs(depth_imbalance) >= self.obi_threshold:
                if side == 'long' and depth_imbalance < -self.obi_threshold:
                    logger.info(f"📉 [深度警告] {symbol} 卖方深度占优: {depth_imbalance:.3f}")
                if side == 'short' and depth_imbalance > self.obi_threshold:
                    logger.info(f"📈 [深度警告] {symbol} 买方深度占优: {depth_imbalance:.3f}")

            # 5. 价格压力检查
            pressure_signal = pressure.get('pressure_signal', 'neutral')
            net_pressure = pressure.get('net_pressure', 0)
            if abs(net_pressure) >= 0.5:
                if side == 'long' and pressure_signal == 'bearish':
                    return False, f"价格压力看空(net={net_pressure:.3f})"
                if side == 'short' and pressure_signal == 'bullish':
                    return False, f"价格压力看多(net={net_pressure:.3f})"

            # 6. 大单/冰山订单检测
            if large_orders.get('has_ask_wall') and side == 'long':
                logger.warning(f"🧊 [大单警告] {symbol} 检测到卖盘墙")
            if large_orders.get('has_bid_wall') and side == 'short':
                logger.warning(f"🧊 [大单警告] {symbol} 检测到买盘墙")

            # 7. 综合信号检查 (高置信度时遵循)
            if confidence >= 0.7:
                if overall_signal == 'sell' and side == 'long':
                    return False, f"综合信号看空(置信度={confidence:.2f})"
                if overall_signal == 'buy' and side == 'short':
                    return False, f"综合信号看多(置信度={confidence:.2f})"

            return True, f"订单簿检查通过: OBI={obi:.3f}, 信号={overall_signal}"

        except Exception as e:
            logger.warning(f"⚠️ [订单簿检查] 异常: {e}")
            return True, f"检查异常: {e}"