NewQuant/data/analysis/Trend.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "id": "initial_id",
   "metadata": {
    "collapsed": true,
    "ExecuteTime": {
     "end_time": "2025-12-08T15:18:02.157580Z",
     "start_time": "2025-12-08T15:18:02.151197Z"
    }
   },
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "import pandas as pd\n",
    "\n",
    "import warnings\n",
    "\n",
    "# 忽略所有警告\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "# --- 0. Configure your file path ---\n",
    "# Please replace 'your_futures_data.csv' with the actual path to your CSV file\n",
    "file_path = 'D:/PyProject/NewQuant/data/data/KQ_m@SHFE_hc/KQ_m@SHFE_hc_min15.csv'\n",
    "\n",
    "sns.set(style='whitegrid')\n",
    "plt.rcParams['font.sans-serif'] = ['SimHei']  # 用来正常显示中文标签\n",
    "plt.rcParams['axes.unicode_minus'] = False  # 用来正常显示负号\n"
   ],
   "outputs": [],
   "execution_count": 42
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2025-12-08T15:18:02.203253Z",
     "start_time": "2025-12-08T15:18:02.157580Z"
    }
   },
   "cell_type": "code",
   "source": [
    "\n",
    "# --- 1. Data Loading and Preprocessing ---\n",
    "def load_and_preprocess_data(file_path):\n",
    "    \"\"\"\n",
    "    Loads historical futures data and performs basic preprocessing.\n",
    "    Assumes data contains 'datetime', 'open', 'high', 'low', 'close', 'volume' columns.\n",
    "    \"\"\"\n",
    "    try:\n",
    "        df = pd.read_csv(file_path, parse_dates=['datetime'], index_col='datetime')\n",
    "        # Ensure data is sorted by time\n",
    "        df = df.sort_index()\n",
    "        # Check and handle missing values\n",
    "        initial_rows = len(df)\n",
    "        df.dropna(inplace=True)\n",
    "        if len(df) < initial_rows:\n",
    "            print(f\"Warning: Missing values found in data, deleted {initial_rows - len(df)} rows.\")\n",
    "\n",
    "        # Check if necessary columns exist\n",
    "        required_columns = ['open', 'high', 'low', 'close', 'volume']\n",
    "        if not all(col in df.columns for col in required_columns):\n",
    "            raise ValueError(f\"CSV file is missing required columns. Please ensure it contains: {required_columns}\")\n",
    "\n",
    "        print(f\"Successfully loaded {len(df)} rows of data.\")\n",
    "        print(\"First 5 rows of data:\")\n",
    "        print(df.head())\n",
    "        return df\n",
    "    except FileNotFoundError:\n",
    "        print(f\"Error: File '{file_path}' not found. Please check the path.\")\n",
    "        return None\n",
    "    except Exception as e:\n",
    "        print(f\"Error during data loading or preprocessing: {e}\")\n",
    "        return None\n",
    "\n",
    "\n",
    "df_raw = load_and_preprocess_data(file_path)\n",
    "print(df_raw)\n",
    "# df_df_raw = df_df_raw[df_df_raw.index >= '2024-01-01']"
   ],
   "id": "1638e05ca7ef1ac8",
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Successfully loaded 26087 rows of data.\n",
      "First 5 rows of data:\n",
      "                       open    high     low   close    volume   open_oi  \\\n",
      "datetime                                                                  \n",
      "2020-12-31 14:45:00  4534.0  4569.0  4528.0  4554.0   50433.0  526437.0   \n",
      "2021-01-04 09:00:00  4507.0  4510.0  4451.0  4468.0  149169.0  523892.0   \n",
      "2021-01-04 09:15:00  4468.0  4473.0  4423.0  4437.0   75329.0  513989.0   \n",
      "2021-01-04 09:30:00  4437.0  4460.0  4435.0  4447.0   47021.0  514326.0   \n",
      "2021-01-04 09:45:00  4447.0  4474.0  4446.0  4473.0   28511.0  509735.0   \n",
      "\n",
      "                     close_oi underlying_symbol  \n",
      "datetime                                         \n",
      "2020-12-31 14:45:00  523892.0       SHFE.hc2105  \n",
      "2021-01-04 09:00:00  513989.0       SHFE.hc2105  \n",
      "2021-01-04 09:15:00  514326.0       SHFE.hc2105  \n",
      "2021-01-04 09:30:00  509735.0       SHFE.hc2105  \n",
      "2021-01-04 09:45:00  509228.0       SHFE.hc2105  \n",
      "                       open    high     low   close    volume    open_oi  \\\n",
      "datetime                                                                   \n",
      "2020-12-31 14:45:00  4534.0  4569.0  4528.0  4554.0   50433.0   526437.0   \n",
      "2021-01-04 09:00:00  4507.0  4510.0  4451.0  4468.0  149169.0   523892.0   \n",
      "2021-01-04 09:15:00  4468.0  4473.0  4423.0  4437.0   75329.0   513989.0   \n",
      "2021-01-04 09:30:00  4437.0  4460.0  4435.0  4447.0   47021.0   514326.0   \n",
      "2021-01-04 09:45:00  4447.0  4474.0  4446.0  4473.0   28511.0   509735.0   \n",
      "...                     ...     ...     ...     ...       ...        ...   \n",
      "2025-09-19 21:30:00  3391.0  3393.0  3387.0  3389.0   15546.0  1406485.0   \n",
      "2025-09-19 21:45:00  3389.0  3392.0  3386.0  3388.0   11803.0  1406923.0   \n",
      "2025-09-19 22:00:00  3388.0  3391.0  3387.0  3387.0    7365.0  1406315.0   \n",
      "2025-09-19 22:15:00  3387.0  3390.0  3385.0  3389.0   10361.0  1405811.0   \n",
      "2025-09-19 22:30:00  3389.0  3390.0  3385.0  3387.0   10758.0  1404506.0   \n",
      "\n",
      "                      close_oi underlying_symbol  \n",
      "datetime                                          \n",
      "2020-12-31 14:45:00   523892.0       SHFE.hc2105  \n",
      "2021-01-04 09:00:00   513989.0       SHFE.hc2105  \n",
      "2021-01-04 09:15:00   514326.0       SHFE.hc2105  \n",
      "2021-01-04 09:30:00   509735.0       SHFE.hc2105  \n",
      "2021-01-04 09:45:00   509228.0       SHFE.hc2105  \n",
      "...                        ...               ...  \n",
      "2025-09-19 21:30:00  1406923.0       SHFE.hc2601  \n",
      "2025-09-19 21:45:00  1406315.0       SHFE.hc2601  \n",
      "2025-09-19 22:00:00  1405811.0       SHFE.hc2601  \n",
      "2025-09-19 22:15:00  1404506.0       SHFE.hc2601  \n",
      "2025-09-19 22:30:00  1402724.0       SHFE.hc2601  \n",
      "\n",
      "[26087 rows x 8 columns]\n"
     ]
    }
   ],
   "execution_count": 43
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2025-12-08T15:18:02.224837Z",
     "start_time": "2025-12-08T15:18:02.203253Z"
    }
   },
   "cell_type": "code",
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import talib\n",
    "from scipy.signal import stft\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from tqdm import tqdm\n",
    "\n",
    "class UniversalAnalyzer:\n",
    "    def __init__(self, df: pd.DataFrame, bars_per_day: int = 23):\n",
    "        \"\"\"\n",
    "        :param df: 包含 ['high', 'low', 'close'] 的原始数据\n",
    "        :param bars_per_day: 每天的K线数 (如小时线=23, 15分钟=96)\n",
    "        \"\"\"\n",
    "        self.df = df.reset_index(drop=True)\n",
    "        self.bars_per_day = bars_per_day\n",
    "        self.high_freq_days = 1.0  # 固定物理常数：过滤日内噪音\n",
    "        print(f\"--- 分析器初始化: {len(df)} Bars, {bars_per_day} Bars/Day ---\")\n",
    "\n",
    "    # ================= 工具函数：STFT 计算核心 =================\n",
    "    def _calc_stft_signal(self, prices, window_days):\n",
    "        window_len = int(window_days * self.bars_per_day)\n",
    "        if window_len % 2 != 0: window_len += 1\n",
    "\n",
    "        # 极短数据保护\n",
    "        if len(prices) < window_len: return np.zeros(len(prices))\n",
    "\n",
    "        # 1. 滚动计算方差进行标准化 (简单起见，这里用全量标准化，严谨可用Rolling)\n",
    "        # 为了速度和形态分析，全量标准化是可以接受的近似\n",
    "        norm = (prices - np.mean(prices)) / (np.std(prices) + 1e-9)\n",
    "\n",
    "        try:\n",
    "            f, t, Zxx = stft(norm, fs=self.bars_per_day, nperseg=window_len,\n",
    "                             noverlap=max(0, window_len // 2), padded=False)\n",
    "        except:\n",
    "            return np.zeros(len(prices))\n",
    "\n",
    "        # 提取当前时间步的能量 (Zxx最后一维是时间，我们需要对齐)\n",
    "        # STFT 输出时间轴通常比原数据短，这里我们需要做对齐处理\n",
    "        # 为简化分析，我们使用逐点滑窗计算的模拟函数\n",
    "        return None # 占位，实际调用下面的 _generate_full_series\n",
    "\n",
    "    def _generate_full_series(self, window_days):\n",
    "        \"\"\"生成全历史 STFT 强度序列\"\"\"\n",
    "        window_len = int(window_days * self.bars_per_day)\n",
    "        if window_len % 2 != 0: window_len += 1\n",
    "\n",
    "        closes = self.df['close'].values\n",
    "        n = len(closes)\n",
    "        signals = np.zeros(n)\n",
    "\n",
    "        # 步长设为1保证精度，设为5-10可加速\n",
    "        step = 1\n",
    "\n",
    "        # 使用 stride_tricks 进行超高速计算 (替代循环)\n",
    "        # 注意：这里计算的是每一个窗口的“瞬时”状态\n",
    "\n",
    "        print(f\"正在计算 Window={window_days} 的信号序列...\")\n",
    "        for i in range(window_len, n, step):\n",
    "            segment = closes[i-window_len:i]\n",
    "            std_val = np.std(segment)\n",
    "            if std_val < 1e-9: continue\n",
    "\n",
    "            norm = (segment - np.mean(segment)) / std_val\n",
    "\n",
    "            try:\n",
    "                f, t, Zxx = stft(norm, fs=self.bars_per_day, nperseg=window_len,\n",
    "                                 noverlap=max(0, window_len // 2), padded=False)\n",
    "\n",
    "                # 只取最后一列频谱\n",
    "                spec = np.abs(Zxx[:, -1]) ** 2\n",
    "                freqs = f\n",
    "\n",
    "                # 频率绑定\n",
    "                low_limit = (1.0 / window_days) + 1e-5 # 加上偏移量防止丢失基频\n",
    "\n",
    "                low_mask = freqs <= low_limit\n",
    "                high_mask = freqs > self.high_freq_days\n",
    "\n",
    "                low_e = np.sum(spec[low_mask])\n",
    "                high_e = np.sum(spec[high_mask])\n",
    "\n",
    "                signals[i] = low_e / (low_e + high_e + 1e-9)\n",
    "            except:\n",
    "                continue\n",
    "\n",
    "        return signals\n",
    "\n",
    "    # ================= 模块 1: 确定止损 (物理常数) =================\n",
    "    def step1_analyze_stop_loss(self):\n",
    "        print(\"\\n=== [Step 1] 止损参数标定 ===\")\n",
    "        high = self.df['high'].values\n",
    "        low = self.df['low'].values\n",
    "        close = self.df['close'].values\n",
    "\n",
    "        atr = talib.ATR(high, low, close, timeperiod=14)\n",
    "        tr = talib.TRANGE(high, low, close)\n",
    "\n",
    "        # 过滤无效数据\n",
    "        mask = (atr > 0) & (~np.isnan(atr))\n",
    "        ratios = tr[mask] / atr[mask]\n",
    "\n",
    "        p99 = np.percentile(ratios, 99)\n",
    "        p99_7 = np.percentile(ratios, 99.7) # 3倍标准差，塔勒布极限\n",
    "\n",
    "        print(f\"建议止损倍数 (基于 P99.7): {p99_7:.2f}\")\n",
    "        suggested = np.ceil(p99_7 * 2) / 2 # 向上取整到0.5\n",
    "        print(f\"-> 最终推荐 Stop Loss ATR Multiplier: {suggested}\")\n",
    "        return suggested\n",
    "\n",
    "    # ================= 模块 2: 确定窗口 (信噪比) =================\n",
    "    def step2_select_window(self, windows=[2.0, 3.0, 4.0, 5.0, 6.0]):\n",
    "        print(\"\\n=== [Step 2] 最佳窗口扫描 ===\")\n",
    "        best_w = None\n",
    "        max_std = -1\n",
    "\n",
    "        stats = []\n",
    "        for w in windows:\n",
    "            sigs = self._generate_full_series(w)\n",
    "            # 去除0值(预热期)\n",
    "            sigs = sigs[sigs > 0]\n",
    "\n",
    "            if len(sigs) == 0: continue\n",
    "\n",
    "            std_val = np.std(sigs)\n",
    "            stats.append((w, std_val))\n",
    "            print(f\"Window {w}: StdDev = {std_val:.4f}\")\n",
    "\n",
    "            if std_val > max_std:\n",
    "                max_std = std_val\n",
    "                best_w = w\n",
    "\n",
    "        print(f\"-> 最终推荐 Spectral Window Days: {best_w} (区分度最高)\")\n",
    "        return best_w\n",
    "\n",
    "    # ================= 模块 3: 确定阈值 (分布+收益) =================\n",
    "    def step3_calibrate_thresholds(self, best_window):\n",
    "        print(\"\\n=== [Step 3] 阈值校准 (分布 & IC分析) ===\")\n",
    "        # 1. 生成信号\n",
    "        signals = self._generate_full_series(best_window)\n",
    "        closes = self.df['close'].values\n",
    "\n",
    "        # 2. 基础分布统计 (PDF) -> 确定 Exit\n",
    "        valid_sigs = signals[signals > 0]\n",
    "        p50 = np.median(valid_sigs)\n",
    "        print(f\"分布中位数 (建议 Exit Threshold): {p50:.2f}\")\n",
    "\n",
    "        # 3. 收益分组分析 (Bucketing) -> 确定 Entry\n",
    "        # 计算未来收益 (持有窗口长度)\n",
    "        hold_period = int(best_window * self.bars_per_day)\n",
    "        fwd_ret = pd.Series(closes).pct_change(hold_period).shift(-hold_period)\n",
    "\n",
    "        # 确定方向: 简单用过去 hold_period 的涨跌作为方向过滤器\n",
    "        # 假设: Signal强 且 过去在涨 -> 做多; Signal强 且 过去在跌 -> 做空\n",
    "        past_ret = pd.Series(closes).pct_change(hold_period).fillna(0)\n",
    "        direction = np.sign(past_ret)\n",
    "\n",
    "        # 策略收益 = 方向 * 未来收益 (假设做对方向)\n",
    "        # 我们只关心 Signal 强度是否能放大收益\n",
    "        df_anl = pd.DataFrame({\n",
    "            'signal': signals,\n",
    "            'fwd_ret': fwd_ret,\n",
    "            'direction': direction\n",
    "        }).dropna()\n",
    "\n",
    "        df_anl = df_anl[df_anl['signal'] > 0]\n",
    "\n",
    "        # 核心假设: 我们在信号强时顺势交易\n",
    "        df_anl['strat_ret'] = np.sign(df_anl['direction']) * df_anl['fwd_ret']\n",
    "\n",
    "        # 分桶\n",
    "        df_anl['bucket'] = pd.cut(df_anl['signal'], bins=np.linspace(0, 1, 21))\n",
    "        grouped = df_anl.groupby('bucket', observed=False)['strat_ret'].mean()\n",
    "        counts = df_anl.groupby('bucket', observed=False)['strat_ret'].count()\n",
    "\n",
    "        # 绘图\n",
    "        fig, ax1 = plt.subplots(figsize=(12, 5))\n",
    "        colors = ['red' if v > 0 else 'green' for v in grouped.values]\n",
    "        grouped.plot(kind='bar', ax=ax1, color=colors, alpha=0.7)\n",
    "        ax1.set_title(f\"Signal Strength vs Profitability (Window={best_window})\")\n",
    "        ax1.set_ylabel(\"Avg Future Return\")\n",
    "\n",
    "        ax2 = ax1.twinx()\n",
    "        counts.plot(kind='line', ax=ax2, color='blue', marker='o')\n",
    "        plt.show()\n",
    "\n",
    "        print(\"请观察图表:\")\n",
    "        print(\"1. 寻找红色柱子稳定出现且较高的区域起始点 -> Entry Threshold\")\n",
    "        print(\"2. 避开深绿色柱子区域 (陷阱区)\")\n",
    "\n",
    "        return grouped\n",
    "\n",
    "    # ================= 模块 4: 热力图验证 (鲁棒性) =================\n",
    "    def step4_heatmap_validation(self, center_window, center_entry, stop_loss_mult):\n",
    "        print(\"\\n=== [Step 4] 热力图鲁棒性验证 ===\")\n",
    "\n",
    "        # 构建搜索网格 (在中心点附近波动)\n",
    "        windows = [center_window - 1.0, center_window - 0.5, center_window, center_window + 0.5, center_window + 1.0]\n",
    "        # 过滤掉不合理的窗口\n",
    "        windows = [w for w in windows if w >= 2.0]\n",
    "\n",
    "        entries = np.arange(max(0.6, center_entry - 0.15), min(0.95, center_entry + 0.15), 0.05)\n",
    "\n",
    "        results = pd.DataFrame(index=np.round(entries, 2), columns=windows)\n",
    "\n",
    "        # 这里使用快速回测逻辑 (简化版)\n",
    "        for w in tqdm(windows):\n",
    "            signals = self._generate_full_series(w)\n",
    "            closes = self.df['close'].values\n",
    "            atr = talib.ATR(self.df['high'], self.df['low'], closes, timeperiod=14)\n",
    "            atr = np.nan_to_num(atr, nan=np.mean(closes)*0.01)\n",
    "\n",
    "            for e in entries:\n",
    "                # 固定 Exit 为 0.45 (或基于 P50 的动态值，这里简化固定)\n",
    "                calmar = self._fast_backtest_calmar(\n",
    "                    signals, closes, atr,\n",
    "                    entry=e, exit=0.45, sl=stop_loss_mult\n",
    "                )\n",
    "                results.loc[np.round(e, 2), w] = calmar\n",
    "\n",
    "        plt.figure(figsize=(10, 6))\n",
    "        sns.heatmap(results.astype(float), annot=True, cmap=\"RdYlGn\", fmt=\".2f\")\n",
    "        plt.title(f\"Robustness Heatmap (SL={stop_loss_mult})\")\n",
    "        plt.gca().invert_yaxis()\n",
    "        plt.show()\n",
    "\n",
    "    def _fast_backtest_calmar(self, signals, closes, atr, entry, exit, sl):\n",
    "        # 极简向量化回测计算 Calmar\n",
    "        # 仅供参数敏感度分析，非精确回测\n",
    "        pos = np.zeros(len(closes))\n",
    "        # 简单的状态机模拟\n",
    "        curr_pos = 0\n",
    "        entry_px = 0\n",
    "\n",
    "        # 预计算方向 (动量)\n",
    "        mom = np.sign(pd.Series(closes).pct_change(5).fillna(0).values)\n",
    "\n",
    "        equity = [1.0]\n",
    "        for i in range(1, len(closes)):\n",
    "            px = closes[i]\n",
    "            sig = signals[i-1]\n",
    "            my_atr = atr[i-1]\n",
    "\n",
    "            pnl = 0\n",
    "            if curr_pos == 0:\n",
    "                if sig > entry:\n",
    "                    curr_pos = mom[i-1] # 顺势\n",
    "                    entry_px = px\n",
    "            elif curr_pos != 0:\n",
    "                # 止损\n",
    "                if curr_pos == 1 and px < (entry_px - sl * my_atr): curr_pos = 0\n",
    "                elif curr_pos == -1 and px > (entry_px + sl * my_atr): curr_pos = 0\n",
    "                # 离场\n",
    "                elif sig < exit: curr_pos = 0\n",
    "\n",
    "                # 计算 PnL\n",
    "                if curr_pos == 1: pnl = (px - closes[i-1]) / closes[i-1]\n",
    "                elif curr_pos == -1: pnl = (closes[i-1] - px) / closes[i-1]\n",
    "\n",
    "            equity.append(equity[-1] * (1 + pnl))\n",
    "\n",
    "        eq_series = pd.Series(equity)\n",
    "        dd = (eq_series / eq_series.cummax() - 1).min()\n",
    "        if dd == 0: return 0\n",
    "        ann_ret = (eq_series.iloc[-1] ** (252*self.bars_per_day/len(eq_series))) - 1\n",
    "        return ann_ret / abs(dd)"
   ],
   "id": "b98755d3f0552e8b",
   "outputs": [],
   "execution_count": 44
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2025-12-08T15:18:02.237244Z",
     "start_time": "2025-12-08T15:18:02.227739Z"
    }
   },
   "cell_type": "code",
   "source": [
    "# 假设是15分钟线，一天有 4小时(夜盘)+5.5小时(日盘) = 9.5小时 ≈ 38根bar?\n",
    "# 或者按照实际交易所 bar 数填写\n",
    "analyzer = UniversalAnalyzer(df_raw, bars_per_day=23)"
   ],
   "id": "57962b7585161e57",
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--- 分析器初始化: 26087 Bars, 23 Bars/Day ---\n"
     ]
    }
   ],
   "execution_count": 45
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2025-12-08T15:18:02.252250Z",
     "start_time": "2025-12-08T15:18:02.246891Z"
    }
   },
   "cell_type": "code",
   "source": [
    "# 这一步直接给你 StopLoss 参数\n",
    "sl_mult = analyzer.step1_analyze_stop_loss()\n",
    "# 记录这个值，例如 4.0"
   ],
   "id": "5a4888719bd6ccf",
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "=== [Step 1] 止损参数标定 ===\n",
      "建议止损倍数 (基于 P99.7): 3.49\n",
      "-> 最终推荐 Stop Loss ATR Multiplier: 3.5\n"
     ]
    }
   ],
   "execution_count": 46
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2025-12-08T15:18:12.422635Z",
     "start_time": "2025-12-08T15:18:02.258765Z"
    }
   },
   "cell_type": "code",
   "source": [
    "# 这一步通过打印的 StdDev 帮你找最敏感的窗口\n",
    "best_window = analyzer.step2_select_window([2.0, 3.0, 4.0, 5.0, 6.0, 8.0])\n",
    "# 记录这个值，例如 5.0 (StdDev最大)"
   ],
   "id": "65c036e6236cff5",
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "=== [Step 2] 最佳窗口扫描 ===\n",
      "正在计算 Window=2.0 的信号序列...\n",
      "Window 2.0: StdDev = 0.2481\n",
      "正在计算 Window=3.0 的信号序列...\n",
      "Window 3.0: StdDev = 0.2468\n",
      "正在计算 Window=4.0 的信号序列...\n",
      "Window 4.0: StdDev = 0.2326\n",
      "正在计算 Window=5.0 的信号序列...\n",
      "Window 5.0: StdDev = 0.2270\n",
      "正在计算 Window=6.0 的信号序列...\n",
      "Window 6.0: StdDev = 0.2172\n",
      "正在计算 Window=8.0 的信号序列...\n",
      "Window 8.0: StdDev = 0.2069\n",
      "-> 最终推荐 Spectral Window Days: 2.0 (区分度最高)\n"
     ]
    }
   ],
   "execution_count": 47
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2025-12-08T15:18:14.229745Z",
     "start_time": "2025-12-08T15:18:12.429188Z"
    }
   },
   "cell_type": "code",
   "source": [
    "# 这一步会画出柱状图\n",
    "# 1. 看控制台输出的 Median -> 这是你的 Exit Threshold (比如 0.42)\n",
    "# 2. 看弹出的图表 -> 找到红色柱子开始稳定且高的位置 -> 这是你的 Entry Threshold (比如 0.85)\n",
    "analyzer.step3_calibrate_thresholds(best_window)"
   ],
   "id": "dce8f46ee6f5c5dd",
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "=== [Step 3] 阈值校准 (分布 & IC分析) ===\n",
      "正在计算 Window=2.0 的信号序列...\n",
      "分布中位数 (建议 Exit Threshold): 0.71\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 1200x500 with 2 Axes>"
      ],
      "image/png": "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
     },
     "metadata": {},
     "output_type": "display_data",
     "jetTransient": {
      "display_id": null
     }
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "请观察图表:\n",
      "1. 寻找红色柱子稳定出现且较高的区域起始点 -> Entry Threshold\n",
      "2. 避开深绿色柱子区域 (陷阱区)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "bucket\n",
       "(0.0, 0.05]   -0.000708\n",
       "(0.05, 0.1]   -0.000032\n",
       "(0.1, 0.15]    0.001019\n",
       "(0.15, 0.2]   -0.000725\n",
       "(0.2, 0.25]    0.001073\n",
       "(0.25, 0.3]    0.000758\n",
       "(0.3, 0.35]    0.001053\n",
       "(0.35, 0.4]   -0.000009\n",
       "(0.4, 0.45]    0.000594\n",
       "(0.45, 0.5]    0.000253\n",
       "(0.5, 0.55]   -0.000740\n",
       "(0.55, 0.6]    0.001504\n",
       "(0.6, 0.65]   -0.000038\n",
       "(0.65, 0.7]   -0.000405\n",
       "(0.7, 0.75]   -0.000288\n",
       "(0.75, 0.8]   -0.000446\n",
       "(0.8, 0.85]    0.000360\n",
       "(0.85, 0.9]    0.000232\n",
       "(0.9, 0.95]    0.000909\n",
       "(0.95, 1.0]   -0.000822\n",
       "Name: strat_ret, dtype: float64"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "execution_count": 48
  },
  {
   "metadata": {
    "ExecuteTime": {
     "end_time": "2025-12-08T15:18:19.894590Z",
     "start_time": "2025-12-08T15:18:14.240383Z"
    }
   },
   "cell_type": "code",
   "source": [
    "# 把上面找到的参数填进去，跑热力图\n",
    "# center_window = 5.0, center_entry = 0.85, sl = 4.0\n",
    "analyzer.step4_heatmap_validation(best_window, 0.85, 4.0)"
   ],
   "id": "1fd393d531324435",
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "=== [Step 4] 热力图鲁棒性验证 ===\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/3 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "正在计算 Window=2.0 的信号序列...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 33%|███▎      | 1/3 [00:01<00:03,  1.77s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "正在计算 Window=2.5 的信号序列...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 67%|██████▋   | 2/3 [00:03<00:01,  1.81s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "正在计算 Window=3.0 的信号序列...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 3/3 [00:05<00:00,  1.83s/it]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 1000x600 with 2 Axes>"
      ],
      "image/png": "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
     },
     "metadata": {},
     "output_type": "display_data",
     "jetTransient": {
      "display_id": null
     }
    }
   ],
   "execution_count": 49
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}