"""
跨期价差均值回归回测：支持任意合约月份组合。
"""
import sqlite3
from collections import defaultdict
from statistics import mean, stdev

DB_PATH = "/Users/vipg/Documents/futures-data-warehouse/db/futures.db"

conn = sqlite3.connect(DB_PATH)
daily = conn.execute(
    "SELECT ts_code, trade_date, close FROM daily WHERE close IS NOT NULL AND close != '' ORDER BY trade_date"
).fetchall()
conn.close()

by_date = defaultdict(dict)
for ts_code, d, close in daily:
    by_date[d][ts_code] = float(close)

dates = sorted(by_date.keys())


def build_spread(month_a, month_b, cross_year=False):
    """
    构建两合约月份的价差序列。
    month_a, month_b: '01','05','09'
    cross_year: True 表示 year_b = year_a + 1（如 09-01 跨年）
    返回 [(date, ts_a, ts_b, p_a, p_b, spread)]
    """
    data = []
    for d in dates:
        items = by_date[d]
        for ts_code in list(items.keys()):
            if not ts_code.endswith(".ZCE") or not ts_code.startswith("FG"):
                continue
            yr, mon = ts_code[2:4], ts_code[4:6]
            if mon == month_a:
                yr_b = str(int(yr) + 1) if cross_year else yr
                ts_b = f"FG{yr_b}{month_b}.ZCE"
                if ts_b in items:
                    p_a = items[ts_code]
                    p_b = items[ts_b]
                    spread = p_b - p_a
                    data.append((d, ts_code, ts_b, p_a, p_b, spread))
    return data


def backtest_spread(spread_data, window=60, entry_z=2.0, exit_z=0.5, max_hold=20):
    equity = 1.0
    position = 0
    entry_spread = 0
    entry_p_a = 0
    entry_date = ""
    hold_days = 0
    trades = []
    daily_eq = []

    for i, (d, ts_a, ts_b, p_a, p_b, spread_val) in enumerate(spread_data):
        if i < window:
            daily_eq.append((d, equity))
            continue

        hist = [spread_data[j][5] for j in range(i - window, i)]
        roll_mean = mean(hist)
        roll_std = stdev(hist)

        signal = 0
        if spread_val > roll_mean + entry_z * roll_std:
            signal = -1  # 价差过大 → 空价差（空远月、多近月）
        elif spread_val < roll_mean - entry_z * roll_std:
            signal = 1

        should_exit = False
        if position != 0:
            hold_days += 1
            if hold_days >= max_hold:
                should_exit = True
            if position == 1 and spread_val >= roll_mean - exit_z * roll_std:
                should_exit = True
            elif position == -1 and spread_val <= roll_mean + exit_z * roll_std:
                should_exit = True

        if position != 0 and should_exit:
            if position == 1:
                pnl = (spread_val - entry_spread) / entry_p_a
            else:
                pnl = (entry_spread - spread_val) / entry_p_a
            equity *= (1 + pnl)
            trades.append((entry_date, d, entry_spread, spread_val, pnl, position))
            position = 0
            hold_days = 0

        if position == 0 and signal != 0:
            position = signal
            entry_spread = spread_val
            entry_p_a = p_a
            entry_date = d
            hold_days = 0

        daily_eq.append((d, equity))

    if position != 0 and spread_data:
        d, _, _, p_a, _, spread_val = spread_data[-1]
        if position == 1:
            pnl = (spread_val - entry_spread) / entry_p_a
        else:
            pnl = (entry_spread - spread_val) / entry_p_a
        equity *= (1 + pnl)
        trades.append((entry_date, d, entry_spread, spread_val, pnl, position))

    return trades, daily_eq


def print_results(trades, daily_eq, label):
    if not trades:
        print(f"\n{label}: 无交易")
        return
    wins = [t for t in trades if t[4] > 0]
    losses = [t for t in trades if t[4] <= 0]
    total_ret = daily_eq[-1][1] - 1.0 if daily_eq else 0
    wr = len(wins) / len(trades) if trades else 0

    returns = []
    for i in range(1, len(daily_eq)):
        if daily_eq[i-1][1] > 0:
            returns.append(daily_eq[i][1] / daily_eq[i-1][1] - 1)
    avg_ret = mean(returns) if returns else 0
    std_ret = stdev(returns) if len(returns) > 1 else 1
    sharpe = (avg_ret / std_ret) * (252 ** 0.5) if std_ret > 0 else 0

    peak = 1.0
    mdd = 0.0
    for _, e in daily_eq:
        if e > peak: peak = e
        dd = (peak - e) / peak
        if dd > mdd: mdd = dd

    print(f"\n{label}")
    print(f"  {'' if trades else '无'}交易次数: {len(trades)}")
    print(f"  总收益率: {total_ret:+.2%}")
    print(f"  夏普比率: {sharpe:.2f}")
    print(f"  最大回撤: {mdd:.2%}")
    print(f"  胜率: {wr:.0%}")
    if losses:
        avg_w = mean(t[4] for t in wins)
        avg_l = mean(t[4] for t in losses)
        print(f"  盈亏比: {abs(avg_w/avg_l):.2f}")


# ── 分析所有有效组合 ──────────────────────────

pairs = [
    ("01", "05", False, "同一年 01-05"),
    ("05", "09", False, "同一年 05-09"),
    ("09", "01", True,  "跨年 09-01"),
]

# 先看各组合的统计特征
print("=" * 60)
print("各合约组合价差统计")
print("=" * 60)
for ma, mb, cross, label in pairs:
    data = build_spread(ma, mb, cross)
    n = len(data)
    vals = [r[5] for r in data]
    pos = sum(1 for v in vals if v > 0)
    avg_s = mean(vals)
    std_s = stdev(vals)
    print(f"\n{label:>12}  {n:>5}天  均值{avg_s:>+7.1f}  σ{std_s:>6.1f}  正{pos:>4}({pos/n:.0%})")

# 回测各组合
print("\n" + "=" * 60)
print("均值回归回测 (window=60  entry=2σ  exit=0.5σ  max_hold=20)")
print("=" * 60)
for ma, mb, cross, label in pairs:
    data = build_spread(ma, mb, cross)
    trades, eq = backtest_spread(data, window=60, entry_z=2.0, exit_z=0.5, max_hold=20)
    print_results(trades, eq, label)

# 当前可交易组合 (202607)
print("\n" + "=" * 60)
print("当前可交易组合分析 (FG2609, FG2701, FG2705)")
print("=" * 60)

today = "20260717"
items = by_date.get(today, {})
for ts in ["FG2609.ZCE", "FG2701.ZCE", "FG2705.ZCE"]:
    if ts in items:
        print(f"  {ts}: {items[ts]}")
    else:
        print(f"  {ts}: 无当日数据")

# 当前配对及统计
pairs_now = [
    ("FG2609.ZCE", "FG2701.ZCE", "09-01(跨年)"),
    ("FG2701.ZCE", "FG2705.ZCE", "01-05(同一年)"),
    ("FG2609.ZCE", "FG2705.ZCE", "09-05(跨年)"),
]

for ts_a, ts_b, label in pairs_now:
    if ts_a not in items or ts_b not in items:
        print(f"\n{label}: {ts_a} 或 {ts_b} 无数据")
        continue

    spread_now = items[ts_b] - items[ts_a]
    print(f"\n{label}:  {ts_a} vs {ts_b}")
    print(f"  当前价差: {spread_now:+.1f}")

    # 找历史均值
    for ma, mb, cross, _ in pairs:
        if (ts_a[4:6] == ma and ts_b[4:6] == mb) or \
           (ts_a[4:6] == mb and ts_b[4:6] == ma and cross):
            data = build_spread(ma, mb, cross)
            if len(data) < 60:
                continue
            vals = [r[5] for r in data]
            # 近 60 天
            recent = vals[-60:]
            avg_r = mean(recent)
            std_r = stdev(recent)
            z = (spread_now - avg_r) / std_r if std_r else 0
            print(f"  近60日均值: {avg_r:.1f}  σ: {std_r:.1f}  z值: {z:.2f}")
            if abs(z) >= 2:
                print(f"  → 偏离 {z:.0f}σ，触发信号！")
            else:
                print(f"  → 未触发 (需 |z|>=2)")