修复项: - export_backtest_detail.py: 统一回测导出脚本的数据源调用逻辑 - test_trading_calendar.py: 交易日历功能测试 - verify_fix_result.py: 修复结果验证 - verify_mode_b.py: 模式 B 验证 策略修复: - momentum.py: 动量因子计算优化 - strategy.py: StrategyBase 数据获取修复(fetch_indices 返回 dict)
478 lines
17 KiB
Python
478 lines
17 KiB
Python
#!/usr/bin/env python3
|
||
"""
|
||
导出回测逐日明细到 JSON,供 HTML 回放器加载。
|
||
|
||
模式 B:指数信号 + ETF 收益(2020-01-01 ~ 2026-05-19)
|
||
|
||
用法:
|
||
python scripts/export_backtest_detail.py
|
||
"""
|
||
|
||
import sys
|
||
import json
|
||
import math
|
||
from pathlib import Path
|
||
|
||
import numpy as np
|
||
import pandas as pd
|
||
import yaml
|
||
|
||
project_root = Path(__file__).parent.parent
|
||
sys.path.insert(0, str(project_root))
|
||
|
||
from dotenv import load_dotenv
|
||
load_dotenv()
|
||
|
||
from datasource.tushare_source import TushareSource
|
||
from datasource.flask_api_source import FlaskAPIDataSource
|
||
from strategies.shared.factors.momentum import MomentumFactor
|
||
from strategies.shared.signals.selectors import TopNSelector
|
||
from framework.execution import BacktestExecutor
|
||
|
||
# ==================== 加载配置 ====================
|
||
config_path = project_root / 'strategies' / 'rotation' / 'config.yaml'
|
||
with open(config_path, 'r', encoding='utf-8') as f:
|
||
config = yaml.safe_load(f)
|
||
|
||
CODE_LIST = config['code_list']
|
||
SELECT_NUM = config['select_num']
|
||
N_DAYS = config['n_days']
|
||
TRADE_COST = config['trade_cost']
|
||
BOND_THRESHOLD = config.get('bond_threshold', {})
|
||
BOND_CODE = BOND_THRESHOLD.get('bond_code', '931862.CSI')
|
||
BOND_RATIO = BOND_THRESHOLD.get('ratio', 1.0)
|
||
|
||
|
||
def fetch_all_data(start_date='2018-01-01', end_date='2026-05-19'):
|
||
ts = TushareSource()
|
||
api = FlaskAPIDataSource() # 默认使用 k3s.tokenpluse.xyz
|
||
|
||
index_data = {}
|
||
etf_data = {}
|
||
etf_code_map = {}
|
||
|
||
# 统一使用 Flask API 获取所有指数数据(与 strategy.py 保持一致)
|
||
print("[指数数据] - 通过 Flask API (k3s服务) 获取")
|
||
index_codes = list(CODE_LIST.keys())
|
||
index_ohlcv_data = api.fetch_batch(index_codes, start_date, end_date)
|
||
|
||
for code, df in index_ohlcv_data.items():
|
||
if df is not None and 'close' in df.columns and len(df) > 0:
|
||
index_data[code] = df
|
||
name = CODE_LIST.get(code, {}).get('name', code)
|
||
print(f" {code} ({name})... {len(df)}天")
|
||
else:
|
||
name = CODE_LIST.get(code, {}).get('name', code)
|
||
print(f" {code} ({name})... 失败")
|
||
|
||
print("\n[ETF数据]")
|
||
etf_nav_data = {}
|
||
for code, cfg in CODE_LIST.items():
|
||
etf_code = cfg.get('etf')
|
||
if etf_code is None:
|
||
continue
|
||
etf_code_map[code] = etf_code
|
||
name = cfg['name']
|
||
print(f" {etf_code} ({name})...", end=' ')
|
||
|
||
df = ts.fetch_etf_adj(etf_code, start_date, end_date)
|
||
if df is not None and 'close_hfq' in df.columns and len(df) > 0:
|
||
adj_ratio = df['close_hfq'] / df['close']
|
||
df['open_hfq'] = df['open'] * adj_ratio
|
||
etf_data[code] = df
|
||
print(f"{len(df)}天", end='')
|
||
else:
|
||
print("失败")
|
||
continue
|
||
|
||
# 获取ETF净值(用于计算溢价率)
|
||
nav_df = ts.fetch_etf_nav(etf_code, start_date, end_date)
|
||
if nav_df is not None and 'nav' in nav_df.columns and len(nav_df) > 0:
|
||
etf_nav_data[code] = nav_df['nav']
|
||
print(f" nav={len(nav_df)}天")
|
||
else:
|
||
print(" nav=无")
|
||
|
||
return index_data, etf_data, etf_code_map, etf_nav_data
|
||
|
||
|
||
def compute_factors(price_data, n_days, trade_dates):
|
||
"""先在原始交易日历上计算因子,再 ffill 对齐到 A 股日历(与 strategy.py 一致)"""
|
||
factor = MomentumFactor(n_days=n_days, weighted=True, crash_filter=True)
|
||
factor_values = {}
|
||
for code, df in price_data.items():
|
||
if 'close' not in df.columns:
|
||
continue
|
||
close_series = df['close'].dropna()
|
||
if len(close_series) == 0:
|
||
continue
|
||
values = factor.compute(pd.DataFrame({'close': close_series}))
|
||
factor_values[code] = values.reindex(trade_dates, method='ffill')
|
||
return pd.DataFrame(factor_values)
|
||
|
||
|
||
def generate_signals(factor_df, group_mapping):
|
||
selector = TopNSelector(
|
||
select_num=SELECT_NUM,
|
||
group_mapping=group_mapping,
|
||
min_score=0.0,
|
||
rebalance_days=1,
|
||
rebalance_threshold=0.0,
|
||
bond_threshold_config=BOND_THRESHOLD
|
||
)
|
||
return selector.generate(factor_df)
|
||
|
||
|
||
def safe_val(v, decimals=4):
|
||
if v is None or (isinstance(v, float) and (math.isnan(v) or math.isinf(v))):
|
||
return None
|
||
if isinstance(v, (np.floating, float)):
|
||
return round(float(v), decimals)
|
||
if isinstance(v, (np.integer, int)):
|
||
return int(v)
|
||
return v
|
||
|
||
|
||
def main():
|
||
from datetime import datetime
|
||
backtest_start = '2020-01-01'
|
||
backtest_end = datetime.now().strftime('%Y-%m-%d') # 动态获取当前日期
|
||
|
||
print("=" * 60)
|
||
print(" 导出回测逐日明细 (模式B: 指数信号 + ETF收益)")
|
||
print("=" * 60)
|
||
|
||
# 1. 获取数据
|
||
print("\n[1] 获取数据...")
|
||
index_data, etf_data, etf_code_map, etf_nav_data = fetch_all_data()
|
||
|
||
# 2. A股交易日历
|
||
print("\n[2] 获取A股交易日历...")
|
||
ts = TushareSource()
|
||
a_share_dates = ts.fetch_trade_cal(backtest_start, backtest_end)
|
||
print(f" {len(a_share_dates)} 天")
|
||
|
||
# 3. 分组映射
|
||
group_mapping = {}
|
||
for code, cfg in CODE_LIST.items():
|
||
if isinstance(cfg, dict):
|
||
group_mapping[code] = cfg.get('market', 'default')
|
||
|
||
valid_codes = [c for c in CODE_LIST if c in index_data]
|
||
|
||
# 4. 计算因子(指数信号)
|
||
print("\n[3] 计算指数动量因子...")
|
||
idx_price_data = {}
|
||
for code in valid_codes:
|
||
if code in index_data and 'close' in index_data[code].columns:
|
||
idx_price_data[code] = index_data[code]
|
||
factor_df = compute_factors(idx_price_data, N_DAYS, a_share_dates)
|
||
print(f" {len(factor_df.columns)} 只, {len(factor_df)} 天")
|
||
|
||
# 5. 生成信号
|
||
print("\n[4] 生成信号...")
|
||
signals = generate_signals(factor_df, group_mapping)
|
||
print(f" {len(signals)} 天")
|
||
|
||
# 6. 准备ETF收益率(模式B)
|
||
print("\n[5] 准备ETF收益率...")
|
||
etf_close_hfq_aligned = {}
|
||
etf_close_aligned = {}
|
||
etf_open_aligned = {}
|
||
etf_close_hfq_raw = {}
|
||
index_close_aligned = {}
|
||
returns_etf = {}
|
||
returns_idx = {}
|
||
|
||
for code in valid_codes:
|
||
# 指数收盘价和收益率
|
||
if code in index_data and 'close' in index_data[code].columns:
|
||
ic = index_data[code]['close'].dropna()
|
||
ic_a = ic.reindex(a_share_dates, method='ffill')
|
||
index_close_aligned[code] = ic_a
|
||
returns_idx[code] = ic_a.pct_change(fill_method=None)
|
||
|
||
# ETF价格和收益率
|
||
etf_code = etf_code_map.get(code)
|
||
if etf_code and code in etf_data:
|
||
df = etf_data[code]
|
||
chfq = df['close_hfq'].dropna()
|
||
chfq_a = chfq.reindex(a_share_dates, method='ffill')
|
||
etf_close_hfq_aligned[code] = chfq_a
|
||
etf_close_hfq_raw[code] = chfq
|
||
returns_etf[f'日收益率_{code}'] = chfq_a.pct_change(fill_method=None)
|
||
|
||
ec = df['close'].reindex(a_share_dates, method='ffill')
|
||
etf_close_aligned[code] = ec
|
||
eo = df['open'].reindex(a_share_dates, method='ffill')
|
||
etf_open_aligned[code] = eo
|
||
elif code in index_data and 'close' in index_data[code].columns:
|
||
ic = index_data[code]['close'].dropna()
|
||
ic_a = ic.reindex(a_share_dates, method='ffill')
|
||
returns_etf[f'日收益率_{code}'] = ic_a.pct_change(fill_method=None)
|
||
|
||
returns_etf_df = pd.DataFrame(returns_etf)
|
||
|
||
# 6.5 溢价率:(ETF收盘价 - 单位净值) / 单位净值
|
||
etf_premium_aligned = {}
|
||
for code in valid_codes:
|
||
if code in etf_nav_data and code in etf_close_aligned:
|
||
nav_raw = etf_nav_data[code]
|
||
nav_raw = nav_raw[~nav_raw.index.duplicated(keep='last')]
|
||
nav = nav_raw.reindex(a_share_dates, method='ffill')
|
||
close = etf_close_aligned[code]
|
||
premium = (close - nav) / nav
|
||
etf_premium_aligned[code] = premium
|
||
|
||
# 7. 执行回测获取净值
|
||
print("\n[6] 执行回测...")
|
||
common_dates = signals.index.intersection(returns_etf_df.index)
|
||
signals_aligned = signals.loc[common_dates]
|
||
returns_aligned = returns_etf_df.loc[common_dates]
|
||
|
||
executor = BacktestExecutor(
|
||
initial_capital=100000,
|
||
trade_cost=TRADE_COST,
|
||
select_num=SELECT_NUM
|
||
)
|
||
portfolio = executor.execute(signals_aligned, returns_aligned)
|
||
result = portfolio.backtest_result
|
||
nav_series_raw = result['策略净值']
|
||
daily_ret_raw = result['策略日收益率']
|
||
|
||
# 扩展到所有common_dates,信号前的日期 nav=1.0, return=0.0
|
||
nav_series = nav_series_raw.reindex(common_dates)
|
||
daily_ret_series = daily_ret_raw.reindex(common_dates, fill_value=0.0)
|
||
first_valid = nav_series.first_valid_index()
|
||
if first_valid is not None:
|
||
nav_series.loc[:first_valid] = nav_series.loc[:first_valid].fillna(1.0)
|
||
nav_series = nav_series.ffill()
|
||
|
||
print(f" 终值: {nav_series.iloc[-1]:.4f}")
|
||
|
||
# 8. 构建逐日明细
|
||
print("\n[7] 构建逐日明细...")
|
||
|
||
# 持仓跟踪状态
|
||
holdings_state = {} # {code: {'entry_date': str, 'entry_price': float}}
|
||
prev_holdings = set()
|
||
|
||
days_list = []
|
||
signal_col = 'signal'
|
||
|
||
for i, date in enumerate(common_dates):
|
||
sig_val = signals_aligned.loc[date, signal_col] if signal_col in signals_aligned.columns else ''
|
||
current_holdings = set(str(sig_val).split(',')) if pd.notna(sig_val) and sig_val else set()
|
||
current_holdings.discard('')
|
||
|
||
# 调仓检测
|
||
added = list(current_holdings - prev_holdings)
|
||
removed = list(prev_holdings - current_holdings)
|
||
is_rebalance = len(added) > 0 or len(removed) > 0
|
||
|
||
# 更新持仓状态
|
||
for code in removed:
|
||
holdings_state.pop(code, None)
|
||
for code in added:
|
||
entry_price_etf = None
|
||
entry_price_idx = None
|
||
if code in etf_close_hfq_aligned:
|
||
ep = etf_close_hfq_aligned[code].get(date)
|
||
if pd.notna(ep):
|
||
entry_price_etf = float(ep)
|
||
if code in index_close_aligned:
|
||
ep = index_close_aligned[code].get(date)
|
||
if pd.notna(ep):
|
||
entry_price_idx = float(ep)
|
||
holdings_state[code] = {
|
||
'entry_date': date.strftime('%Y-%m-%d'),
|
||
'entry_price_etf': entry_price_etf,
|
||
'entry_price_idx': entry_price_idx,
|
||
}
|
||
|
||
# 动态阈值
|
||
factor_scores = {}
|
||
for code in valid_codes:
|
||
if code in factor_df.columns:
|
||
v = factor_df.loc[date, code] if date in factor_df.index else np.nan
|
||
if pd.notna(v):
|
||
factor_scores[code] = float(v)
|
||
|
||
bond_score = factor_scores.get(BOND_CODE)
|
||
if BOND_THRESHOLD.get('enabled') and bond_score is not None and bond_score >= 0:
|
||
threshold = bond_score * BOND_RATIO
|
||
else:
|
||
threshold = 0.0
|
||
|
||
# 排名(按动量降序,排除BOND)
|
||
non_bond_scores = {k: v for k, v in factor_scores.items()
|
||
if group_mapping.get(k) != 'BOND'}
|
||
sorted_codes = sorted(non_bond_scores.keys(),
|
||
key=lambda c: non_bond_scores[c], reverse=True)
|
||
rank_map = {c: r + 1 for r, c in enumerate(sorted_codes)}
|
||
# BOND不参与排名
|
||
if BOND_CODE in factor_scores:
|
||
rank_map[BOND_CODE] = None
|
||
|
||
# 每标的详情
|
||
assets = {}
|
||
for code in valid_codes:
|
||
asset = {}
|
||
|
||
# 指数收盘价
|
||
if code in index_close_aligned:
|
||
v = index_close_aligned[code].get(date)
|
||
asset['index_close'] = safe_val(v, 2)
|
||
else:
|
||
asset['index_close'] = None
|
||
|
||
# 动量
|
||
mom = factor_scores.get(code)
|
||
asset['momentum'] = safe_val(mom, 4)
|
||
|
||
# 排名
|
||
asset['rank'] = rank_map.get(code)
|
||
|
||
# 阈值
|
||
asset['threshold'] = safe_val(threshold, 4)
|
||
asset['above_threshold'] = mom >= threshold if mom is not None else False
|
||
|
||
# ETF价格
|
||
if code in etf_close_aligned:
|
||
asset['etf_close'] = safe_val(etf_close_aligned[code].get(date), 3)
|
||
else:
|
||
asset['etf_close'] = None
|
||
|
||
if code in etf_open_aligned:
|
||
asset['etf_open'] = safe_val(etf_open_aligned[code].get(date), 3)
|
||
else:
|
||
asset['etf_open'] = None
|
||
|
||
if code in etf_close_hfq_aligned:
|
||
asset['etf_close_hfq'] = safe_val(etf_close_hfq_aligned[code].get(date), 4)
|
||
else:
|
||
asset['etf_close_hfq'] = None
|
||
|
||
# 溢价率
|
||
if code in etf_premium_aligned:
|
||
asset['premium'] = safe_val(etf_premium_aligned[code].get(date), 4)
|
||
else:
|
||
asset['premium'] = None
|
||
|
||
# ETF日收益率
|
||
ret_col = f'日收益率_{code}'
|
||
if ret_col in returns_etf_df.columns:
|
||
asset['etf_return_ctc'] = safe_val(returns_etf_df.loc[date, ret_col], 6)
|
||
else:
|
||
asset['etf_return_ctc'] = None
|
||
|
||
# 指数日收益率
|
||
if code in returns_idx:
|
||
asset['index_return'] = safe_val(returns_idx[code].get(date), 6)
|
||
else:
|
||
asset['index_return'] = None
|
||
|
||
# 持仓状态
|
||
is_held = code in current_holdings
|
||
asset['is_held'] = is_held
|
||
|
||
if is_held and code in holdings_state:
|
||
hs = holdings_state[code]
|
||
asset['entry_date'] = hs['entry_date']
|
||
asset['entry_price_etf'] = safe_val(hs['entry_price_etf'], 4)
|
||
asset['entry_price_idx'] = safe_val(hs['entry_price_idx'], 4)
|
||
|
||
entry_dt = pd.Timestamp(hs['entry_date'])
|
||
trading_days_held = len(common_dates[(common_dates >= entry_dt) & (common_dates <= date)])
|
||
asset['holding_days'] = trading_days_held
|
||
|
||
# ETF累计收益
|
||
if hs['entry_price_etf'] and hs['entry_price_etf'] > 0:
|
||
cur = etf_close_hfq_aligned[code].get(date) if code in etf_close_hfq_aligned else None
|
||
if cur and pd.notna(cur):
|
||
asset['cum_return_etf'] = safe_val(float(cur) / hs['entry_price_etf'] - 1, 4)
|
||
else:
|
||
asset['cum_return_etf'] = None
|
||
else:
|
||
asset['cum_return_etf'] = None
|
||
|
||
# 指数累计收益
|
||
if hs['entry_price_idx'] and hs['entry_price_idx'] > 0:
|
||
cur = index_close_aligned[code].get(date) if code in index_close_aligned else None
|
||
if cur and pd.notna(cur):
|
||
asset['cum_return_idx'] = safe_val(float(cur) / hs['entry_price_idx'] - 1, 4)
|
||
else:
|
||
asset['cum_return_idx'] = None
|
||
else:
|
||
asset['cum_return_idx'] = None
|
||
else:
|
||
asset['entry_date'] = None
|
||
asset['entry_price_etf'] = None
|
||
asset['entry_price_idx'] = None
|
||
asset['holding_days'] = 0
|
||
asset['cum_return_etf'] = None
|
||
asset['cum_return_idx'] = None
|
||
|
||
assets[code] = asset
|
||
|
||
# 构建当天记录
|
||
nav_val = nav_series.loc[date] if date in nav_series.index else None
|
||
ret_val = daily_ret_series.loc[date] if date in daily_ret_series.index else None
|
||
|
||
day_record = {
|
||
'date': date.strftime('%Y-%m-%d'),
|
||
'nav': safe_val(nav_val, 4),
|
||
'daily_return': safe_val(ret_val, 6),
|
||
'is_rebalance': is_rebalance,
|
||
'holdings': sorted(list(current_holdings)),
|
||
'added': sorted(added),
|
||
'removed': sorted(removed),
|
||
'assets': assets
|
||
}
|
||
days_list.append(day_record)
|
||
prev_holdings = current_holdings
|
||
|
||
# 9. 构建元数据
|
||
codes_meta = {}
|
||
for code, cfg in CODE_LIST.items():
|
||
codes_meta[code] = {
|
||
'name': cfg['name'],
|
||
'etf': cfg.get('etf'),
|
||
'market': cfg.get('market')
|
||
}
|
||
|
||
output = {
|
||
'meta': {
|
||
'mode': 'B: 指数信号 + ETF收益',
|
||
'start_date': common_dates[0].strftime('%Y-%m-%d'),
|
||
'end_date': common_dates[-1].strftime('%Y-%m-%d'),
|
||
'total_days': len(common_dates),
|
||
'select_num': SELECT_NUM,
|
||
'n_days': N_DAYS,
|
||
'trade_cost': TRADE_COST,
|
||
'bond_threshold': {
|
||
'enabled': BOND_THRESHOLD.get('enabled', False),
|
||
'bond_code': BOND_CODE,
|
||
'ratio': BOND_RATIO
|
||
},
|
||
'codes': codes_meta
|
||
},
|
||
'days': days_list
|
||
}
|
||
|
||
# 10. 输出
|
||
output_path = project_root / 'results' / 'backtest_detail.json'
|
||
print(f"\n[8] 写入 {output_path}...")
|
||
with open(output_path, 'w', encoding='utf-8') as f:
|
||
json.dump(output, f, ensure_ascii=False)
|
||
|
||
file_size_mb = output_path.stat().st_size / 1024 / 1024
|
||
print(f" 大小: {file_size_mb:.1f} MB")
|
||
print(f" 天数: {len(days_list)}")
|
||
print(f" 标的: {len(valid_codes)}")
|
||
print(" 完成!")
|
||
|
||
|
||
if __name__ == '__main__':
|
||
main()
|