bd7d85817a
三版本 A/B/C 止损策略对比回测 - A: 固定止损 12% - B: ATR x2.5 动态止损 - C: 混合自适应（低波动固定8%/中波动ATR×2.5/高波动ATR×2.0） 含仓位分级、成交量确认、CSV缓存机制 已验证三只港股持仓：01833 / 09886 / 09982 待补全：data/1833.csv 和 data/9886.csv（在外网运行 download_data.py）
269 lines
11 KiB
Python
269 lines
11 KiB
Python
"""
|
||
港股 AI v4 A/B 回测 — 支持本地CSV缓存，绕过yfinance限速
|
||
用法：
|
||
1. 正常运行：python3 stock_backtest_v4_ab.py
|
||
2. 强制重新下载：python3 stock_backtest_v4_ab.py --refresh
|
||
"""
|
||
|
||
import yfinance as yf
|
||
import pandas as pd
|
||
import numpy as np
|
||
import time, os, sys
|
||
import warnings
|
||
warnings.filterwarnings('ignore')
|
||
|
||
CACHE_DIR = "data"
|
||
os.makedirs(CACHE_DIR, exist_ok=True)
|
||
FORCE_REFRESH = "--refresh" in sys.argv
|
||
|
||
STOCKS = {
|
||
"平安好医生": "1833.HK",
|
||
"叮当健康": "9886.HK",
|
||
"中原建业": "9982.HK",
|
||
}
|
||
PERIOD = "2y"
|
||
INITIAL_CAPITAL = 10000.0
|
||
W_TECH, W_FUNDAMENTAL, W_SENTIMENT = 0.50, 0.30, 0.20
|
||
BUY_THRESH, SELL_THRESH = 1.5, -1.5
|
||
|
||
# A: 固定止损
|
||
A_TRAILING_STOP = 0.12
|
||
A_VOL_CONFIRM = 1.2
|
||
|
||
# B: ATR动态止损
|
||
B_ATR_MULT = 2.5
|
||
B_ATR_PERIOD = 14
|
||
B_VOL_CONFIRM = 1.2
|
||
B_MIN_STOP = 0.08
|
||
B_MAX_STOP = 0.35
|
||
|
||
def position_ratio(score):
|
||
if score >= 5: return 1.0
|
||
elif score >= 3: return 0.6
|
||
return 0.3
|
||
|
||
FUNDAMENTAL_TIMELINE = {
|
||
"平安好医生": [
|
||
{"from": "2024-01-01", "score": -3.0},
|
||
{"from": "2024-08-01", "score": -1.0},
|
||
{"from": "2025-01-01", "score": 0.0},
|
||
{"from": "2025-08-01", "score": 1.0},
|
||
],
|
||
"叮当健康": [
|
||
{"from": "2024-01-01", "score": -3.0},
|
||
{"from": "2024-06-01", "score": -2.0},
|
||
{"from": "2025-01-01", "score": -1.0},
|
||
{"from": "2025-09-01", "score": 1.0},
|
||
],
|
||
"中原建业": [
|
||
{"from": "2024-01-01", "score": -3.0},
|
||
{"from": "2024-06-01", "score": -4.0},
|
||
{"from": "2025-01-01", "score": -4.0},
|
||
{"from": "2025-10-01", "score": -5.0},
|
||
],
|
||
}
|
||
SENTIMENT_TIMELINE = {
|
||
"平安好医生": [
|
||
{"from": "2024-01-01", "score": -1.0},
|
||
{"from": "2024-10-01", "score": 1.0},
|
||
{"from": "2025-01-01", "score": 2.0},
|
||
{"from": "2026-01-01", "score": 3.0},
|
||
],
|
||
"叮当健康": [
|
||
{"from": "2024-01-01", "score": -2.0},
|
||
{"from": "2024-08-01", "score": -1.0},
|
||
{"from": "2025-04-01", "score": 1.0},
|
||
{"from": "2025-10-01", "score": 2.0},
|
||
],
|
||
"中原建业": [
|
||
{"from": "2024-01-01", "score": -2.0},
|
||
{"from": "2024-06-01", "score": -3.0},
|
||
{"from": "2025-01-01", "score": -3.0},
|
||
{"from": "2025-10-01", "score": -4.0},
|
||
],
|
||
}
|
||
|
||
def get_snapshot(tl, date):
|
||
score = tl[0]["score"]
|
||
for e in tl:
|
||
if str(date.date()) >= e["from"]: score = e["score"]
|
||
else: break
|
||
return score
|
||
|
||
def calc_rsi(s, p=14):
|
||
d = s.diff()
|
||
g = d.clip(lower=0).ewm(com=p-1, min_periods=p).mean()
|
||
l = (-d.clip(upper=0)).ewm(com=p-1, min_periods=p).mean()
|
||
return 100 - 100 / (1 + g / l)
|
||
|
||
def calc_macd(s, fast=12, slow=26, sig=9):
|
||
m = s.ewm(span=fast, adjust=False).mean() - s.ewm(span=slow, adjust=False).mean()
|
||
return m - m.ewm(span=sig, adjust=False).mean()
|
||
|
||
def calc_atr(df, period=14):
|
||
hi, lo, cl = df["High"], df["Low"], df["Close"]
|
||
tr = pd.concat([(hi-lo), (hi-cl.shift(1)).abs(), (lo-cl.shift(1)).abs()], axis=1).max(axis=1)
|
||
return tr.ewm(com=period-1, min_periods=period).mean()
|
||
|
||
def score_tech(row):
|
||
s = 0
|
||
if row.RSI < 30: s += 3
|
||
elif row.RSI < 45: s += 1
|
||
elif row.RSI > 70: s -= 3
|
||
elif row.RSI > 55: s -= 1
|
||
if row.MACD_h > 0 and row.MACD_h_p <= 0: s += 3
|
||
elif row.MACD_h < 0 and row.MACD_h_p >= 0: s -= 3
|
||
elif row.MACD_h > 0: s += 1
|
||
else: s -= 1
|
||
if row.MA5 > row.MA20 > row.MA60: s += 2
|
||
elif row.MA5 < row.MA20 < row.MA60: s -= 2
|
||
if row.Close > row.MA20 and row.Close_p <= row.MA20_p: s += 1
|
||
elif row.Close < row.MA20 and row.Close_p >= row.MA20_p: s -= 1
|
||
return float(np.clip(s, -10, 10))
|
||
|
||
def load_data(ticker):
|
||
"""优先读CSV缓存，否则从yfinance下载并缓存"""
|
||
sym = ticker.replace(".HK", "")
|
||
fp = os.path.join(CACHE_DIR, f"{sym}.csv")
|
||
if os.path.exists(fp) and not FORCE_REFRESH:
|
||
df = pd.read_csv(fp, index_col=0, parse_dates=True)
|
||
print(f" 📂 读取缓存: {fp} ({len(df)} 行)")
|
||
return df
|
||
print(f" 🌐 下载数据: {ticker}")
|
||
df = yf.download(ticker, period=PERIOD, auto_adjust=True, progress=False)
|
||
if df.empty: return None
|
||
if isinstance(df.columns, pd.MultiIndex): df.columns = df.columns.droplevel(1)
|
||
df.to_csv(fp)
|
||
print(f" 💾 已缓存: {fp}")
|
||
return df
|
||
|
||
def prepare_df(ticker):
|
||
df = load_data(ticker)
|
||
if df is None or len(df) < 60: return None
|
||
c = df["Close"]
|
||
df["RSI"] = calc_rsi(c)
|
||
h = calc_macd(c)
|
||
df["MACD_h"] = h
|
||
df["MACD_h_p"] = h.shift(1)
|
||
for p in [5, 20, 60]: df[f"MA{p}"] = c.rolling(p).mean()
|
||
df["MA20_p"] = df["MA20"].shift(1)
|
||
df["Close_p"] = c.shift(1)
|
||
df["Vol20"] = df["Volume"].rolling(20).mean()
|
||
df["ATR"] = calc_atr(df, B_ATR_PERIOD)
|
||
return df.dropna()
|
||
|
||
def simulate(name, df, use_atr=False):
|
||
capital, position, entry_price = INITIAL_CAPITAL, 0, 0.0
|
||
highest_price, trailing_pct = 0.0, 0.0
|
||
trades = []
|
||
vc = A_VOL_CONFIRM if not use_atr else B_VOL_CONFIRM
|
||
|
||
for date, row in df.iterrows():
|
||
f = get_snapshot(FUNDAMENTAL_TIMELINE[name], date)
|
||
s = get_snapshot(SENTIMENT_TIMELINE[name], date)
|
||
t = score_tech(row)
|
||
score = W_TECH*t + W_FUNDAMENTAL*f + W_SENTIMENT*s
|
||
price = float(row["Close"])
|
||
vol = float(row["Volume"])
|
||
vol20 = float(row["Vol20"])
|
||
|
||
if score >= BUY_THRESH and position == 0 and capital > price:
|
||
if vol >= vol20 * vc:
|
||
ratio = position_ratio(score)
|
||
shares = int(capital * ratio / price)
|
||
if shares > 0:
|
||
position, entry_price, highest_price = shares, price, price
|
||
capital -= shares * price
|
||
if use_atr:
|
||
raw = float(row["ATR"]) * B_ATR_MULT / price
|
||
trailing_pct = float(np.clip(raw, B_MIN_STOP, B_MAX_STOP))
|
||
note = f"仓{ratio*100:.0f}% 量比{vol/vol20:.1f}x ATR止损{trailing_pct*100:.1f}%"
|
||
else:
|
||
trailing_pct = A_TRAILING_STOP
|
||
note = f"仓{ratio*100:.0f}% 量比{vol/vol20:.1f}x 固定止损{trailing_pct*100:.0f}%"
|
||
trades.append({"操作":"买入","日期":date.date(),"价格":round(price,4),
|
||
"股数":shares,"评分":round(score,2),"备注":note})
|
||
|
||
elif position > 0:
|
||
highest_price = max(highest_price, price)
|
||
stop_price = highest_price * (1 - trailing_pct)
|
||
if price <= stop_price or score <= SELL_THRESH:
|
||
pnl = position * (price - entry_price)
|
||
pct = pnl / (position * entry_price) * 100
|
||
reason = (f"移动止损 高点{highest_price:.3f}→止损{stop_price:.3f}"
|
||
if price <= stop_price else f"评分卖出({score:.1f})")
|
||
capital += position * price
|
||
trades.append({"操作":"卖出","日期":date.date(),"价格":round(price,4),
|
||
"股数":position,"评分":round(score,2),
|
||
"盈亏%":f"{pct:+.1f}%","备注":reason})
|
||
position, highest_price, trailing_pct = 0, 0.0, 0.0
|
||
|
||
last = float(df["Close"].iloc[-1])
|
||
total = capital + position * last
|
||
if position > 0:
|
||
pct = (last - entry_price) / entry_price * 100
|
||
trades.append({"操作":"未平仓","日期":"持仓中","价格":round(last,4),
|
||
"股数":position,"评分":"-","盈亏%":f"{pct:+.1f}%","备注":"-"})
|
||
return total, trades
|
||
|
||
def run_ab(name, ticker):
|
||
print(f"\n{'='*70}")
|
||
print(f" {name} ({ticker})")
|
||
print(f"{'='*70}")
|
||
df = prepare_df(ticker)
|
||
if df is None:
|
||
print(" ⚠️ 数据不足，跳过")
|
||
return None
|
||
|
||
avg_atr_pct = df["ATR"].mean() / df["Close"].mean() * 100
|
||
est_stop = np.clip(df["ATR"].mean()*B_ATR_MULT/df["Close"].mean(), B_MIN_STOP, B_MAX_STOP)*100
|
||
bh = (float(df["Close"].iloc[-1]) / float(df["Close"].iloc[0]) - 1) * 100
|
||
print(f" ATR均值波动: {avg_atr_pct:.1f}% → B动态止损估算: {est_stop:.1f}% 买持: {bh:+.1f}%")
|
||
|
||
total_A, tA = simulate(name, df, use_atr=False)
|
||
total_B, tB = simulate(name, df, use_atr=True)
|
||
retA = (total_A - INITIAL_CAPITAL) / INITIAL_CAPITAL * 100
|
||
retB = (total_B - INITIAL_CAPITAL) / INITIAL_CAPITAL * 100
|
||
|
||
for label, trades in [("版本A 固定止损12%", tA), ("版本B ATR动态止损", tB)]:
|
||
print(f"\n 【{label}】")
|
||
if not trades: print(" 无信号"); continue
|
||
cols = [c for c in ["操作","日期","价格","股数","评分","盈亏%","备注"] if c in pd.DataFrame(trades).columns]
|
||
print(pd.DataFrame(trades)[cols].to_string(index=False))
|
||
|
||
nA = len([t for t in tA if t["操作"] in ("买入","卖出")])
|
||
nB = len([t for t in tB if t["操作"] in ("买入","卖出")])
|
||
print(f"\n {'':22} {'A 固定12%':>12} {'B ATR动态':>12} {'买入持有':>10}")
|
||
print(f" {'策略总收益':<22} {retA:>+11.1f}% {retB:>+11.1f}% {bh:>+9.1f}%")
|
||
print(f" {'超额收益α':<22} {retA-bh:>+11.1f}% {retB-bh:>+11.1f}%")
|
||
print(f" {'交易次数':<22} {nA:>12} {nB:>12}")
|
||
w = "B ✅" if retB > retA else ("A ✅" if retA > retB else "平手")
|
||
print(f" {'胜出':<22} {'':>23} → {w} (B-A: {retB-retA:+.1f}%)")
|
||
return {"name":name,"A":retA,"B":retB,"BH":bh,"nA":nA,"nB":nB,"atr":avg_atr_pct}
|
||
|
||
if __name__ == "__main__":
|
||
print("\n🔬 港股 AI v4 A/B 回测 — ATR动态止损 vs 固定止损")
|
||
print(f" A: 固定止损{A_TRAILING_STOP*100:.0f}% | B: ATR×{B_ATR_MULT}动态({B_MIN_STOP*100:.0f}%~{B_MAX_STOP*100:.0f}%)")
|
||
print(f" 仓位分级 评分1.5-3→30% | 3-5→60% | >5→100%\n")
|
||
|
||
results = []
|
||
for i, (name, ticker) in enumerate(STOCKS.items()):
|
||
if i > 0: time.sleep(3)
|
||
r = run_ab(name, ticker)
|
||
if r: results.append(r)
|
||
|
||
if results:
|
||
print(f"\n{'='*70}")
|
||
print(" 📋 A/B 最终汇总")
|
||
print(f"{'='*70}")
|
||
print(f" {'股票':<12} {'ATR%':>6} {'A收益':>9} {'B收益':>9} {'买持':>9} {'B-A':>8} {'胜者':>5}")
|
||
print(f" {'-'*64}")
|
||
for r in results:
|
||
w = "B✅" if r["B"]>r["A"] else ("A✅" if r["A"]>r["B"] else "平")
|
||
print(f" {r['name']:<12} {r['atr']:>5.1f}% {r['A']:>+8.1f}% {r['B']:>+8.1f}% "
|
||
f"{r['BH']:>+8.1f}% {r['B']-r['A']:>+7.1f}% {w:>5}")
|
||
avg_a = np.mean([r["A"] for r in results])
|
||
avg_b = np.mean([r["B"] for r in results])
|
||
avg_bh = np.mean([r["BH"] for r in results])
|
||
print(f" {'平均':<12} {'':>6} {avg_a:>+8.1f}% {avg_b:>+8.1f}% {avg_bh:>+8.1f}% {avg_b-avg_a:>+7.1f}%")
|