Backtesting.py 是 Python 下的一个回测框架，但是你不用需要在他的类里实现什么指标。基本上你只需要加载历史数据，然后转换成 pandas 的 dataframe，然后用 pandas_ta 这个库计算指标和信号，然后再在 Backtesting.py 的策略类实现信号买卖就可以了。

螺纹钢2405 5 分钟图的简单回测

弄一个简单的 supertrend 策略，添加了止盈和止损。

• 最大收益 108%
• 最终收益 108%
• 最大回撤 -8.4%
• 历时 89 天
• 21 笔交易，包括多空
• 绿色是成功的交易段
• 红色是失败的交易段

Start   2023-08-17 21:35:00+08:00
End 2024-02-07 11:25:00+08:00
Duration    173 days 13:50:00
Exposure Time [%]   78.8
Equity Final [$]    108185.079963
Equity Peak [$] 108185.079963
Return [%]  8.18508
Buy & Hold Return [%]   4.480874
Return (Ann.) [%]   18.295331
Volatility (Ann.) [%]   12.531943
Sharpe Ratio    1.459896
Sortino Ratio   2.518898
Calmar Ratio    2.182325
Max. Drawdown [%]   -8.383414
Avg. Drawdown [%]   -0.462499
Max. Drawdown Duration  84 days 17:30:00
Avg. Drawdown Duration  2 days 13:01:00
# Trades    21
Win Rate [%]    47.619048
Best Trade [%]  4.325431
Worst Trade [%] -1.802946
Avg. Trade [%]  0.380934
Max. Trade Duration 25 days 12:55:00
Avg. Trade Duration 6 days 12:46:00
Profit Factor   1.797981
Expectancy [%]  0.396803
SQN 0.961494
_strategy   MyStrategy
_equity_curve   Equity Draw...
_trades Size EntryBar ExitBar EntryPrice Exi...

from backtesting import Backtest, Strategy
from backtesting.lib import crossover

def ind_supertrend(data):
    supert = ta.supertrend(
        data.High.s, data.Low.s, data.Close.s, length=10, multiplier=10
    )["SUPERT_10_10.0"]
    return supert

class MyStrategy(Strategy):
    def init(self):
        self.supert = self.I(ind_supertrend, self.data)

    def next(self):
        price = self.data.Close
        atr = self.data.ATRr_14

        if crossover(price, self.supert):
            self.position.close()
            sl = price - atr * 10
            tp = price + atr * 25
            self.buy(sl = sl, tp = tp)
        elif crossover(self.supert, price):
            self.position.close()
            sl = price + atr * 10
            tp = price - atr * 25
            self.sell(sl = sl, tp = tp)

bt = Backtest(
    df_new,             # df_new 是要加载 OHLCV 数据及指标数据
    MyStrategy,
    cash=100_000,
    commission=0,
    hedging=True,       # 可做空
    #exclusive_orders=True,
    trade_on_close=True,
)
stats = bt.run()
bt.plot()

stats.to_frame()

pd.options.display.max_columns = None
pd.options.display.max_rows = None

stats._trades

这个是按着涨跌比例来赚钱，不是商品期货的涨跌点，也无杠杆，所以只需要检验策略的胜率和回撤即可。

参考

• https://github.com/kernc/backtesting.py/blob/master/doc/examples/Quick%20Start%20User%20Guide.ipynb
• https://greyhoundanalytics.com/blog/custom-indicators-in-backtestingpy/
• https://www.youtube.com/watch?v=k6i903uNZ6E
• https://github.com/cgohlke/talib-build/releases
• https://kernc.github.io/backtesting.py/doc/backtesting/backtesting.html#backtesting.backtesting.Trade
• https://github.com/zunda-lab/fx_backtest/blob/main/src/unit_backtesting.py#L67
• https://qiita.com/TKfumi/items/4744f34a10863f0d3cfb
• https://qiita.com/TKfumi/items/ffb617898211a794bf9e
• https://qiita.com/TKfumi/items/17301ca962a0da9fcbce
• https://github.com/mpquant/MyTT
• https://gitee.com/wkingnet/stock-analysis/blob/master/func_TDX.py