你该不该止损？

交易者必须自律，这是一个身为投资者/交易者时常听见的话
每笔交易都要有进场出场与止损的计划
止损是一种风险管理工具，目的让你能结束交易来防止更多的亏损

今天我们就来研究看看止损到底是不是有效的方法
至于止损该设多少，每个交易者都不同

常见的有5% 10%，这也得考虑你的风险偏好

数据读取

我从Yahoo Finance上下载了KLSE指数的历史数据，

要分析数据得先读取这个csv文件。

import pandas as pd

KLSE = pd.read_csv('KLSE.csv')
KLSE = KLSE.replace(',','', regex=True)
KLSE['Open'] = KLSE['Open'].astype(float)
KLSE['Close'] = KLSE['Close'].astype(float)
KLSE['Change'] = (KLSE['Close'] - KLSE['Close'].shift(1))/KLSE['Open'] + 1

KLSE.head()

	Date	Open	High	Low	Close	Adj Close	Volume	Change
0	4-Jan-10	1272.31	1275.75	1272.25	1275.75	1275.75	56508200	NaN
1	5-Jan-10	1278.26	1290.55	1278.26	1288.24	1288.24	136646600	1.009771
2	6-Jan-10	1288.86	1296.44	1288.02	1293.17	1293.17	117740300	1.003825
3	7-Jan-10	1293.69	1299.70	1290.36	1291.42	1291.42	115024400	0.998647
4	8-Jan-10	1294.93	1295.51	1290.86	1292.98	1292.98	74587200	1.001205

策略规则

我们以最简单的双均线策略 + 5% 止损进行回测，
具体规则：

1. 快线穿过慢线时买入
2. 快线跌破慢线时卖出
3. 账面亏损大于 5% 就卖出
4. 不考虑交易成本

def simulate(df,fast,slow,cutloss = False):
    
    import talib

    df['fast'] = talib.SMA(df['Close'],fast)
    df['slow'] = talib.SMA(df['Close'],slow)
    
    df.dropna(inplace = True)

    gold_cross = df[df['fast'] > df['slow']].index
    df.loc[gold_cross,'Cross'] = 1

    gold_cross = df[df['fast'] < df['slow']].index
    df.loc[gold_cross,'Cross'] = 0

    df['Buy'] = df['Cross'].diff()

    df['Return'] = df['Cross']*df['Change']

    def norm(x):
        if x == 0:
            return 1
        else:
            return x

    df['Return'] = df['Return'].apply(lambda x: norm(x))
    df['Nav'] = (df['Return']).cumprod()

    DD = 1 - df['Nav']/df['Nav'].cummax()
    
    #===================================
    if cutloss:
        # set cutloss to 5.5% for extra buffer
        c = DD[DD > 0.055].index
        df.loc[c,'Cross'] = 0

        df['Buy'] = df['Cross'].diff()

        df['Return'] = df['Cross']*df['Change']

        df['Return'] = df['Return'].apply(lambda x: norm(x))
        df['Nav'] = (df['Return']).cumprod()

    #====================================
    
    # divide by 252 because generally a year has 252 trading days
    num_periods = df.shape[0]/252
    rety = ((df['Nav'].iloc[-1] / df['Nav'].iloc[0]) ** (1 / (num_periods - 1)) - 1)*100.0
    
    price_in = df.loc[df['Buy'] == 1,'Close'].values
    price_out = df.loc[df['Buy'] == -1,'Close'].values
    
    if len(price_out) > len(price_in):
        price_out = price_out[:len(price_in)]

    if len(price_in) > len(price_out):
        price_in = price_in[:len(price_out)]

    VictoryRatio = ((price_out - price_in)>0).mean()*100.0
    DD = 1 - df['Nav']/df['Nav'].cummax()
    MDD = max(DD)*100.0
    
    return df, round(rety, 2), round(VictoryRatio, 2), round(MDD,2)

策略表现

我们测试4个组合的策略，

1. 10天线 和 20天线 (短期交易)，无止损
2. 20天线 和 50天线 (中期交易)，无止损
3. 10天线 和 20天线 (短期交易)，5%止损
4. 20天线 和 50天线 (中期交易)，5%止损

MA1020,cagr1020,vr1020,mdd1020 = simulate(KLSE.copy(),10,20,False)
MA2050,cagr2050,vr2050,mdd2050 = simulate(KLSE.copy(),20,50,False)
MA1020_cutloss,cagr1020_cutloss,vr1020_cutloss,mdd1020_cutloss = simulate(KLSE.copy(),10,20,True)
MA2050_cutloss,cagr2050_cutloss,vr2050_cutloss,mdd2050_cutloss = simulate(KLSE.copy(),20,50,True)

KLSE['KLSE'] = (KLSE['Change']).cumprod()
import matplotlib.pyplot as plt
plt.style.use('seaborn')

ax = MA1020['Nav'].plot(figsize=(10, 6))
MA2050['Nav'].plot(ax=ax)
MA1020_cutloss['Nav'].plot(ax=ax)
MA2050_cutloss['Nav'].plot(ax=ax)

KLSE['KLSE'].plot(ax=ax)
# plt.plot( 'Date','Nav', data = MA2050, marker='', color='olive', linewidth=2)
ax.legend(['MA1020','MA2050','MA1020 5% Cutloss','MA2050 5% Cutloss','KLSE']);
plt.show()


from prettytable import PrettyTable

t = PrettyTable(['Strategy', 'CAGR', 'Win Rate', 'Max Drawdown'])
t.add_row(['MA1020', cagr1020,vr1020,mdd1020])
t.add_row(['MA2050', cagr2050,vr2050,mdd2050])
t.add_row(['MA1020 5% Cutloss', cagr1020_cutloss,vr1020_cutloss,mdd1020_cutloss])
t.add_row(['MA2050 5% Cutloss', cagr2050_cutloss,vr2050_cutloss,mdd2050_cutloss])

print(t)

+-------------------+------+----------+--------------+
|      Strategy     | CAGR | Win Rate | Max Drawdown |
+-------------------+------+----------+--------------+
|       MA1020      | 6.44 |  50.77   |     8.52     |
|       MA2050      | 3.14 |  30.77   |    12.15     |
| MA1020 5% Cutloss | 7.19 |  45.21   |     5.44     |
| MA2050 5% Cutloss | 4.54 |  20.59   |     5.46     |
+-------------------+------+----------+--------------+

想法

从回测结果看来，
止损看起来是有效的风险管理技巧

设下5%止损的策略都比无止损的策略来的好
虽然赢率(Win Rate)较低

这可能是有止损的会更常认赔出局，这也可以说是好事
想要自己测试的读者可以从Yahoo Finance下载想要的个股数据来分析看看。

纯属分享，无买卖建议