本平台基于Jupyter实现，预置数个交易策略及其回测代码，用于课堂教学和实验室研究。

策略示例

因子策略之成交量排序

In [81]:

import pandas as pd
import matplotlib.pyplot as plt
from matplotlib.font_manager import FontProperties
import matplotlib as mpl
import datetime
import numpy as np
import statsmodels.api as sm
import math
import warnings
import seaborn as sns

sns.set()
warnings.filterwarnings('ignore')
plt.rcParams['font.sans-serif']=['SimHei']
plt.rcParams['axes.unicode_minus'] = False
import akshare as ak

In [82]:

import akshare as ak

df = ak.index_zh_a_hist(symbol="000300", period="daily", start_date="20050301", end_date="20210531")[['日期','开盘','收盘','最高','最低','成交量']]
df
df['ma20'] = pd.Series.rolling(df['收盘'],window = 20).mean()

In [83]:

#计算因子，成交量排名因子

import numpy as np
def cal_volumesort(df,n):
    df['paiming'] = np.nan
    for i in range(n,df.shape[0]):
        sorted_volume = sorted(df['成交量'][i-n:i+1])
        df['paiming'][i] = sorted_volume.index(df['成交量'][i])+1
    df['paiming_guiyi'] =(2*df['paiming'] - n -2)/n
    return df
dff = cal_volumesort(df,40)
dff

Out[83]:

	日期	开盘	收盘	最高	最低	成交量	ma20	paiming	paiming_guiyi
0	2005-03-01	1039.35	1035.93	1042.74	1031.17	9962090	NaN	NaN	NaN
1	2005-03-02	1036.37	1021.32	1045.76	1021.00	14225135	NaN	NaN	NaN
2	2005-03-03	1019.88	1027.71	1028.40	1014.75	9203085	NaN	NaN	NaN
3	2005-03-04	1027.93	1023.67	1031.85	1022.18	8173641	NaN	NaN	NaN
4	2005-03-07	1024.48	1029.87	1031.34	1024.48	7932949	NaN	NaN	NaN
…	…	…	…	…	…	…	…	…	…
3948	2021-05-25	5165.04	5318.48	5324.36	5161.52	182996838	5113.5180	41.0	1.00
3949	2021-05-26	5326.12	5320.59	5344.28	5308.74	169607096	5122.7750	38.0	0.85
3950	2021-05-27	5311.37	5338.23	5378.48	5286.05	148462856	5135.8245	29.0	0.40
3951	2021-05-28	5338.73	5321.09	5360.28	5288.65	164605806	5147.3530	37.0	0.80
3952	2021-05-31	5318.08	5331.57	5331.63	5281.69	144603269	5157.9695	26.0	0.25

3953 rows × 9 columns

In [84]:

#策略实现：
# def run_original_strategy(df,N,S):
#     df = df.copy()
#     df = df.copy()
#     df = cal_volumesort(df,N)
#     df['flag'] = 0
#     df['position'] = 0
#     position = 0
#     df = df.dropna().reset_index(drop=True)
#     for i in range(df.shape[0]):
#         if df.loc[i,'paiming_guiyi']>S and position == 0:
#             df.loc[i,'flag']=1
#             df.loc[i+1,'position'] = 1
#             position = 1
#         elif df.loc[i,'paiming_guiyi']<S and position == 1:
#             df.loc[i,'flag']=-1
#             df.loc[i+1,'position'] = 0
#             position = 0
#         else:
#             df['position'][i+1] = df['position'][i] 
#     df['pct_change'] = df['收盘'].pct_change(1)
#     df['net_value'] = (1+(df['pct_change'])*df.position).cumprod()
#     df['benshen'] = (1+df['pct_change']).cumprod()
#     return df.dropna()
# df1 = run_original_strategy(df,40,0.5)
# df1

#添加交易费用
def run_original_strategy(df,N,S):
    df = df.copy()
    df = df.copy()
    df = cal_volumesort(df,N)
    df['flag'] = 0
    df['position'] = 0
    position = 0
    df = df.dropna().reset_index(drop=True)
    

    df['trade_fee'] = 0.0
    fee_rate = 0.001
    for i in range(df.shape[0]):
        if df.loc[i,'paiming_guiyi']>S and position == 0:
            df.loc[i,'flag']=1
            df.loc[i, 'trade_fee'] = df.loc[i, '收盘'] * fee_rate
            df.loc[i+1,'position'] = 1
            position = 1
        elif df.loc[i,'paiming_guiyi']<S and position == 1:
            df.loc[i,'flag']=-1
            df.loc[i, 'trade_fee'] = df.loc[i, '收盘'] * fee_rate
            df.loc[i+1,'position'] = 0
            position = 0
        else:
            df['position'][i+1] = df['position'][i] 
    df['pct_change'] = df['收盘'].pct_change(1)
    df['pct_change_adj'] = df['pct_change'] - df['trade_fee'] / df['收盘']
    df['net_value'] = (1+(df['pct_change'])*df.position).cumprod()
    df['net_value_n'] = (1+(df['pct_change_adj'])*df.position).cumprod()
    df['benshen'] = (1+df['pct_change']).cumprod()
    return df.dropna()
df1 = run_original_strategy(df,40,0.5)
df1

#如果再加入调仓周期呢（比如每月最后一天调仓）

Out[84]:

	日期	开盘	收盘	最高	最低	成交量	ma20	paiming	paiming_guiyi	flag	position	trade_fee	pct_change	pct_change_adj	net_value	net_value_n	benshen
1	2005-04-27	938.57	926.60	938.91	925.90	10780611.0	964.6700	22.0	0.05	0.0	0.0	0.00000	-0.011184	-0.011184	1.000000	1.000000	0.988816
2	2005-04-28	923.53	942.07	945.50	914.83	14343486.0	964.6635	34.0	0.65	1.0	0.0	0.94207	0.016695	0.015695	1.000000	1.000000	1.005325
3	2005-04-29	940.81	932.40	942.45	929.81	11235419.0	962.3760	23.0	0.10	-1.0	1.0	0.93240	-0.010265	-0.011265	0.989735	0.988735	0.995006
4	2005-05-09	934.65	909.17	937.39	909.17	8529115.0	959.7260	7.0	-0.70	0.0	0.0	0.00000	-0.024914	-0.024914	0.989735	0.988735	0.970216
5	2005-05-10	905.54	913.08	913.39	892.31	10494331.0	957.6000	18.0	-0.15	0.0	0.0	0.00000	0.004301	0.004301	0.989735	0.988735	0.974389
…	…	…	…	…	…	…	…	…	…	…	…	…	…	…	…	…	…
3908	2021-05-25	5165.04	5318.48	5324.36	5161.52	182996838.0	5113.5180	41.0	1.00	1.0	0.0	5.31848	0.031595	0.030595	9.077395	6.360596	5.675588
3909	2021-05-26	5326.12	5320.59	5344.28	5308.74	169607096.0	5122.7750	38.0	0.85	0.0	1.0	0.00000	0.000397	0.000397	9.080997	6.363119	5.677840
3910	2021-05-27	5311.37	5338.23	5378.48	5286.05	148462856.0	5135.8245	29.0	0.40	-1.0	1.0	5.33823	0.003315	0.002315	9.111104	6.377852	5.696664
3911	2021-05-28	5338.73	5321.09	5360.28	5288.65	164605806.0	5147.3530	37.0	0.80	1.0	0.0	5.32109	-0.003211	-0.004211	9.111104	6.377852	5.678373
3912	2021-05-31	5318.08	5331.57	5331.63	5281.69	144603269.0	5157.9695	26.0	0.25	-1.0	1.0	5.33157	0.001970	0.000970	9.129048	6.384036	5.689557

3912 rows × 17 columns

In [48]:

#跑图
import matplotlib.pyplot as plt
fig = plt.figure(figsize = (10,4),dpi = 100)
x = df1['日期']
    
plt.plot(x,df1['net_value'],label = '策略净值',color = 'r')
plt.plot(x,df1['net_value_n'],label='策略净值（交易费率）',color='b')
plt.plot(x,df1['benshen'],label = '沪深300指数',color = 'c')
    
# xlab = range(0,df1.shape[0]-1,int(df1.shape[0]/12))
# plt.xticks(xlab,df1.loc[xlab,'日期'],rotation = 30)
plt.legend(loc = 'best')
plt.show()

In [30]:

#计算回测指标(每一年的年化收益率，夏普比率，最大回撤，)

df2 = df1.copy()
df2['日期'] = pd.to_datetime(df2['日期'])
#年化收益率
df2.set_index('日期',inplace=True)
annual_value = df2['net_value'].resample('Y')
list(annual_value)

#计算每一年有两种方法：resample('y')或者groupby(df['日期'].dt.year)

Out[30]:

[(Timestamp('2005-12-31 00:00:00', freq='A-DEC'),
  日期
  2005-04-27    1.000000
  2005-04-28    1.000000
  2005-04-29    0.989735
  2005-05-09    0.989735
  2005-05-10    0.989735
                  ...   
  2005-12-26    1.051362
  2005-12-27    1.047920
  2005-12-28    1.049698
  2005-12-29    1.062362
  2005-12-30    1.052582
  Name: net_value, Length: 168, dtype: float64),
 (Timestamp('2006-12-31 00:00:00', freq='A-DEC'),
  日期
  2006-01-04    1.073076
  2006-01-05    1.093251
  2006-01-06    1.105676
  2006-01-09    1.111625
  2006-01-10    1.114931
                  ...   
  2006-12-25    1.719127
  2006-12-26    1.719127
  2006-12-27    1.719127
  2006-12-28    1.719127
  2006-12-29    1.772196
  Name: net_value, Length: 241, dtype: float64),
 (Timestamp('2007-12-31 00:00:00', freq='A-DEC'),
  日期
  2007-01-04    1.794806
  2007-01-05    1.799834
  2007-01-08    1.850784
  2007-01-09    1.910287
  2007-01-10    1.958806
                  ...   
  2007-12-24    2.290320
  2007-12-25    2.294578
  2007-12-26    2.315787
  2007-12-27    2.360871
  2007-12-28    2.348001
  Name: net_value, Length: 242, dtype: float64),
 (Timestamp('2008-12-31 00:00:00', freq='A-DEC'),
  日期
  2008-01-02    2.368599
  2008-01-03    2.384843
  2008-01-04    2.411946
  2008-01-07    2.444028
  2008-01-08    2.431475
                  ...   
  2008-12-25    2.581121
  2008-12-26    2.581121
  2008-12-29    2.581121
  2008-12-30    2.581121
  2008-12-31    2.581121
  Name: net_value, Length: 246, dtype: float64),
 (Timestamp('2009-12-31 00:00:00', freq='A-DEC'),
  日期
  2009-01-05    2.581121
  2009-01-06    2.581121
  2009-01-07    2.581121
  2009-01-08    2.581121
  2009-01-09    2.581121
                  ...   
  2009-12-25    2.944005
  2009-12-28    2.944005
  2009-12-29    2.944005
  2009-12-30    2.944005
  2009-12-31    2.944005
  Name: net_value, Length: 244, dtype: float64),
 (Timestamp('2010-12-31 00:00:00', freq='A-DEC'),
  日期
  2010-01-04    2.944005
  2010-01-05    2.944005
  2010-01-06    2.944005
  2010-01-07    2.944005
  2010-01-08    2.944005
                  ...   
  2010-12-27    3.246186
  2010-12-28    3.246186
  2010-12-29    3.246186
  2010-12-30    3.246186
  2010-12-31    3.246186
  Name: net_value, Length: 242, dtype: float64),
 (Timestamp('2011-12-31 00:00:00', freq='A-DEC'),
  日期
  2011-01-04    3.246186
  2011-01-05    3.231918
  2011-01-06    3.231918
  2011-01-07    3.231918
  2011-01-10    3.172283
                  ...   
  2011-12-26    3.534381
  2011-12-27    3.534381
  2011-12-28    3.534381
  2011-12-29    3.534381
  2011-12-30    3.534381
  Name: net_value, Length: 244, dtype: float64),
 (Timestamp('2012-12-31 00:00:00', freq='A-DEC'),
  日期
  2012-01-04    3.534381
  2012-01-05    3.534381
  2012-01-06    3.556443
  2012-01-09    3.556443
  2012-01-10    3.674733
                  ...   
  2012-12-25    3.583772
  2012-12-26    3.597267
  2012-12-27    3.578195
  2012-12-28    3.630099
  2012-12-31    3.692892
  Name: net_value, Length: 243, dtype: float64),
 (Timestamp('2013-12-31 00:00:00', freq='A-DEC'),
  日期
  2013-01-04    3.695029
  2013-01-07    3.711979
  2013-01-08    3.696376
  2013-01-09    3.696376
  2013-01-10    3.696376
                  ...   
  2013-12-25    4.111625
  2013-12-26    4.111625
  2013-12-27    4.111625
  2013-12-30    4.111625
  2013-12-31    4.111625
  Name: net_value, Length: 238, dtype: float64),
 (Timestamp('2014-12-31 00:00:00', freq='A-DEC'),
  日期
  2014-01-02    4.111625
  2014-01-03    4.111625
  2014-01-06    4.111625
  2014-01-07    4.111625
  2014-01-08    4.111625
                  ...   
  2014-12-25    5.126517
  2014-12-26    5.126517
  2014-12-29    5.140830
  2014-12-30    5.143939
  2014-12-31    5.143939
  Name: net_value, Length: 245, dtype: float64),
 (Timestamp('2015-12-31 00:00:00', freq='A-DEC'),
  日期
  2015-01-05    5.143939
  2015-01-06    5.143261
  2015-01-07    5.147117
  2015-01-08    5.147117
  2015-01-09    5.147117
                  ...   
  2015-12-25    6.774694
  2015-12-28    6.774694
  2015-12-29    6.774694
  2015-12-30    6.774694
  2015-12-31    6.774694
  Name: net_value, Length: 244, dtype: float64),
 (Timestamp('2016-12-31 00:00:00', freq='A-DEC'),
  日期
  2016-01-04    6.774694
  2016-01-05    6.774694
  2016-01-06    6.774694
  2016-01-07    6.774694
  2016-01-08    6.774694
                  ...   
  2016-12-26    7.278410
  2016-12-27    7.278410
  2016-12-28    7.278410
  2016-12-29    7.278410
  2016-12-30    7.278410
  Name: net_value, Length: 244, dtype: float64),
 (Timestamp('2017-12-31 00:00:00', freq='A-DEC'),
  日期
  2017-01-03    7.278410
  2017-01-04    7.278410
  2017-01-05    7.278410
  2017-01-06    7.278410
  2017-01-09    7.278410
                  ...   
  2017-12-25    7.636594
  2017-12-26    7.636594
  2017-12-27    7.636594
  2017-12-28    7.636594
  2017-12-29    7.636594
  Name: net_value, Length: 244, dtype: float64),
 (Timestamp('2018-12-31 00:00:00', freq='A-DEC'),
  日期
  2018-01-02    7.636594
  2018-01-03    7.681415
  2018-01-04    7.713961
  2018-01-05    7.713961
  2018-01-08    7.753866
                  ...   
  2018-12-24    7.019752
  2018-12-25    7.019752
  2018-12-26    7.019752
  2018-12-27    7.019752
  2018-12-28    7.019752
  Name: net_value, Length: 243, dtype: float64),
 (Timestamp('2019-12-31 00:00:00', freq='A-DEC'),
  日期
  2019-01-02    7.019752
  2019-01-03    7.019752
  2019-01-04    7.019752
  2019-01-07    7.062367
  2019-01-08    7.047106
                  ...   
  2019-12-25    7.986407
  2019-12-26    7.986407
  2019-12-27    7.978552
  2019-12-30    8.096781
  2019-12-31    8.126437
  Name: net_value, Length: 244, dtype: float64),
 (Timestamp('2020-12-31 00:00:00', freq='A-DEC'),
  日期
  2020-01-02    8.236851
  2020-01-03    8.222409
  2020-01-06    8.191345
  2020-01-07    8.252701
  2020-01-08    8.157661
                  ...   
  2020-12-25    9.292844
  2020-12-28    9.292844
  2020-12-29    9.292844
  2020-12-30    9.292844
  2020-12-31    9.292844
  Name: net_value, Length: 243, dtype: float64),
 (Timestamp('2021-12-31 00:00:00', freq='A-DEC'),
  日期
  2021-01-04    9.393471
  2021-01-05    9.573183
  2021-01-06    9.660863
  2021-01-07    9.832034
  2021-01-08    9.799526
                  ...   
  2021-05-25    9.077395
  2021-05-26    9.080997
  2021-05-27    9.111104
  2021-05-28    9.111104
  2021-05-31    9.129048
  Name: net_value, Length: 97, dtype: float64)]

In [85]:

import numpy as np
df3 = df1.copy()
df3['日期'] = pd.to_datetime(df3['日期'])
group_data = df3.groupby(df3['日期'].dt.year)
annual_returns = {}
annual_vol = {}
max_down = {}
sharp={}
trade_winrate={}
trade_count={}
for year,group in group_data:
    #年化收益率
    annual_returns[year] = (group['net_value'].iloc[-1] - group['net_value'].iloc[0])/group['net_value'].iloc[0]
    #年化波动率
    annual_vol[year] = group['pct_change'].std()* (252 ** 0.5)
    #最大回撤
    max_rate = pd.Series.rolling(group['net_value'],window = len(group),min_periods=1).max()
    max_down[year] = ((max_rate - group['net_value'])/max_rate).max()
    #计算夏普比率
    sharp[year] =  annual_returns[year]/annual_vol[year]
    #做多胜率
    sell_trade = np.array(group[group['flag'] < 0].net_value)
    buy_trade = np.array(group[group['flag'] > 0].net_value)
    if group[group['flag']!=0]['flag'].iloc[0] < 0:
        buy_trade = np.insert(buy_trade,0,group['net_value'].iloc[0])
    if group[group['flag']!=0]['flag'].iloc[-1] == 1:
        sell_trade = np.append(sell_trade,group['net_value'].iloc[-1])
    trade_pct = (sell_trade - buy_trade)/buy_trade
    trade_winrate[year] = len(trade_pct[trade_pct>0])/len(trade_pct)
    #交易次数
    trade_count[year] = len(buy_trade)*2
results_df = pd.DataFrame({
    '年化收益率':annual_returns,
    '年化波动率':annual_vol,
    '最大回撤':max_down,
    '夏普比率':sharp,
    '做多胜率':trade_winrate,
    '交易次数':trade_count
})
results_df

Out[85]:

	年化收益率	年化波动率	最大回撤	夏普比率	做多胜率	交易次数
2005	0.052582	0.208093	0.050736	0.252685	0.562500	32
2006	0.651510	0.222871	0.047632	2.923258	0.782609	46
2007	0.308220	0.368108	0.162072	0.837309	0.714286	42
2008	0.089725	0.484516	0.167409	0.185185	0.478261	46
2009	0.140592	0.326617	0.115427	0.430448	0.739130	46
2010	0.102643	0.251362	0.087140	0.408347	0.590909	44
2011	0.088779	0.206636	0.035459	0.429643	0.608696	46
2012	0.044848	0.204371	0.133545	0.219447	0.481481	54
2013	0.112745	0.222203	0.055411	0.507397	0.458333	48
2014	0.251072	0.192548	0.068150	1.303947	0.520000	50
2015	0.317025	0.394255	0.128959	0.804110	0.846154	26
2016	0.074353	0.222238	0.050307	0.334562	0.521739	46
2017	0.049212	0.101465	0.033810	0.485013	0.464286	56
2018	-0.080775	0.214243	0.162613	-0.377022	0.500000	60
2019	0.157653	0.198537	0.053411	0.794073	0.476190	42
2020	0.128203	0.227709	0.078808	0.563015	0.785714	28
2021	-0.028150	0.222799	0.114079	-0.126345	0.714286	14

In [86]:

#循环n和s，计算年化收益和最大回测
import pandas as pd
S=[0.2,0.3,0.4,0.6,0.7,0.8]
N=[10,15,20,25,30,35,40,45,50,55,60]
annual_rate = pd.DataFrame(index=N,columns=S)
max_down = pd.DataFrame(index=N,columns=S)
# print(annual_rate.index)
# print(max_down)
#年化收益率
for i in N:
    for j in S:
        df = run_original_strategy(df,i,j)
        dff = df.copy()
        annual_rate.loc[i,j] = (dff['net_value'].iloc[-1]/dff['net_value'].iloc[0])**(1/17)-1
        max_down.loc[i,j] = dff['pct_change'].std()*((252*17) ** 0.5)
print(annual_rate)
print(max_down)

          0.2        0.3        0.4        0.6        0.7         0.8
10   0.142209  0.0956948  0.0827609  0.0661283  0.0672422  0.00470183
15   0.116233   0.106971   0.110715  0.0815936  0.0752791    0.045797
20   0.130223    0.13268   0.121639   0.104067  0.0852824   0.0526898
25   0.124471   0.108583    0.10413  0.0977906  0.0885462   0.0709694
30   0.108111  0.0934919  0.0735281  0.0817172  0.0793524   0.0825152
35  0.0687295  0.0787064  0.0604639  0.0800909  0.0751698   0.0880687
40  0.0693751   0.080896  0.0727802  0.0739836  0.0692706   0.0715929
45   0.058269  0.0690542  0.0679535  0.0656895  0.0686604   0.0632936
50  0.0676958   0.060232  0.0536289  0.0553279  0.0549616   0.0520086
55  0.0585059    0.06025  0.0575609  0.0523109  0.0543271     0.05069
60  0.0472388  0.0574737  0.0531075  0.0470148  0.0485322   0.0399418
         0.2       0.3       0.4       0.6       0.7       0.8
10   1.11352   1.11358   1.11426   1.11487     1.116   1.11278
15   1.11337   1.11446   1.11555   1.11674   1.11811    1.1198
20    1.1216    1.1241   1.12618    1.1284   1.13086   1.13292
25    1.1315   1.13204   1.13236   1.13495   1.13764   1.13941
30   1.13621   1.12595   1.12818   1.11895   1.11207   1.10798
35   1.10809   1.10433   1.09652   1.08361     1.064    1.0478
40   1.03548   1.01029  0.991673  0.979649  0.976349   0.97763
45  0.960713   0.95712   0.95398  0.956324  0.951205  0.949798
50  0.948967  0.946349  0.947248  0.952209  0.954829   0.95465
55  0.950977  0.955286  0.960424  0.963298  0.967207  0.964548
60  0.964005  0.965237  0.972303  0.977945  0.988824   1.00102

In [ ]:

量化金融交易策略模拟回测平台

策略示例

因子策略之成交量排序

评论

发表回复