-
Notifications
You must be signed in to change notification settings - Fork 2
/
pairs_trading_strat_backtest.py
137 lines (91 loc) · 5.77 KB
/
pairs_trading_strat_backtest.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
import pandas as pd
import numpy as np
import pickle
from functools import reduce
import itertools
import statsmodels.api as sm
import matplotlib.pyplot as plt
pd.set_option('display.max_columns', None) # or 1000
li_pair_stop_trade_dfs = pickle.load(open('data/backtest/li_pair_stop_trade_dfs.pkl', 'rb'))
apparels_closes = pickle.load(open('data/backtest/apparels_closes.pkl', 'rb'))
apparels_closes.index = pd.to_datetime(apparels_closes.index)
# print(li_pair_stop_trade_dfs[0].dropna())
def get_UB_touch(ts, UB):
crit1 = ts.shift(1) < UB.shift(1)
crit2 = ts > UB
return ts[(crit1) & (crit2)].rename('UB_touch')
def get_LB_touch(ts, LB):
crit1 = ts.shift(1) > LB.shift(1)
crit2 = ts < LB
return ts[(crit1) & (crit2)].rename('LB_touch')
port_stats = pd.DataFrame([], columns=['port_name', 'PnL', 'return'])
dollar_port_li = []
for ind,pair in enumerate(li_pair_stop_trade_dfs):
pair_name = pair.columns.tolist()[0]
pair['stop_trading'] = pair['estimates_actuals_Date Actual'].notna()
df_pair = pd.merge(apparels_closes[[col for col in apparels_closes.columns.tolist() if col in pair.columns[0]]]
, pair, left_index=True, right_index=True, how='left').sort_index()
# print(df_pair)
alpha=0.5
span=50
df_pair['mean'] = df_pair[pair_name].ewm(span=span).mean()
df_pair['std'] = df_pair[pair_name].ewm(span=span).std()
df_pair = df_pair['2016-04-03'::]
z = 1.645
df_pair['UB'], df_pair['LB'] = df_pair['mean'] + df_pair['std']*z , df_pair['mean'] - df_pair['std']*z
#print(df_pair.columns)
# plt.plot(df_pair[['UB', 'LB', pair_name]])
# plt.show()
UB_touch_df, LB_touch_df = get_UB_touch(df_pair[pair_name], df_pair['UB']), get_LB_touch(df_pair[pair_name], df_pair['LB'])
# stop_trade_df['stop_trading'] = False
# stop_trade_df['']
# print(UB_touch_df)
list_trading_df = [df_pair, UB_touch_df, LB_touch_df]
trading_df = reduce(lambda X, x: pd.merge_asof(X.sort_index(), x.sort_index(), left_index=True, right_index=True,
direction='forward', tolerance=pd.Timedelta('1d')), list_trading_df)
# print(trading_df)
# print(trading_df.columns)
# print(trading_df['UB_touch'].notna())
# print(trading_df['stop_trading'])
# print(trading_df[['stop_trading', 'UB_touch']].groupby(['stop_trading', 'UB_touch']).agg('count'))
# print(trading_df[(trading_df['UB_touch'] > 0 )])
# trading_df['net_long'] = trading_df[['UB_touch']
# trading_df['net_short'] = trading_df['LB_touch'].apply(lambda x: True if x > 0 else False)
# print(trading_df['net_long'].dropna(), print(trading_df['stop_trading'].dropna()))
trading_df['net_long'] = (trading_df['UB_touch'] > 0) & (trading_df['stop_trading'] == False)
trading_df['net_short'] = (trading_df['LB_touch'] > 0) & (trading_df['stop_trading'] == False)
# print(trading_df.net_short[trading_df['net_short'] == True])
trading_df['net_trade'] = trading_df.net_long.astype(int) + trading_df.net_short.astype(int)*-1
# print(trading_df.net_trade[trading_df['net_trade'] == 1])
# print(trading_df.columns)
exit_days = 10
trading_df.loc[trading_df.net_trade == 1, 'enter_net_book_UB'] = trading_df['net_trade']*-1*trading_df[trading_df.columns.tolist()[0]] + \
trading_df['net_trade']*trading_df[trading_df.columns.tolist()[1]]
trading_df.loc[trading_df.net_trade == 1, 'exit_net_book_UB'] = trading_df['net_trade']*trading_df[trading_df.columns.tolist()[0]].shift(exit_days) + \
trading_df['net_trade']*-1*trading_df[trading_df.columns.tolist()[1]].shift(exit_days)
trading_df['net_book_UB'] = trading_df['enter_net_book_UB'] + trading_df['exit_net_book_UB']
# print(trading_df['net_book_UB'].fillna(0).cumsum())
trading_df.loc[trading_df.net_trade == -1, 'enter_net_book_LB'] = trading_df['net_trade']*trading_df[trading_df.columns.tolist()[0]] + \
trading_df['net_trade']*-1*trading_df[trading_df.columns.tolist()[1]]
trading_df.loc[trading_df.net_trade == -1, 'exit_net_book_LB'] = trading_df['net_trade']*-1*trading_df[trading_df.columns.tolist()[0]].shift(exit_days) + \
trading_df['net_trade']*trading_df[trading_df.columns.tolist()[1]].shift(exit_days)
trading_df['net_book_LB'] = trading_df['enter_net_book_LB'] + trading_df['exit_net_book_LB']
# print(trading_df['net_book_LB'].fillna(0).cumsum())
trading_df['net_port_position'] = trading_df['net_book_UB'].fillna(0).cumsum() + trading_df['net_book_LB'].fillna(0).cumsum()
invested_amount = abs(trading_df['enter_net_book_UB'].fillna(0)).cumsum() + abs(trading_df['enter_net_book_LB'].fillna(0)).cumsum()
# print(pair[pair.columns.tolist()[0]].name)
port_tup = pd.Series([pair[pair.columns.tolist()[0]].name, trading_df['net_port_position'][-1],
trading_df['net_port_position'][-1] / invested_amount[-1]],
['port_name', 'PnL', 'return'])
dollar_port_li.append(trading_df['net_port_position'])
# print(port_tup)
port_stats = port_stats.append([port_tup], ignore_index=True)
port_stats['return'] = np.power(port_stats['return'] + 1, 1/3) - 1
print(port_stats)
cum_dollar_port = reduce(lambda X, x: pd.merge(X.sort_index(), x.sort_index(), left_index=True, right_index=True,
how='outer'), dollar_port_li)
cum_dollar_port.index = pd.to_datetime(cum_dollar_port.index)
cum_dollar_port = cum_dollar_port.fillna(method='ffill')
cum_dollar_port = cum_dollar_port.fillna(0)
cum_dollar_port = cum_dollar_port.sort_values(by=cum_dollar_port.index[-1],axis=1, ascending=False)
cum_dollar_port.to_csv("data/backtest/pairs_trading_cum_dollar_port.csv")