-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
175 lines (132 loc) · 5.54 KB
/
utils.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
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
import numpy as np
import pandas as pd
def cal_method_bucket_loss(df, name):
t = df.copy()
t['delta_bucket'] = t['delta'].apply(round_tenth)
if 'MVdelta' in name:
name_diff = name
else:
name_diff = name + '_diff'
t[name_diff] = t['dV'] - t[name]*t['dS']
loss = t.groupby(['ticker', 'delta_bucket'])[name_diff].apply(lambda x: np.mean(x**2)).unstack()
loss['overall'] = t.groupby('ticker')[name_diff].apply(lambda x: np.mean(x**2))
return loss
def cal_method_bucket_gain(df, N=25):
name_list = df.columns[N:].to_list()
loss0 = cal_method_bucket_loss(df, 'delta')
gain_all = {}
for name in name_list:
loss = cal_method_bucket_loss(df, name)
gain = 1 - loss/loss0
gain_all[name] = gain
return gain_all
def round_tenth(x, TYPE='C'):
# 0.05 <= x <= 0.95
for i in range(1,10):
if x >=-0.05 + i/10 and x < 0.05 + i/10:
return i
return 9
def round_tenth_put(x):
for i in range(9,0,-1):
if x >=-0.05 - i/10 and x < 0.05 - i/10:
return i
return 1
def cal_bucket_hedge_error_all(df, name='delta', residual=False, TYPE='C'):
k = df
k = k[k['month']>36].copy()
if TYPE == 'C':
k['delta_bucket'] = k['delta'].apply(round_tenth)
elif TYPE == 'P':
k['delta_bucket'] = k['delta'].apply(round_tenth_put)
if not residual:
k['hedge_error'] = k['dV'] - k[name]*k['dS']
k['hedge_error_sq'] = k['hedge_error']**2
else:
k['hedge_error_sq'] = k[name]**2
overall = k['hedge_error_sq'].mean()
bucket = k.groupby('delta_bucket')['hedge_error_sq'].mean()
return overall, bucket
def cal_bucket_gain_all(k, name, residual, subzero=False, TYPE='C', mean=True):
overall, bucket = cal_bucket_hedge_error_all(k, 'delta', False, TYPE=TYPE)
toverall, tbucket = cal_bucket_hedge_error_all(k, name, residual, TYPE=TYPE)
gainoveral = 1 - toverall / overall
gainbucket = 1 - tbucket / bucket
if subzero:
if TYPE == 'C':
k['delta_bucket'] = k['delta'].apply(round_tenth)
elif TYPE == 'P':
k['delta_bucket'] = k['delta'].apply(round_tenth_put)
for d in k['delta_bucket'].unique():
if gainbucket[d] < 0:
k.loc[k['delta_bucket']==d, name] = k.loc[k['delta_bucket']==d, 'delta']
overall, bucket = cal_bucket_hedge_error_all(k, 'delta', False, TYPE=TYPE)
toverall, tbucket = cal_bucket_hedge_error_all(k, name, residual, TYPE=TYPE)
gainoveral = 1 - toverall / overall
gainbucket = 1 - tbucket / bucket
gainbucket[gainbucket<1e-6] = 0
return gainoveral, gainbucket
def cal_gain_all_for_all_method(df, TYPE='C', subzero=False, N=25):
res = {}
for name in df.columns[N:]:
print('='*50)
print(name)
if 'MV' in name or 'xgb' in name:
gain, gain_bucket = cal_bucket_gain_all(df, name, True, subzero=subzero, TYPE=TYPE)
else:
gain, gain_bucket = cal_bucket_gain_all(df, name, False, subzero=subzero, TYPE=TYPE)
print(gain)
print(gain_bucket)
gain_bucket['overall'] = gain
res[name] = gain_bucket
return pd.concat(res, axis=1).T
def cal_bucket_gain(df, name='delta', residual=False, TYPE='C'):
k = df
k = k[k['month']>36].copy()
if TYPE == 'C':
k['delta_bucket'] = k['delta'].apply(round_tenth)
elif TYPE == 'P':
k['delta_bucket'] = k['delta'].apply(round_tenth_put)
if not residual:
k['hedge_error'] = k['dV'] - k[name]*k['dS']
k['hedge_error_sq'] = k['hedge_error']**2
else:
k['hedge_error_sq'] = k[name]**2
overall = {}
bucket = {}
for month in k['month'].unique():
kk = k[k['month']==month]
overall[month] = kk['hedge_error_sq'].mean()
bucket[month] = kk.groupby('delta_bucket')['hedge_error_sq'].mean()
return pd.Series(overall), pd.concat(bucket, axis=1)
def cal_average_bucket_gain(k, name, residual, overall, bucket, TYPE='C', mean=True):
toverall, tbucket = cal_bucket_gain(k, name, residual, TYPE=TYPE)
gainoveral = 1 - toverall / overall
gainbucket = 1 - tbucket / bucket
if not mean:
return gainoveral, gainbucket
return gainoveral.mean(), gainbucket.mean(axis=1)
def cal_gain_for_all_method(df, TYPE='C', N=25):
overall, bucket = cal_bucket_gain(df, name='delta', residual=False, TYPE=TYPE)
for name in df.columns[N:]:
print('='*50)
print(name)
if 'MV' in name :
gain, gain_bucket = cal_average_bucket_gain(df, name, True, overall, bucket, TYPE=TYPE)
if ('DNN' in name or 'RNN' in name) and 'MV' not in name:
gain, gain_bucket = cal_average_bucket_gain(df, name, False, overall, bucket, TYPE=TYPE)
print(gain)
print(gain_bucket)
def generate_OTM(ticker='AAPL', TYPE='C'):
folder = './data_'+ticker+'/'
filename = ticker + '_Option_'+TYPE+'_A'
df = pd.read_hdf(folder+filename+'.h5')
if TYPE == 'C':
suffix = '_OTM'
df[df['delta']<=0.5] .to_hdf('./data_'+ticker+'/'+filename+suffix+'.h5', key=filename+suffix)
suffix = '_OTM2'
df[df['delta']<0.55] .to_hdf('./data_'+ticker+'/'+filename+suffix+'.h5', key=filename+suffix)
elif TYPE == 'P':
suffix = '_OTM'
df[df['delta']>=-0.5] .to_hdf('./data_'+ticker+'/'+filename+suffix+'.h5', key=filename+suffix)
suffix = '_OTM2'
df[df['delta']>-0.55] .to_hdf('./data_'+ticker+'/'+filename+suffix+'.h5', key=filename+suffix)