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
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
|
from .db import _engine, dbconn
from dates import bond_cal
import numpy as np
from .utils import tenor_t
from functools import lru_cache
from pyisda.curve import SpreadCurve
from multiprocessing import Pool
from yieldcurve import get_curve
import datetime
import pandas as pd
def insert_quotes():
"""
backpopulate some version i+1 quotes one day before they start trading so
that we get continuous time series when we compute returns.
We can also do it in sql as follows:
INSERT INTO index_quotes(date, index, series, version, tenor, closeprice)
SELECT date, index, series, version+1, tenor, (factor1*closeprice-100*0.355)/factor2
FROM index_quotes
WHERE index='HY' and series=23 and date='2017-02-02'
"""
dates = pd.DatetimeIndex(['2014-05-21', '2015-02-19', '2015-03-05', '2015-06-23'])
df = pd.read_sql_query("SELECT DISTINCT ON (date) * FROM index_quotes "
"WHERE index='HY' AND tenor='5yr' "
"ORDER BY date, series DESC, version DESC",
_engine, parse_dates=['date'], index_col=['date'])
df = df.loc[dates]
for tup in df.itertuples():
result = _engine.execute("SELECT indexfactor, cumulativeloss FROM index_version "
"WHERE index = 'HY' AND series=%s AND version in (%s, %s)"
"ORDER BY version",
(tup.series, tup.version, tup.version+1))
factor1, cumloss1 = result.fetchone()
factor2, cumloss2 = result.fetchone()
recovery = 1-(cumloss2-cumloss1)
version2_price = (factor1 * tup.closeprice - 100*recovery)/factor2
print(version2_price)
_engine.execute("INSERT INTO index_quotes(date, index, series, version, tenor, closeprice)"
"VALUES(%s, %s, %s, %s, %s, %s)",
(tup.Index, 'HY', tup.series, tup.version+1, tup.tenor, version2_price))
def get_index_quotes(index=None, series=None, tenor=None, from_date=None,
years=3, remove_holidays=True, source='MKIT'):
args = locals().copy()
del args['remove_holidays']
if args['years'] is not None:
args['date'] = (pd.Timestamp.now() - pd.DateOffset(years=years)).date()
del args['years']
if args['from_date']:
args['date'] = args['from_date']
del args['from_date']
def make_str(key, val):
if isinstance(val, list):
op = "IN"
return "{} IN %({})s".format(key, key)
elif isinstance(val, datetime.date):
op = ">="
else:
op = "="
return "{} {} %({})s".format(key, op, key)
where_clause = " AND ".join(make_str(k, v)
for k, v in args.items() if v is not None)
sql_str = "SELECT * FROM index_quotes_pre"
if where_clause:
sql_str = " WHERE ".join([sql_str, where_clause])
def make_params(args):
return {k: tuple(v) if isinstance(v, list) else v
for k, v in args.items() if v is not None}
df = pd.read_sql_query(sql_str, _engine, parse_dates=['date'],
index_col=['date', 'index', 'series', 'version'],
params=make_params(args))
df.tenor = df.tenor.astype(tenor_t)
df = df.set_index('tenor', append=True)
df.sort_index(inplace=True)
# get rid of US holidays
if remove_holidays:
dates = df.index.levels[0]
if index in ['IG', 'HY']:
holidays = bond_cal().holidays(start=dates[0], end=dates[-1])
df = df.loc(axis=0)[dates.difference(holidays), :, :]
return df
def index_returns(df=None, index=None, series=None, tenor=None, from_date=None,
years=3, per=1):
"""computes spreads and price returns
Parameters
----------
df : pandas.DataFrame
index : str or List[str], optional
index type, one of 'IG', 'HY', 'EU', 'XO'
series : int or List[int], optional
tenor : str or List[str], optional
tenor in years e.g: '3yr', '5yr'
date : datetime.date, optional
starting date
years : int, optional
limits many years do we go back starting from today.
per: int, optional
calculate returns across different time frames
"""
if df is None:
df = get_index_quotes(index, series, tenor, from_date, years)
spread_return = (df.
groupby(level=['index', 'series', 'tenor', 'version']).
close_spread.
pct_change(periods=per))
price_return = (df.
groupby(level=['index', 'series', 'tenor', 'version']).
close_price.
diff() / 100)
df = pd.concat([spread_return, price_return], axis=1,
keys=['spread_return', 'price_return'])
df = df.groupby(level=['date', 'index', 'series', 'tenor']).nth(0)
coupon_data = pd.read_sql_query("SELECT index, series, tenor, coupon * 1e-4 AS coupon, "
"maturity FROM "
"index_maturity WHERE coupon is NOT NULL",
_engine,
index_col=['index', 'series', 'tenor'])
df = df.reset_index('date').join(coupon_data).reset_index('tenor')
# for some reason pandas doesn't keep the categories, so we have to
# do this little dance
df.tenor = df.tenor.astype(tenor_t)
df = df.set_index('tenor', append=True)
df['day_frac'] = (df.groupby(level=['index', 'series', 'tenor'])['date'].
transform(lambda s: s.
diff().
astype('timedelta64[D]') / 360))
df['price_return'] += df.day_frac * df.coupon
df = df.drop(['day_frac', 'coupon', 'maturity'], axis=1)
return df.set_index(['date'], append=True)
def get_singlenames_quotes(indexname, date):
conn = dbconn('serenitasdb')
with conn.cursor() as c:
c.execute("SELECT * FROM curve_quotes(%s, %s)", vars=(indexname, date))
return list(c)
def build_curve(r, tenors, currency="USD"):
if r['date'] is None:
raise ValueError(f"Curve for {r['cds_ticker']} is missing")
spread_curve = 1e-4 * np.array(r['spread_curve'], dtype='float')
upfront_curve = 1e-2 * np.array(r['upfront_curve'], dtype='float')
recovery_curve = np.array(r['recovery_curve'], dtype='float')
yc = get_curve(r['date'], currency)
try:
sc = SpreadCurve(r['date'], yc, None, None, None,
tenors, spread_curve, upfront_curve, recovery_curve,
ticker=r['cds_ticker'])
except ValueError as e:
print(r[0], e)
return None
return sc
def build_curves(quotes, args):
return [build_curve(q, *args) for q in quotes if q is not None]
def build_curves_dist(quotes, args, workers=4):
# about twice as fast as the non distributed version
# non thread safe for some reason so need ProcessPool
with Pool(workers) as pool:
r = pool.starmap(build_curve, [(q, *args) for q in quotes], 30)
return r
@lru_cache(maxsize=16)
def _get_singlenames_curves(index_type, series, trade_date, tenors):
sn_quotes = get_singlenames_quotes(f"{index_type.lower()}{series}",
trade_date)
currency = "EUR" if index_type in ["XO", "EU"] else "USD"
args = (np.array(tenors), currency)
return build_curves_dist(sn_quotes, args)
def get_singlenames_curves(index_type, series, trade_date,
tenors=(0.5, 1, 2, 3, 4, 5, 7, 10)):
if isinstance(trade_date, pd.Timestamp):
trade_date = trade_date.date()
return _get_singlenames_curves(index_type, series,
min(datetime.date.today(), trade_date),
tenors)
def get_tranche_quotes(index_type, series, tenor, date=datetime.date.today()):
conn = dbconn('serenitasdb')
with conn.cursor() as c:
c.callproc("get_tranche_quotes", (index_type, series, tenor, date))
return pd.DataFrame.from_records(dict(d) for d in c)
|
