-
Notifications
You must be signed in to change notification settings - Fork 86
/
badj_rating.py
164 lines (129 loc) · 6.94 KB
/
badj_rating.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
#!/usr/bin/env python
from regress import *
from loaddata import *
from util import *
def calc_badj_daily(daily_df, horizon):
print "Caculating daily badj..."
result_df = filter_expandable(daily_df)
print "Calculating badj0..."
result_df['badj0'] = result_df['log_ret'] / result_df['pbeta']
result_df['badj0_B'] = winsorize_by_date(result_df[ 'badj0' ])
print result_df.columns
result_df['rating'] = -1 * result_df['rating_diff_mean'].fillna(0)
result_df.loc[ result_df['rating'] != 0, 'badj0_B'] = np.nan
result_df = result_df.dropna(subset=['badj0_B'])
demean = lambda x: (x - x.mean())
indgroups = result_df[['badj0_B', 'gdate', 'ind1']].groupby(['gdate', 'ind1'], sort=True).transform(demean)
result_df['badj0_B_ma'] = indgroups['badj0_B']
print "Calulating lags..."
for lag in range(1,horizon+1):
shift_df = result_df.unstack().shift(lag).stack()
result_df['badj'+str(lag)+'_B_ma'] = shift_df['badj0_B_ma']
return result_df
def calc_badj_intra(intra_df):
print "Calculating badj intra..."
result_df = filter_expandable(intra_df)
print "Calulating badjC..."
result_df['badjC'] = (result_df['overnight_log_ret'] + (np.log(result_df['iclose']/result_df['dopen']))) / result_df['pbeta']
result_df['badjC_B'] = winsorize_by_ts(result_df[ 'badjC' ])
result_df['rating'] = -1 * pd.rolling_sum(result_df['rating_diff_mean'].fillna(0).unstack(level='sid'), 28).stack()
result_df.loc[ result_df['rating'] != 0, 'badjC_B'] = np.nan
result_df = result_df.dropna(subset=['badjC_B'])
print "Calulating badjC_ma..."
demean = lambda x: (x - x.mean())
indgroups = result_df[['badjC_B', 'giclose_ts', 'ind1']].groupby(['giclose_ts', 'ind1'], sort=True).transform(demean)
result_df['badjC_B_ma'] = indgroups['badjC_B']
print "Calculated {} values".format(len(result_df['badjC_B_ma'].dropna()))
print result_df.xs(testid, level='sid')['badjC_B_ma']
return result_df
def badj_fits(daily_df, intra_df, horizon, name, middate=None):
insample_intra_df = intra_df
insample_daily_df = daily_df
outsample_intra_df = intra_df
if middate is not None:
insample_intra_df = intra_df[ intra_df['date'] < middate ]
insample_daily_df = daily_df[ daily_df.index.get_level_values('date') < middate ]
outsample_intra_df = intra_df[ intra_df['date'] >= middate ]
outsample_intra_df['badj_b'] = np.nan
outsample_intra_df[ 'badjC_B_ma_coef' ] = np.nan
for lag in range(1, horizon+1):
outsample_intra_df[ 'badj' + str(lag) + '_B_ma_coef' ] = np.nan
fits_df = pd.DataFrame(columns=['horizon', 'coef', 'indep', 'tstat', 'nobs', 'stderr'])
for lag in range(1,horizon+1):
fitresults_df = regress_alpha(insample_daily_df, 'badj0_B_ma', lag, True, 'daily')
fits_df = fits_df.append(fitresults_df, ignore_index=True)
plot_fit(fits_df, "badj_daily_"+name+"_" + df_dates(insample_daily_df))
fits_df.set_index(keys=['indep', 'horizon'], inplace=True)
coef0 = fits_df.ix['badj0_B_ma'].ix[horizon].ix['coef']
outsample_intra_df[ 'badjC_B_ma_coef' ] = coef0
print "Coef0: {}".format(coef0)
for lag in range(1,horizon):
coef = coef0 - fits_df.ix['badj0_B_ma'].ix[lag].ix['coef']
print "Coef{}: {}".format(lag, coef)
outsample_intra_df[ 'badj'+str(lag)+'_B_ma_coef' ] = coef
outsample_intra_df['badj_b'] = outsample_intra_df['badjC_B_ma'] * outsample_intra_df['badjC_B_ma_coef']
for lag in range(1,horizon):
outsample_intra_df[ 'badj_b'] += outsample_intra_df['badj'+str(lag)+'_B_ma'] * outsample_intra_df['badj'+str(lag)+'_B_ma_coef']
print "SEAN"
print insample_intra_df.xs(testid, level='sid')['badjC_B_ma'].describe()
print outsample_intra_df.xs(testid, level='sid')['badjC_B_ma'].describe()
return outsample_intra_df
def calc_badj_forecast(daily_df, intra_df, horizon, middate):
daily_results_df = calc_badj_daily(daily_df, horizon)
forwards_df = calc_forward_returns(daily_df, horizon)
daily_results_df = pd.concat( [daily_results_df, forwards_df], axis=1)
intra_results_df = calc_badj_intra(intra_df)
intra_results_df = merge_intra_data(daily_results_df, intra_results_df)
res5 = badj_fits( daily_results_df, intra_results_df, horizon, "ot_eq", middate)
result_df = pd.concat([res5], verify_integrity=True).sort()
# sector_name = 'Energy'
# print "Running badj for sector {}".format(sector_name)
# sector_df = daily_results_df[ daily_results_df['sector_name'] == sector_name ]
# sector_intra_results_df = intra_results_df[ intra_results_df['sector_name'] == sector_name ]
# result1_df = badj_fits(sector_df, sector_intra_results_df, horizon, "in", middate)
# print "Running badj for not sector {}".format(sector_name)
# sector_df = daily_results_df[ daily_results_df['sector_name'] != sector_name ]
# sector_intra_results_df = intra_results_df[ intra_results_df['sector_name'] != sector_name ]
# result2_df = badj_fits(sector_df, sector_intra_results_df, horizon, "ex", middate)
# result_df = pd.concat([result1_df, result2_df], verify_integrity=True)
return result_df
if __name__=="__main__":
parser = argparse.ArgumentParser(description='G')
parser.add_argument("--start",action="store",dest="start",default=None)
parser.add_argument("--end",action="store",dest="end",default=None)
parser.add_argument("--mid",action="store",dest="mid",default=None)
args = parser.parse_args()
start = args.start
end = args.end
lookback = 30
horizon = 3
pname = "./badj_b" + start + "." + end
start = dateparser.parse(start)
end = dateparser.parse(end)
middate = dateparser.parse(args.mid)
freq="15Min"
loaded = False
try:
daily_df = pd.read_hdf(pname+"_daily.h5", 'table')
intra_df = pd.read_hdf(pname+"_intra.h5", 'table')
loaded = True
except:
print "Did not load cached data..."
if not loaded:
uni_df = get_uni(start, end, lookback)
BARRA_COLS = ['ind1', 'pbeta']
barra_df = load_barra(uni_df, start, end, BARRA_COLS)
PRICE_COLS = ['close', 'overnight_log_ret']
price_df = load_prices(uni_df, start, end, PRICE_COLS)
DBAR_COLS = ['close', 'dopen', 'dvolume']
intra_df = load_daybars(price_df[['ticker']], start, end, DBAR_COLS, freq)
daily_df = merge_barra_data(price_df, barra_df)
intra_df = merge_intra_data(daily_df, intra_df)
analyst_df = load_ratings_hist(price_df[['ticker']], start, end, False, True)
intra_df = pd.merge(intra_df, analyst_df, left_index=True, right_index=True, how='left')
daily_df = merge_intra_eod(daily_df, intra_df)
daily_df.to_hdf(pname+"_daily.h5", 'table', complib='zlib')
intra_df.to_hdf(pname+"_intra.h5", 'table', complib='zlib')
result_df = calc_badj_forecast(daily_df, intra_df, horizon, middate)
print result_df
dump_alpha(result_df, 'badj_b')