/
sim.py
executable file
·937 lines (784 loc) · 40 KB
/
sim.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
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
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
#!/usr/bin/env python
from __future__ import division, unicode_literals
"""Simulate a sequence of trades and portfolio performance over time
Examples:
$ python sim.py trade 50000 orders.csv values.csv
# input file
$ cat orders.csv:
2008, 12, 3, AAPL, BUY, 130
2008, 12, 8, AAPL, SELL, 130
2008, 12, 5, IBM, BUY, 50
# output file
$ cat values.csv
2008, 12, 3, 50000.25
2008, 12, 4, 50010.25
2008, 12, 5, 50250.125
"""
import argparse
import sys
import csv
import datetime
import re
import math
import itertools
import os
import json
import copy
from collections import Mapping
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
from dateutil.parser import parse as parse_date
from scipy import integrate
import QSTK.qstkutil.DataAccess as da
import QSTK.qstkutil.qsdateutil as du
import QSTK.qstkstudy.EventProfiler as ep
from pug import debug
from pug.nlp import util
# from pug.decorators import memoize
#t = qstk.dateutil.getNYSEdays(datetime.datetime(2010,1,1), datetime.datetime(2010,2,1), datetime.timedelta(hours=16))
dataobj = da.DataAccess('Yahoo')
DATE_SEP = re.compile(r'[^0-9]')
def get_price(symbol='$SPY', date=(2010,1,1), price='actual_close'):
# if hasattr(symbol, '__iter__'):
# return [get_price(sym, date=date, price=price) for sym in symbol]
if isinstance(date, basestring):
date = DATE_SEP.split(date)
if isinstance(date, (tuple, list)):
date = datetime.datetime(*[int(i) for i in date])
if isinstance(date, datetime.date) or date.hour < 9 or date.hour > 16:
date = datetime.datetime(date.year, date.month, date.day, 16)
symbol = str(symbol).upper().strip()
if symbol == '$CASH':
return 1.0
try:
sym_price = dataobj.get_data([date], [symbol], [price])[0][symbol][0]
print 'found {0} price of {1} on {2}'.format(symbol, sym_price, date)
return sym_price
except IndexError:
raise
except:
print 'BAD DATE ({0}) or SYMBOL ({1})'.format(date, symbol)
return None
def portfolio_value(portfolio, date, price='close'):
"""Total value of a portfolio (dict mapping symbols to numbers of shares)
$CASH used as symbol for USD
"""
value = 0.0
for (sym, sym_shares) in portfolio.iteritems():
sym_price = None
if sym_shares:
sym_price = get_price(symbol=sym, date=date, price=price)
# print sym, sym_shares, sym_price
# print last_date, k, price
if sym_price != None:
if np.isnan(sym_price):
print 'Invalid price, shares, value, total: ', sym_price, sym_shares, (float(sym_shares) * float(sym_price)) if sym_shares and sym_price else 'Invalid', value
if sym_shares:
return float('nan')
else:
# print ('{0} shares of {1} = {2} * {3} = {4}'.format(sym_shares, sym, sym_shares, sym_price, sym_shares * sym_price))
value += sym_shares * sym_price
# print 'new price, value = {0}, {1}'.format(sym_price, value)
return value
################################################
# General ticker symbol pricing and charting
def chart(
symbols=("AAPL", "GLD", "GOOG", "$SPX", "XOM", "msft"),
start=datetime.datetime(2008, 1, 1),
end=datetime.datetime(2009, 12, 31), # data stops at 2013/1/1
normalize=True,
):
"""Display a graph of the price history for the list of ticker symbols provided
Arguments:
symbols (list of str): Ticker symbols like "GOOG", "AAPL", etc
start (datetime): The date at the start of the period being analyzed.
end (datetime): The date at the end of the period being analyzed.
normalize (bool): Whether to normalize prices to 1 at the start of the time series.
"""
start = util.normalize_date(start or datetime.date(2008, 1, 1))
end = util.normalize_date(end or datetime.date(2009, 12, 31))
symbols = [s.upper() for s in symbols]
timeofday = datetime.timedelta(hours=16)
timestamps = du.getNYSEdays(start, end, timeofday)
ls_keys = ['open', 'high', 'low', 'close', 'volume', 'actual_close']
ldf_data = da.get_data(timestamps, symbols, ls_keys)
d_data = dict(zip(ls_keys, ldf_data))
na_price = d_data['close'].values
if normalize:
na_price /= na_price[0, :]
plt.clf()
plt.plot(timestamps, na_price)
plt.legend(symbols)
plt.ylabel('Adjusted Close')
plt.xlabel('Date')
plt.savefig('chart.pdf', format='pdf')
plt.grid(True)
plt.show()
return na_price
def chart_series(series, market_sym='$SPX', price='actual_close', normalize=True):
"""Display a graph of the price history for the list of ticker symbols provided
Arguments:
series (dataframe, list of str, or list of tuples):
datafram (Timestamp or Datetime for index)
other columns are float y-axis values to be plotted
list of str: 1st 3 comma or slash-separated integers are the year, month, day
others are float y-axis values
list of tuples: 1st 3 integers are year, month, day
others are float y-axis values
market_sym (str): ticker symbol of equity or comodity to plot along side the series
price (str): which market data value ('close', 'actual_close', 'volume', etc) to use
for the market symbol for comparison to the series
normalize (bool): Whether to normalize prices to 1 at the start of the time series.
"""
series = util.make_dataframe(series)
start = util.normalize_date(series.index[0] or datetime.datetime(2008, 1, 1))
end = util.normalize_date(series.index[-1] or datetime.datetime(2009, 12, 28))
timestamps = du.getNYSEdays(start, end, datetime.timedelta(hours=16))
if market_sym:
if isinstance(market_sym, basestring):
market_sym = [market_sym.upper().strip()]
reference_prices = da.get_data(timestamps, market_sym, [price])[0]
reference_dict = dict(zip(market_sym, reference_prices))
for sym, market_data in reference_dict.iteritems():
series[sym] = pd.Series(market_data, index=timestamps)
# na_price = reference_dict[price].values
# if normalize:
# na_price /= na_price[0, :]
series.plot()
# plt.clf()
# plt.plot(timestamps, na_price)
# plt.legend(symbols)
# plt.ylabel(price.title())
# plt.xlabel('Date')
# # plt.savefig('portfolio.chart_series.pdf', format='pdf')
plt.grid(True)
plt.show()
return series
def normalize_symbols(symbols, *args):
"""Return a list of uppercase strings like "GOOG", "$SPX, "XOM"...
Arguments:
symbols (str or list of str): list of market ticker symbols to normalize
If `symbols` is a str a get_symbols_from_list() call is used to retrieve the list of symbols
Returns:
list of str: list of cananical ticker symbol strings (typically after .upper().strip())
Examples:
>>> normalize_symbols("Goog")
["GOOG"]
>>> normalize_symbols(" $SPX ", " aaPL ")
["$SPX", "AAPL"]
>>> normalize_symbols(["$SPX", ["GOOG", "AAPL"]])
["$SPX", "GOOG", "AAPL"]
>>> normalize_symbols(" $Spy, Goog, aAPL ")
["$SPY", "GOOG", "AAPL"]
"""
if ( (hasattr(symbols, '__iter__') and not any(symbols))
or (isinstance(symbols, (list, tuple, Mapping)) and not symbols)):
return []
if isinstance(symbols, basestring):
# get_symbols_from_list seems robust to string normalizaiton like .upper()
try:
return list(set(dataobj.get_symbols_from_list(symbols)))
except:
return [s.upper().strip() for s in symbols.split(',')]
else:
ans = []
for sym in (list(symbols) + list(args)):
tmp = normalize_symbols(sym)
ans = ans + tmp
return list(set(ans))
def clean_dataframe(df):
"""Fill NaNs with the previous value, the next value or if all are NaN then 1.0"""
df = df.fillna(method='ffill')
df = df.fillna(method='bfill')
df = df.fillna(1.0)
return df
def clean_dataframes(dfs):
"""Fill NaNs with the previous value, the next value or if all are NaN then 1.0
TODO:
Linear interpolation and extrapolation
Arguments:
dfs (list of dataframes): list of dataframes that contain NaNs to be removed
Returns:
list of dataframes: list of dataframes with NaNs replaced by interpolated values
"""
if isinstance(dfs, (list)):
for df in dfs:
df = clean_dataframe(df)
return dfs
else:
return [clean_dataframe(dfs)]
def price_dataframe(symbols='sp5002012',
start=datetime.datetime(2008, 1, 1),
end=datetime.datetime(2009, 12, 31),
price_type='actual_close',
cleaner=clean_dataframe,
):
"""Retrieve the prices of a list of equities as a DataFrame (columns = symbols)
Arguments:
symbols (list of str): Ticker symbols like "GOOG", "AAPL", etc
e.g. ["AAPL", " slv ", GLD", "GOOG", "$SPX", "XOM", "msft"]
start (datetime): The date at the start of the period being analyzed.
end (datetime): The date at the end of the period being analyzed.
Yahoo data stops at 2013/1/1
"""
if isinstance(price_type, basestring):
price_type = [price_type]
start = util.normalize_date(start or datetime.date(2008, 1, 1))
end = util.normalize_date(end or datetime.date(2009, 12, 31))
symbols = normalize_symbols(symbols)
t = du.getNYSEdays(start, end, datetime.timedelta(hours=16))
df = clean_dataframes(dataobj.get_data(t, symbols, price_type))
if not df or len(df) > 1:
return cleaner(df)
else:
return cleaner(df[0])
def portfolio_prices(
symbols=("AAPL", "GLD", "GOOG", "$SPX", "XOM", "msft"),
start=datetime.datetime(2005, 1, 1),
end=datetime.datetime(2011, 12, 31), # data stops at 2013/1/1
normalize=True,
allocation=None,
price_type='actual_close',
):
"""Calculate the Sharpe Ratio and other performance metrics for a portfolio
Arguments:
symbols (list of str): Ticker symbols like "GOOG", "AAPL", etc
start (datetime): The date at the start of the period being analyzed.
end (datetime): The date at the end of the period being analyzed.
normalize (bool): Whether to normalize prices to 1 at the start of the time series.
allocation (list of float): The portion of the portfolio allocated to each equity.
"""
symbols = normalize_symbols(symbols)
start = util.normalize_date(start)
end = util.normalize_date(end)
if allocation is None:
allocation = [1. / len(symbols)] * len(symbols)
if len(allocation) < len(symbols):
allocation = list(allocation) + [1. / len(symbols)] * (len(symbols) - len(allocation))
total = np.sum(allocation.sum)
allocation = np.array([(float(a) / total) for a in allocation])
timestamps = du.getNYSEdays(start, end, datetime.timedelta(hours=16))
ls_keys = [price_type]
ldf_data = da.get_data(timestamps, symbols, ls_keys)
d_data = dict(zip(ls_keys, ldf_data))
na_price = d_data[price_type].values
if normalize:
na_price /= na_price[0, :]
na_price *= allocation
return np.sum(na_price, axis=1)
# General ticker symbol pricing and charting
###################################################################
################################################
# Bolinger band charts and indicator values
def series_bollinger(series, window=20, sigma=1., plot=False):
mean = pd.rolling_mean(series, window=window)
std = pd.rolling_std(series, window=window)
df = pd.DataFrame({'value': series, 'mean': mean, 'upper': mean + sigma * std, 'lower': mean - sigma * std})
bollinger_values = (series - pd.rolling_mean(series, window=window)) / (pd.rolling_std(series, window=window))
if plot:
df.plot()
pd.DataFrame({'bollinger': bollinger_values}).plot()
plt.show()
return bollinger_values
def frame_bollinger(df, window=20, sigma=1., plot=False):
bol = pd.DataFrame()
for col in df.columns:
bol[col] = series_bollinger(df[col], plot=False)
return bol
def symbol_bollinger(symbol='GOOG',
start=datetime.datetime(2008, 1, 1), end=datetime.datetime(2009, 12, 31), price_type='close', cleaner=clean_dataframe,
window=20, sigma=1.):
"""Calculate the Bolinger indicator value
>>> symbol_bollinger("goog", '2008-1-1', '2008-2-1')[-1] # doctest: +ELLIPSIS, +NORMALIZE_WHITESPACE
-1.8782...
"""
symbols = normalize_symbols(symbol)
prices = price_dataframe(symbols, start=start, end=end, price_type=price_type, cleaner=cleaner)
return series_bollinger(prices[symbols[0]], window=window, sigma=sigma, plot=False)
def symbols_bollinger(symbols='sp5002012',
start=datetime.datetime(2008, 1, 1), end=datetime.datetime(2009, 12, 31), price_type='adjusted_close', cleaner=clean_dataframe,
window=20, sigma=1.):
"""Calculate the Bolinger for a list or set of symbols
Example:
>>> symbols_bollinger(["AAPL", "GOOG", "IBM", "MSFT"], '10-12-01', '10-12-30')[-5:] # doctest: +NORMALIZE_WHITESPACE
GOOG AAPL IBM MSFT
2010-12-23 16:00:00 1.298178 1.185009 1.177220 1.237684
2010-12-27 16:00:00 1.073603 1.371298 0.590403 0.932911
2010-12-28 16:00:00 0.745548 1.436278 0.863406 0.812844
2010-12-29 16:00:00 0.874885 1.464894 2.096242 0.752602
2010-12-30 16:00:00 0.634661 0.793493 1.959324 0.498395
"""
symbols = normalize_symbols(symbols)
prices = price_dataframe(symbols, start=start, end=end, price_type=price_type, cleaner=cleaner)
return frame_bollinger(prices, window=window, sigma=sigma, plot=False)
def bollinger_events(symbols, start=None, end=None, price_type='close', window=20, market_symbol='SPY', threshold=-2, threshold_market=1.4, threshold_yest=None):
threshold = threshold or -2.0
threshold_yest = threshold_yest or threshold
threshold_market = threshold_market or threshold * (-0.65)
bol = symbols_bollinger(symbols, start, end, price_type=price_type, window=window)
# bol = bol.fillna(0.0).values
market = bol.copy()
market_series = symbols_bollinger([market_symbol]*len(bol.columns), start, end, price_type=price_type, window=window)
for sym in market.columns:
market[sym] = market_series
market = market.values
bolv = bol.values
if threshold >= 0:
events = (bolv[1:] > threshold) & (bolv[:-1] <= threshold_yest) & (market[1:] < threshold_market)
else:
events =(bolv[1:] < threshold) & (bolv[:-1] >= threshold_yest) & (market[1:] >= threshold_market)
return pd.DataFrame(events, index=bol.index[1:], columns=bol.columns)
# Bolinger band charts and indicator values
###################################################################
###################################################################
# sim analyze: compute statistics
def metrics(prices, fudge=False, sharpe_days=252., baseline='$SPX'):
"""Calculate the volatiliy, average daily return, Sharpe ratio, and cumulative return
Arguments:
prices (file or basestring or iterable): path to file or file pointer or sequence of prices/values of a portfolio or equity
fudge (bool): Whether to use Tucker Balche's erroneous division by N or the more accurate N-1 for stddev of returns
sharpe_days: Number of trading days in a year. Sharpe ratio = sqrt(sharpe_days) * total_return / std_dev_of_daily_returns
Examples:
>>> metrics(np.array([1,2,3,4])) == {'mean': 0.61111111111111105, 'return': 4.0, 'sharpe': 34.245718429742873, 'std': 0.28327886186626583}
True
>>> metrics(portfolio_prices(symbols=['AAPL', 'GLD', 'GOOG', 'XOM'], start=datetime.datetime(2011,1,1), end=datetime.datetime(2011,12,31), allocations=[0.4, 0.4, 0.0, 0.2])
... ) == {'std': 0.0101467067654, 'mean': 0.000657261102001, 'sharpe': 1.02828403099, 'return': 1.16487261965}
True
"""
if isinstance(prices, basestring) and os.path.isfile(prices):
prices = open(prices, 'rU')
if isinstance(prices, file):
values = {}
csvreader = csv.reader(prices, dialect='excel', quoting=csv.QUOTE_MINIMAL)
for row in csvreader:
# print row
values[tuple(int(s) for s in row[:3])] = row[-1]
prices.close()
prices = [v for (k,v) in sorted(values.items())]
print prices
if isinstance(prices[0], (tuple, list)):
prices = [row[-1] for row in prices]
if sharpe_days == None:
sharpe_days = len(prices)
prices = np.array([float(p) for p in prices])
if not isinstance(fudge, bool) and fudge:
fudge = float(fudge)
elif fudge == True or (isinstance(fudge, float) and fudge == 0.0):
fudge = (len(prices) - 1.) / len(prices)
else:
fudge = 1.0
daily_returns = np.diff(prices) / prices[0:-1]
# print daily_returns
end_price = float(prices[-1])
start_price = (prices[0])
mean = fudge * np.average(daily_returns)
variance = fudge * np.sum((daily_returns - mean) * (daily_returns - mean)) / float(len(daily_returns))
results = {
'standared deviation of daily returns': math.sqrt(variance),
'variance of daily returns': variance,
'average daily return': mean,
'Sharpe ratio': mean * np.sqrt(sharpe_days) / np.sqrt(variance),
'total return': end_price / start_price,
'final value': end_price,
'starting value': start_price,
}
results['return rate'] = results['total return'] - 1.0
return results
def analyze(args):
print 'Report for {0}...'.format(args.infile)
report = metrics(args.infile, fudge=args.fudge, sharpe_days=args.sharpe_days)
print report
return report
# sim analyze: compute statistics
###################################################################
def prices(symbol='$DJI', start=datetime.datetime(2008,1,1), end=datetime.datetime(2009,12,31)):
start = util.normalize_date(start or datetime.date(2008, 1, 1))
end = util.normalize_date(end or datetime.date(2009, 12, 31))
symbol = symbol.upper()
timeofday = datetime.timedelta(hours=16)
timestamps = du.getNYSEdays(start, end, timeofday)
ls_keys = ['open', 'high', 'low', 'close', 'volume', 'actual_close']
ldf_data = da.get_data(timestamps, [symbol], ls_keys)
d_data = dict(zip(ls_keys, ldf_data))
na_price = d_data['close'].values
return na_price[:,0]
def simulate(symbols=("AAPL", "GLD", "GOOG", "$SPX", "XOM", "msft"),
start=datetime.datetime(2005, 1, 1),
end=datetime.datetime(2011, 12, 31), # data stops at 2013/1/1
normalize=True,
allocation=None,
fudge=True,
):
p = portfolio_prices(symbols=symbols, start=start, end=end, normalize=normalize, allocation=allocation)
return metrics(p, fudge=fudge)
def optimize_allocation(symbols=("AAPL", "GLD", "GOOG", "$SPX", "XOM"),
start=datetime.datetime(2005, 1, 1),
end=datetime.datetime(2011, 12, 31),
normalize=True,
):
N = len(symbols)
alloc = itertools.product(range(11), repeat=N-1)
best_results = [0, 0, 0, 0]
for a in alloc:
if sum(a) > 10:
continue
last_alloc = 10 - sum(a)
allocation = 0.1 * np.array(list(a) + [last_alloc])
results = simulate(symbols=symbols, start=start, end=end, normalize=normalize,
allocation=allocation)
if results[2] > best_results[2]:
best_results = results
best_allocation = allocation
print allocation
print results
return best_results, best_allocation
##############################################################
## Generate Buy/Sell orders based on event triggers
def buy_on_drop(symbol_set="sp5002012",
dataobj=dataobj,
start=datetime.datetime(2008, 1, 3),
end=datetime.datetime(2009, 12, 28),
market_sym='$SPX',
threshold=6,
sell_delay=5,
):
'''Compute and display an "event profile" for multiple sets of symbols'''
if symbol_set:
if isinstance(symbol_set, basestring):
if symbol_set.lower().startswith('sp'):
symbol_set = dataobj.get_symbols_from_list(symbol_set.lower())
else:
symbol_set = [sym.stip().upper() for sym in symbol_set.split(",")]
else:
symbol_set = dataobj.get_symbols_from_list("sp5002012")
if market_sym:
symbol_set.append(market_sym)
print "Starting Event Study, retrieving data for the {0} symbol list...".format(symbol_set)
market_data = get_clean_prices(symbol_set, dataobj=dataobj, start=start, end=end)
print "Finding events for {0} symbols between {1} and {2}...".format(len(symbol_set), start, end)
trigger_kwargs={'threshold': threshold}
events = find_events(symbol_set, market_data, market_sym=market_sym, trigger=drop_below, trigger_kwargs=trigger_kwargs)
csvwriter = csv.writer(getattr(args, 'outfile', open('buy_on_drop_outfile.csv', 'w')), dialect='excel', quoting=csv.QUOTE_MINIMAL)
for order in generate_orders(events, sell_delay=sell_delay, sep=None):
csvwriter.writerow(order)
print "Creating Study report for {0} events...".format(len(events))
ep.eventprofiler(events, market_data,
i_lookback=20, i_lookforward=20,
s_filename='Event report--buy on drop below {0} for {1} symbols.pdf'.format(threshold, len(symbol_set)),
b_market_neutral=True,
b_errorbars=True,
s_market_sym=market_sym,
)
return events
def event(args):
return buy_on_drop(symbol_set=args.symbols,
dataobj=dataobj,
start=args.start,
end=args.end,
market_sym=args.baseline,
threshold=args.threshold,
sell_delay=args.delay)
def trade(args):
"""Simulate a sequence of trades indicated in `infile` and write the portfolio value time series to `outfile`"""
print args
print vars(args)['funds']
print args.funds
portfolio = { '$CASH': args.funds }
print portfolio
csvreader = csv.reader(args.infile, dialect='excel', quoting=csv.QUOTE_MINIMAL)
csvwriter = csv.writer(args.outfile, dialect='excel', quoting=csv.QUOTE_MINIMAL)
detailed = not args.simple
history = []
portfolio_history = []
#trading_days = du.getNYSEdays(datetime.datetime(2010,01,01), datetime.datetime(2012,01,01), datetime.timedelta(hours=16))
for row in csvreader:
# print '-'*30 + ' CSV Row ' + '-'*30
# print ', '.join(row)
trade_date = datetime.datetime(*[int(i) for i in (row[:3] + [16])])
if history:
last_date = datetime.datetime(*(history[-1][:3] + [16])) + datetime.timedelta(days=1)
# print (date.date() - last_date).days
while (trade_date - last_date).days > 0:
print 'Filling in the blanks for {0}'.format(last_date)
value = portfolio_value(portfolio, last_date, price='close')
print ' the portfolio value on that date is: {0}'.format(value)
assert(value != None)
# NaN for porfolio value indicates a non-trading day
if not np.isnan(value):
history += [[last_date.year, last_date.month, last_date.day]
+ (["$CASH", "0.0", "0.0"] if args.trades else [])
+ ([json.dumps(portfolio)] if detailed else []) + [value]]
portfolio_history += [datetime.datetime(last_date.year, last_date.month, last_date.day, 16), portfolio]
csvwriter.writerow(history[-1])
last_date += datetime.timedelta(days=1)
trade_symbol = row[3]
trade_shares = float(row[5])
trade_sign = 1 - 2 * int(row[4].strip().upper()[0]=='S')
# If this the first row in the CSV and the symbol is $CASH then it represents an initial deposit (Sell) or withdrawal (Buy) of cash
# otherwise se need to add or deduct whatever security was bought or sold.
# if not (trade_symbol == '$CASH') or history:
portfolio[trade_symbol] = portfolio.get(trade_symbol, 0) + trade_sign * trade_shares
trade_price = get_price(symbol=trade_symbol, date=trade_date, price='close')
while trade_price == None or np.isnan(trade_price) or float(trade_price) == float('nan'):
trade_date += datetime.timedelta(days=1)
trade_price = get_price(symbol=trade_symbol, date=trade_date, price='close')
#print trade_date, trade_symbol, trade_sign, trade_shares, trade_price
if trade_price and trade_shares and trade_sign in (-1, 1):
print 'spending cash: {0}'.format(trade_sign * trade_shares * trade_price)
portfolio['$CASH'] = portfolio.get('$CASH',0.) - trade_sign * trade_shares * trade_price
else:
print 'ERROR: bad price, sign, shares: ', trade_price, trade_sign, trade_shares
history += [[trade_date.year, trade_date.month, trade_date.day, trade_symbol, trade_sign, trade_shares] + ([json.dumps(portfolio)] if detailed else []) + [portfolio_value(portfolio, trade_date, price='close')]]
csvwriter.writerow(history[-1])
return metrics(history)
## Generate Buy/Sell orders based on event triggers
##############################################################
##############################################
## Event Studies
def event_happened(**kwargs):
"""Function that takes as input various prices (today, yesterday, etc) and returns True if an "event" has been triggered
Examples:
Event is found if the symbol is down more then 3% while the market is up more then 2%:
return bool(kwargs['return_today'] <= -0.03 and kwargs['market_return_today'] >= 0.02)
"""
return bool(kwargs['price_today'] < 8.0 and kwargs['price_yest'] >= 8.0)
def drop_below(threshold=5, **kwargs):
"""Trigger function that returns True if the price falls below the threshold
price_today < threshold and price_yest >= threshold
"""
if (
# kwargs['price_today'] and kwargs['price_yest'] and
# not np.isnan(kwargs['price_today'] and not kwargs['price_yest'] and
kwargs['price_today'] < threshold and kwargs['price_yest'] >= threshold
):
return True
else:
return False
def generate_orders(events, sell_delay=5, sep=','):
"""Generate CSV orders based on events indicated in a DataFrame
Arguments:
events (pandas.DataFrame): Table of NaNs or 1's, one column for each symbol.
1 indicates a BUY event. -1 a SELL event. nan or 0 is a nonevent.
sell_delay (float): Number of days to wait before selling back the shares bought
sep (str or None): if sep is None, orders will be returns as tuples of `int`s, `float`s, and `str`s
otherwise the separator will be used to join the order parameters into the yielded str
Returns:
generator of str: yielded CSV rows in the format (yr, mo, day, symbol, Buy/Sell, shares)
"""
sell_delay = float(unicode(sell_delay)) or 1
for i, (t, row) in enumerate(events.iterrows()):
for sym, event in row.to_dict().iteritems():
# print sym, event, type(event)
# return events
if event and not np.isnan(event):
# add a sell event `sell_delay` in the future within the existing `events` DataFrame
# modify the series, but only in the future and be careful not to step on existing events
if event > 0:
sell_event_i = min(i + sell_delay, len(events) - 1)
sell_event_t = events.index[sell_event_i]
sell_event = events[sym][sell_event_i]
if np.isnan(sell_event):
events[sym][sell_event_t] = -1
else:
events[sym][sell_event_t] += -1
order = (t.year, t.month, t.day, sym, 'Buy' if event > 0 else 'Sell', abs(event) * 100)
if isinstance(sep, basestring):
yield sep.join(order)
yield order
def orders_from_events(events, sell_delay=5, num_shares=100):
"""Create a DataFrame of orders (signed share quantities) based on event triggers (T/F or 0/1 matrix)
Arguments:
events (DataFrame): mask table to indicate occurrence of buy event (1 = buy, NaN/0/False = do nothing)
sell_delay (int): number of days after the buy order to initiate a sell order of those shares
num_shares (int): number of shares to buy and sell at each event
Returns:
DataFrame: Signed integer numbers of shares to buy (+) or sell (-)
columns are stock ticker symbols
index is datetime at end of trading day (16:00 in NY)
"""
buy = events.copy() * num_shares
sell = -1 * pd.DataFrame(buy.copy().values[:-sell_delay], index=buy.index[sell_delay:], columns=buy.columns)
sell = pd.concat([0 * buy.iloc[:sell_delay], sell])
for i in range(sell_delay):
sell.iloc[-1] -= buy.iloc[-sell_delay + i]
orders = buy + sell
return orders
def portfolio_from_orders(orders, funds=1e5, price_type='close'):
"""Create a DataFrame of portfolio holdings (#'s' of shares for the symbols and dates)
Appends the "$CASH" symbol to the porfolio and initializes it to `funds` indicated.
Appends the symbol "total_value" to store the total value of cash + stocks at each timestamp.
The symbol holdings are found by multipling each element of the orders matrix by the
price matrix for those symbols and then computing a cumulative sum of those purchases.
portfolio["$CASH"] = funds - (orders * prices).sum(axis=1).cumsum()
portfolio["total_value"] = portfolio["$CASH"] + (orders.cumsum() * prices).sum(axis=1)
"""
portfolio = orders.copy()
prices = price_dataframe(orders.columns, start=orders.index[0], end=orders.index[-1],
price_type=price_type, cleaner=clean_dataframe)
portfolio["$CASH"] = funds - (orders * prices).sum(axis=1).cumsum()
portfolio["total_value"] = portfolio["$CASH"] + (orders.cumsum() * prices).sum(axis=1)
return portfolio
######### Machine Learning or Optimization of Prediction
##
def integrated_change(ts, integrator=integrate.trapz, clip_floor=None, clip_ceil=float('inf')):
"""Total value * time above the starting value within a TimeSeries"""
if clip_floor is None:
clip_floor = ts[0]
if clip_ceil < clip_floor:
polarity = -1
offset, clip_floor, clip_ceil, = clip_ceil, clip_ceil, clip_floor
else:
polarity, offset = 1, clip_floor
clipped_values = np.clip(ts.values - offset, clip_floor, clip_ceil)
print polarity, offset, clip_floor, clip_ceil
print clipped_values
integrator_types = set(['trapz', 'cumtrapz', 'simps', 'romb'])
if integrator in integrator_types:
integrator = integrate.__getattribute__(integrator)
integrator = integrator or integrate.trapz
# datetime units converted to seconds (since 1/1/1970)
return integrator(clipped_values, ts.index.astype(np.int64) / 10**9)
def clipping_start_end(ts, capacity=100):
"""Start and end index that clips the price/value of a time series the most
Assumes that the integrated maximum includes the peak (instantaneous maximum).
Arguments:
ts (TimeSeries): Time series to attempt to clip to as low a max value as possible
capacity (float): Total "funds" or "energy" available for clipping (integrated area under time series)
Returns:
2-tuple: Timestamp of the start and end of the period of the maximum clipped integrated increase
"""
ts_sorted = ts.order(ascending=False)
i, t0, t1, integral = 1, None, None, 0
while integral <= capacity and i+1 < len(ts):
i += 1
t0_within_capacity = t0
t1_within_capacity = t1
t0 = min(ts_sorted.index[:i])
t1 = max(ts_sorted.index[:i])
integral = integrated_change(ts[t0:t1])
print i, t0, ts[t0], t1, ts[t1], integral
if t0_within_capacity and t1_within_capacity:
return t0_within_capacity, t1_within_capacity
# argmax = ts.argmax() # index of the maximum value
def filter_integrated_increase(factors, coef):
"""Multiply linear coeficients by"""
pass
##
######### Machine Learning or Optimization of Prediction
def find_events(symbols, d_data, market_sym='$SPX', trigger=drop_below, trigger_kwargs={}):
'''Return dataframe of 1's (event happened) and NaNs (no event), 1 column for each symbol'''
df_close = d_data['actual_close']
ts_market = df_close[market_sym]
print "Finding `{0}` events with kwargs={1} for {2} ticker symbols".format(trigger.func_name, trigger_kwargs, len(symbols))
print 'Trigger docstring says:\n\n{0}\n\n'.format(trigger.func_doc)
# Creating an empty dataframe
df_events = copy.deepcopy(df_close)
df_events = df_events * np.NAN
# Time stamps for the event range
ldt_timestamps = df_close.index
for s_sym in symbols:
if s_sym == market_sym:
continue
for i in range(1, len(ldt_timestamps)):
# Calculating the returns for this timestamp
kwargs = dict(trigger_kwargs)
kwargs['price_today'] = df_close[s_sym].ix[ldt_timestamps[i]]
kwargs['price_yest'] = df_close[s_sym].ix[ldt_timestamps[i - 1]]
kwargs['return_today'] = (kwargs['price_today'] / (kwargs['price_yest'] or 1.)) - 1
kwargs['market_price_today'] = ts_market.ix[ldt_timestamps[i]]
kwargs['market_price_yest'] = ts_market.ix[ldt_timestamps[i - 1]]
kwargs['market_return_today'] = (kwargs['market_price_today'] / (kwargs['market_price_yest'] or 1.)) - 1
if trigger(**kwargs):
df_events[s_sym].ix[ldt_timestamps[i]] = 1
print 'Found {0} events where priced dropped below {1}.'.format(df_events.sum(axis=1).sum(axis=0), trigger_kwargs['threshold'])
return df_events
def get_clean_prices(symbols=None,
dataobj=dataobj,
start=None,
end=None,
market_sym='$SPX',
reset_cache=True):
start = util.normalize_date(start or datetime.date(2008, 1, 1))
end = util.normalize_date(end or datetime.date(2009, 12, 31))
symbols = normalize_symbols(symbols)
symbols += [market_sym]
print "Calculating timestamps for {0} SP500 symbols".format(len(symbols))
ldt_timestamps = du.getNYSEdays(start, end, datetime.timedelta(hours=16))
ls_keys = ['open', 'high', 'low', 'close', 'volume', 'actual_close']
print "Retrieving data for {0} SP500 symbols between {1} and {2}.".format(len(symbols), start, end)
ldf_data = dataobj.get_data(ldt_timestamps, symbols, ls_keys, )
d_data = dict(zip(ls_keys, ldf_data))
for s_key in ls_keys:
print 'cleaning nans from the column {0}'.format(repr(s_key))
d_data[s_key] = d_data[s_key].fillna(method='ffill')
d_data[s_key] = d_data[s_key].fillna(method='bfill')
d_data[s_key] = d_data[s_key].fillna(1.0)
return d_data
## Event Studies
##############################################
def build_args_parser(parser=None):
# create the top-level parser for this "sim" module
parser = argparse.ArgumentParser(prog='sim', description='Simulate trading and predictive analytics algorithms.')
parser.add_argument('--source',
default='Yahoo',
choices=('Yahoo', 'Google', 'Bloomberg'),
help='Name of financial data source to use in da.DataAccess("Name")')
subparsers = parser.add_subparsers(help='`sim command` help')
# create the parser for the "trade" command
parser_trade = subparsers.add_parser('trade', help='Simulate a sequence of trades')
parser_trade.add_argument('funds', type=float,
nargs='?',
default=50000.0,
help='Initial funds (cash, USD) in portfolio.')
parser_trade.add_argument('infile', nargs='?', type=argparse.FileType('rU'),
help='Path to input CSV file containing a list of trades: y,m,d,sym,BUY/SELL,shares',
default=sys.stdin)
parser_trade.add_argument('outfile', nargs='?', type=argparse.FileType('w'),
help='Path to output CSV file where a time series of dates and portfolio values will be written',
default=sys.stdout)
parser_trade.add_argument('--detailed', action='store_true',
help='Whether to output a json string containing the portfolio allocation as the 2nd-to-last CSV column')
parser_trade.add_argument('--trades', action='store_true',
help='Whether to output the buy/sell events along with the total value of the portfolio in the output.')
parser_trade.add_argument('--simple', action='store_true',
help='Whether to supress output of a json string containing the portfolio allocation as the 2nd-to-last CSV column')
parser_trade.set_defaults(func=trade)
# create the parser for the "analyze" command
parser_analyze = subparsers.add_parser('analyze', help='Analyze a time series (sequence) of prices (typically portfolio values)')
parser_analyze.add_argument('infile', nargs='?', type=argparse.FileType('rU'),
help='Path to input CSV file containing sequence of prices (portfolio values) in the last column. Typically each line should be a (yr, mo, dy, price) CSV string for each trading day in the sequence',
default=sys.stdin)
parser_analyze.add_argument('--fudgefactor', type=float, default=False,
help="Value to multiply the total return by to make it match Tucker Balche's incorrect math")
parser_analyze.add_argument('--fudge', action='store_true',
help="Whether to fudge by N-1")
parser_analyze.add_argument('--sharpe_days', type=float, default=252.0,
help="Value to multiple the daily average return by to get the yearly return for Sharpe ratio calculation.")
parser_analyze.set_defaults(func=analyze)
# create the parser for the "event" command (event studies)
parser_event = subparsers.add_parser('event', help='Generate a sequence of trades based on an event study (trigger events)')
parser_event.add_argument('outfile', nargs='?', type=argparse.FileType('w'),
help="Path to output CSV file to contain the trades (yr, mo, day, symbol, 'Buy'/'Sell', shares)",
default=sys.stdout)
parser_event.add_argument('--price', type=str, default="actual_close",
help="Which price to trigger on (close, actual_close, volume)")
parser_event.add_argument('--threshold', type=float, default=5.0,
help="Buy equities whenever they fall below this actual_close price")
parser_event.add_argument('--start', type=parse_date, default='2008-01-01',
help="Start of time period to perform event study.")
parser_event.add_argument('--end', type=parse_date, default='2009-12-31',
help="End of time period to perform event study.")
parser_event.add_argument('--delay', type=float, default=5,
help="Number of days to hold the stock before selling it.")
parser_event.add_argument('--symbols', type=str, default="sp5002012",
help="Which stocks to search for events for (sp5002012, sp5002008, all, ...).")
parser_event.add_argument('--baseline', type=str, default="$SPX",
help="Which stocks to search for events for (sp5002012, sp5002008, all, ...).")
parser_event.set_defaults(func=event)
return parser
if __name__ == '__main__':
# build the parser and then use it to parse the arguments in sys.args
args = build_args_parser().parse_args()
# run `sim()` or `analyze()` or whatever function is indicated by the `subparser.set_defaults()` for `args.main`
args.func(args)