-
Notifications
You must be signed in to change notification settings - Fork 452
/
portfolio_optimization.py
252 lines (224 loc) · 10.1 KB
/
portfolio_optimization.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
'''
This is a follow up of https://letianzj.github.io/portfolio-management-one.html
It backtests four portfolios: GMV, tangent, maximum diversification and risk parity
and compare them with equally-weighted portfolio
'''
import os
import numpy as np
import pandas as pd
import pytz
from datetime import datetime, timezone
import quanttrader as qt
from scipy.optimize import minimize
import matplotlib.pyplot as plt
import empyrical as ep
import pyfolio as pf
# set browser full width
from IPython.core.display import display, HTML
display(HTML("<style>.container { width:100% !important; }</style>"))
# ------------------ help functions -------------------------------- #
def minimum_vol_obj(wo, cov):
w = wo.reshape(-1, 1)
sig_p = np.sqrt(np.matmul(w.T, np.matmul(cov, w)))[0, 0] # portfolio sigma
return sig_p
def maximum_sharpe_negative_obj(wo, mu_cov):
w = wo.reshape(-1, 1)
mu = mu_cov[0].reshape(-1, 1)
cov = mu_cov[1]
obj = np.matmul(w.T, mu)[0, 0]
sig_p = np.sqrt(np.matmul(w.T, np.matmul(cov, w)))[0, 0] # portfolio sigma
obj = -1 * obj/sig_p
return obj
def maximum_diversification_negative_obj(wo, cov):
w = wo.reshape(-1, 1)
w_vol = np.matmul(w.T, np.sqrt(np.diag(cov).reshape(-1, 1)))[0, 0]
port_vol = np.sqrt(np.matmul(w.T, np.matmul(cov, w)))[0, 0]
ratio = w_vol / port_vol
return -ratio
# this is also used to verify rc from optimal w
def calc_risk_contribution(wo, cov):
w = wo.reshape(-1, 1)
sigma = np.sqrt(np.matmul(w.T, np.matmul(cov, w)))[0, 0]
mrc = np.matmul(cov, w)
rc = (w * mrc) / sigma # element-wise multiplication
return rc
def risk_budget_obj(wo, cov_wb):
w = wo.reshape(-1, 1)
cov = cov_wb[0]
wb = cov_wb[1].reshape(-1, 1) # target/budget in percent of portfolio risk
sig_p = np.sqrt(np.matmul(w.T, np.matmul(cov, w)))[0, 0] # portfolio sigma
risk_target = sig_p * wb
asset_rc = calc_risk_contribution(w, cov)
f = np.sum(np.square(asset_rc - risk_target.T)) # sum of squared error
return f
class PortfolioOptimization(qt.StrategyBase):
def __init__(self, nlookback=200, model='gmv'):
super(PortfolioOptimization, self).__init__()
self.nlookback = nlookback,
self.model = model
self.current_time = None
def on_tick(self, tick_event):
self.current_time = tick_event.timestamp
# print('Processing {}'.format(self.current_time))
# wait for enough bars
for symbol in self.symbols:
df_hist = self._data_board.get_hist_price(symbol, self.current_time)
if df_hist.shape[0] < self.nlookback:
return
# wait for month end
time_index = self._data_board.get_hist_time_index()
time_loc = time_index.get_loc(self.current_time)
if (time_loc != len(time_index)-1) & (time_index[time_loc].month == time_index[time_loc+1].month):
return
npv = self._position_manager.current_total_capital
n_stocks = len(self.symbols)
TOL = 1e-12
prices = None
for symbol in self.symbols:
price = self._data_board.get_hist_price(symbol, self.current_time)['Close'].iloc[-self.nlookback:]
price = np.array(price)
if prices is None:
prices = price
else:
prices = np.c_[prices, price]
rets = prices[1:,:]/prices[0:-1, :]-1.0
mu = np.mean(rets, axis=0)
cov = np.cov(rets.T)
w = np.ones(n_stocks) / n_stocks # default
try:
if self.model == 'gmv':
w0 = np.ones(n_stocks) / n_stocks
cons = ({'type': 'eq', 'fun': lambda w: np.sum(w) - 1.0}, {'type': 'ineq', 'fun': lambda w: w})
res = minimize(minimum_vol_obj, w0, args=cov, method='SLSQP', constraints=cons, tol=TOL, options={'disp': True})
if not res.success:
print(f'{self.model} Optimization failed')
w = res.x
elif self.model == 'sharpe':
w0 = np.ones(n_stocks) / n_stocks
cons = ({'type': 'eq', 'fun': lambda w: np.sum(w) - 1.0}, {'type': 'ineq', 'fun': lambda w: w})
res = minimize(maximum_sharpe_negative_obj, w0, args=[mu, cov], method='SLSQP', constraints=cons, tol=TOL, options={'disp': True})
w = res.x
elif self.model == 'diversified':
w0 = np.ones(n_stocks) / n_stocks
cons = ({'type': 'eq', 'fun': lambda x: np.sum(x) - 1.0}) # weights sum to one
bnds = tuple([(0, 1)] * n_stocks)
res = minimize(maximum_diversification_negative_obj, w0, bounds=bnds, args=cov, method='SLSQP', constraints=cons, tol=TOL, options={'disp': True})
w = res.x
elif self.model == 'risk_parity':
w0 = np.ones(n_stocks) / n_stocks
w_b = np.ones(n_stocks) / n_stocks # risk budget/target, percent of total portfolio risk (in this case equal risk)
# bnds = ((0,1),(0,1),(0,1),(0,1)) # alternative, use bounds for weights, one for each stock
cons = ({'type': 'eq', 'fun': lambda x: np.sum(x) - 1.0}, {'type': 'ineq', 'fun': lambda x: x})
res = minimize(risk_budget_obj, w0, args=[cov, w_b], method='SLSQP', constraints=cons, tol=TOL, options={'disp': True})
w = res.x
except Exception as e:
print(f'{self.model} Optimization failed; {str(e)}')
i = 0
for sym in self.symbols:
current_size = self._position_manager.get_position_size(sym)
current_price = self._data_board.get_hist_price(sym, self.current_time)['Close'].iloc[-1]
target_size = (int)(npv * w[i] / current_price)
self.adjust_position(sym, size_from=current_size, size_to=target_size, timestamp=self.current_time)
print('REBALANCE ORDER SENT, %s, Price: %.2f, Percentage: %.2f, Target Size: %.2f' %
(sym,
current_price,
w[i],
target_size))
i += 1
if __name__ == '__main__':
etfs = ['SPY', 'EFA', 'TIP', 'GSG', 'VNQ']
models = ['gmv', 'sharpe', 'diversified', 'risk_parity']
benchmark = etfs
init_capital = 100_000.0
test_start_date = datetime(2010,1,1, 8, 30, 0, 0, pytz.timezone('America/New_York'))
test_end_date = datetime(2019,12,31, 6, 0, 0, 0, pytz.timezone('America/New_York'))
dict_results = dict()
for model in models:
dict_results[model] = dict()
# SPY: S&P 500
# EFA: MSCI EAFE
# TIP: UST
# GSG: GSCI
# VNQ: REITs
strategy = PortfolioOptimization()
strategy.set_capital(init_capital)
strategy.set_symbols(etfs)
strategy.set_params({'nlookback': 200, 'model': model})
backtest_engine = qt.BacktestEngine(test_start_date, test_end_date)
backtest_engine.set_capital(init_capital) # capital or portfolio >= capital for one strategy
for symbol in etfs:
data = qt.util.read_ohlcv_csv(os.path.join('../data/', f'{symbol}.csv'))
backtest_engine.add_data(symbol, data)
backtest_engine.set_strategy(strategy)
ds_equity, df_positions, df_trades = backtest_engine.run()
# save to excel
qt.util.save_one_run_results('./output', ds_equity, df_positions, df_trades, batch_tag=model)
ds_ret = ds_equity.pct_change().dropna()
ds_ret.name = model
dict_results[model]['equity'] = ds_equity
dict_results[model]['return'] = ds_ret
dict_results[model]['positions'] = df_positions
dict_results[model]['transactions'] = df_trades
# ------------------------- Evaluation and Plotting -------------------------------------- #
bm = pd.DataFrame()
for s in etfs:
df_temp = qt.util.read_ohlcv_csv(os.path.join('../data/', f'{s}.csv'))
df_temp = df_temp['Close']
df_temp.name = s
bm = pd.concat([bm, df_temp], axis=1)
bm_ret = bm.pct_change().dropna()
bm_ret.index = pd.to_datetime(bm_ret.index)
bm_ret = bm_ret.loc[dict_results[models[0]]['return'].index]
bm_ret['benchmark'] = bm_ret.mean(axis=1) # 20% each
bm_value = init_capital * (bm_ret + 1).cumprod()
perf_stats_all = pd.DataFrame()
for m in models:
perf_stats_strat = pf.timeseries.perf_stats(dict_results[m]['return'])
perf_stats_strat.name = m
perf_stats_all = pd.concat([perf_stats_all, perf_stats_strat], axis=1)
perf_stats_bm = pf.timeseries.perf_stats(bm_ret.benchmark)
perf_stats_bm.name = 'equal_weights'
perf_stats_all = pd.concat([perf_stats_all, perf_stats_bm], axis=1)
print(perf_stats_all)
# portfolio values
portfolio_value_all = pd.DataFrame()
for m in models:
port_value = dict_results[m]['positions'].sum(axis=1)
port_value.name = m
portfolio_value_all = pd.concat([portfolio_value_all, port_value], axis=1)
port_value = bm_value.benchmark.copy()
port_value.name = 'equal_weights'
portfolio_value_all = pd.concat([portfolio_value_all, port_value], axis=1)
fig, ax = plt.subplots(2, 1, figsize=(5, 12))
portfolio_value_all.plot(ax=ax[0])
bm_value[etfs].plot(ax=ax[1])
fig.tight_layout()
plt.show()
# monthly returns
fig, ax = plt.subplots(5, 1, figsize=(10, 35))
i = 0
for m in models:
pf.plotting.plot_monthly_returns_heatmap(dict_results[m]['return'], ax[i])
ax[i].title.set_text(m)
i += 1
pf.plotting.plot_monthly_returns_heatmap(bm_ret['benchmark'], ax[i])
ax[i].title.set_text('equal weighted')
fig.tight_layout()
plt.show()
# positions
fig, ax = plt.subplots(4, 1, figsize=(25, 25))
etfs_plus_cash = etfs+['cash']
i = 0
for m in models:
sum_ = dict_results[m]['positions'].sum(axis=1)
pcts = []
for etf in etfs_plus_cash:
pct = dict_results[m]['positions'][etf] / sum_
pcts.append(pct)
print(pcts[0].shape, len(pcts))
ax[i].stackplot(pcts[0].index, pcts, labels=etfs_plus_cash)
ax[i].legend(loc='upper left')
ax[i].title.set_text(m)
i += 1
fig.tight_layout()
plt.show()