-
Notifications
You must be signed in to change notification settings - Fork 653
/
combining-smart-factors-momentum-and-market-portfolio.py
218 lines (180 loc) · 9.07 KB
/
combining-smart-factors-momentum-and-market-portfolio.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
# region imports
from AlgorithmImports import *
# endregion
# https://quantpedia.com/strategies/combining-smart-factors-momentum-and-market-portfolio/
#
# The investment universe consists of factors from the Alpha Architect’s Factor Investing Data Library (factor for all major investment styles such as Value, Quality, Momentum, Size and Volatility)
# based on the top 1500 US stocks. Firstly construct the fast and slow signals for each factor. The fast signal is the past one-month return, and the slow signal is the past twelve-months return.
# For each type of signal, to obtain the weights, cross-sectionally rank signals’ based on their absolute values. The weight for the individual slow or fast signal is equal to the corresponding rank
# divided by the sum of all ranks and multiplied by the signal’s sign (equations 3 and 4 in the paper). For the dynamically blended strategy (smart factors strategy), each factor has a final weight of
# three-quarters of the weight of fast signal plus one-quarter of the weight of slow signal (equation 12). Nextly, consider the top 1500 US stocks as the market portfolio. The combined smart factors
# and market strategy finds the weights of the market and factor portfolio using past moving averages of the returns. The combined strategy looks back on the past twelve months, and twelve MAs of the
# returns. Suppose the MA for active investing (factor momentum) is larger than MA for market portfolio, then the active investing scores one point. Otherwise, the market portfolio gets one point.
# Therefore, each month, the weight of the factor momentum and market portfolio is determined by the number of “winning” (loosing) moving averages (equations 13 and 14). The strategy is rebalanced monthly.
#
# QC Implementation:
# - IWM etf is used as 'market' portfolio.
import numpy as np
class CombiningSmartFactorsMomentumandMarketPortfolio(QCAlgorithm):
def Initialize(self):
self.SetStartDate(2000, 1, 1)
self.SetCash(100000)
self.symbols = {
"momentum": "US_EQUAL_DECILE_1500_12_2m_L_S",
"value": "US_EQUAL_DECILE_1500_B_M_L_S",
"quality": "US_EQUAL_DECILE_1500_ROA_L_S",
"size": "US_EQUAL_DECILE_1500_Size_L_S",
"volatility": "US_EQUAL_DECILE_1500_Volatility_L_S",
}
# monthly price data
self.data = {}
self.long_period = 13
self.short_period = 2
self.monthly_returns = {}
self.monthly_returns_period = 12
for symbol, equity_symbol in self.symbols.items():
data = self.AddData(USEquity, equity_symbol, Resolution.Daily)
data.SetLeverage(10)
data.SetFeeModel(CustomFeeModel(self))
self.data[symbol] = RollingWindow[float](self.long_period)
self.market = self.AddEquity("IWM", Resolution.Daily).Symbol
self.data[self.market] = RollingWindow[float](self.short_period)
self.monthly_returns["smart_factors"] = RollingWindow[float](
self.monthly_returns_period
)
self.monthly_returns["market"] = RollingWindow[float](
self.monthly_returns_period
)
self.selection_flag = False
self.Schedule.On(
self.DateRules.MonthStart(self.market),
self.TimeRules.AfterMarketOpen(self.market),
self.Rebalance,
)
def OnSecuritiesChanged(self, changes):
for security in changes.AddedSecurities:
security.SetFeeModel(CustomFeeModel(self))
security.SetLeverage(5)
def OnData(self, data):
# store factor monthly prices
for symbol, equity_symbol in self.symbols.items():
if equity_symbol in data and data[equity_symbol]:
price = data[equity_symbol].Value
self.data[symbol].Add(price)
# store market prices
if self.market in data and data[self.market]:
market_price = data[self.market].Value
self.data[self.market].Add(market_price)
def Rebalance(self):
slow_momentum = {}
fast_momentum = {}
# calculate both momentum values
for symbol, equity_symbol in self.symbols.items():
if self.data[symbol].IsReady:
slow_momentum[symbol] = (
self.data[symbol][0] / self.data[symbol][self.long_period - 1] - 1
)
fast_momentum[symbol] = self.data[symbol][0] / self.data[symbol][1] - 1
if len(fast_momentum) != 0:
# momentum ranking
total_weight = {}
# weights
rank_sum = sum([x for x in range(1, len(slow_momentum) + 1)])
sorted_by_slow_momentum = sorted(
slow_momentum.items(), key=lambda x: abs(x[1]), reverse=False
)
slow_weight = {}
for i, (symbol, momentum) in enumerate(sorted_by_slow_momentum):
rank = i + 1
slow_weight[symbol] = (rank / rank_sum) * np.sign(momentum)
sorted_by_fast_momentum = sorted(
fast_momentum.items(), key=lambda x: abs(x[1]), reverse=False
)
fast_weight = {}
for i, (symbol, momentum) in enumerate(sorted_by_fast_momentum):
rank = i + 1
fast_weight[symbol] = (rank / rank_sum) * np.sign(momentum)
# total weight
for symbol, equity_symbol in self.symbols.items():
if symbol in slow_momentum and symbol in fast_momentum:
s_weight = slow_weight[symbol]
f_weight = fast_weight[symbol]
total_weight[symbol] = 0.75 * f_weight + 0.25 * s_weight
# retrun calculation for market and smart factors
if self.data[self.market].IsReady:
market_return = self.data[self.market][0] / self.data[self.market][1] - 1
self.monthly_returns["market"].Add(market_return)
# smart factor return calculation
smart_factors_return = 0
for symbol, momentum_1M in fast_momentum.items():
if symbol in total_weight:
w = total_weight[symbol]
symbol_ret = w * momentum_1M
smart_factors_return += symbol_ret
if smart_factors_return != 0:
self.monthly_returns["smart_factors"].Add(smart_factors_return)
score = {}
traded_weight = {}
# calculate 12 SMA's
if (
self.monthly_returns["smart_factors"].IsReady
and self.monthly_returns["market"].IsReady
):
score["smart_factors"] = 0
score["market"] = 0
for sma_period in range(1, 13):
factor_returns = [x for x in self.monthly_returns["smart_factors"]][
:sma_period
]
market_returns = [x for x in self.monthly_returns["market"]][
:sma_period
]
factor_mean_return = np.mean(factor_returns)
market_mean_return = np.mean(market_returns)
if factor_mean_return > market_mean_return:
score["smart_factors"] += 1
else:
score["market"] += 1
total_score = score["market"] + score["smart_factors"]
if total_score != 0:
traded_weight["market"] = score["market"] / total_score
traded_weight["smart_factors"] = (
score["smart_factors"] / total_score
)
# order execution
# market
self.SetHoldings(self.market, traded_weight["market"])
# smart factors
for symbol, equity_symbol in self.symbols.items():
if symbol in total_weight:
w = total_weight[symbol]
self.SetHoldings(
equity_symbol, traded_weight["smart_factors"] * w
)
class USEquity(PythonData):
def GetSource(self, config, date, isLiveMode):
return SubscriptionDataSource(
"data.quantpedia.com/backtesting_data/equity/us_ew_decile/{0}.csv".format(
config.Symbol.Value
),
SubscriptionTransportMedium.RemoteFile,
FileFormat.Csv,
)
# File example.
# date;equity
# 1992-01-31;0.98
def Reader(self, config, line, date, isLiveMode):
data = USEquity()
data.Symbol = config.Symbol
if not line[0].isdigit():
return None
split = line.split(";")
# Prevent lookahead bias.
data.Time = datetime.strptime(split[0], "%Y-%m-%d") + timedelta(days=1)
data.Value = float(split[1])
return data
# Custom fee model
class CustomFeeModel(FeeModel):
def GetOrderFee(self, parameters):
fee = parameters.Security.Price * parameters.Order.AbsoluteQuantity * 0.00005
return OrderFee(CashAmount(fee, "USD"))