商品趋势跟随策略中的季节性

1import pandas as pd
2from AlgorithmImports import *
3import numpy as np
4class SeasonalityWithinTrendFollowing(QCAlgorithm):
5def Initialize(self):
6 self.SetStartDate(1990, 1, 1)
7 self.SetCash(100000)
8 
9 self.symbols = ["CME_S1",   # Soybean Futures, Continuous Contract
10                 "CME_W1",   # Wheat Futures, Continuous Contract
11                 "CME_SM1",  # Soybean Meal Futures, Continuous Contract
12                 "CME_BO1",  # Soybean Oil Futures, Continuous Contract
13                 "CME_C1",   # Corn Futures, Continuous Contract
14                 "CME_O1",   # Oats Futures, Continuous Contract
15                 "CME_LC1",  # Live Cattle Futures, Continuous Contract
16                 "CME_FC1",  # Feeder Cattle Futures, Continuous Contract
17                 "CME_LN1",  # Lean Hog Futures, Continuous Contract
18                 "CME_GC1",  # Gold Futures, Continuous Contract
19                 "CME_SI1",  # Silver Futures, Continuous Contract
20                 "CME_PL1",  # Platinum Futures, Continuous Contract
21                 "CME_CL1",  # Crude Oil Futures, Continuous Contract
22                 "CME_HG1",  # Copper Futures, Continuous Contract
23                 "CME_LB1",  # Random Length Lumber Futures, Continuous Contract
24                 # "CME_NG1",  # Natural Gas (Henry Hub) Physical Futures, Continuous Contract
25                 "CME_PA1",  # Palladium Futures, Continuous Contract 
26                 "CME_RR1",  # Rough Rice Futures, Continuous Contract
27                 "ICE_RS1",  # Canola Futures, Continuous Contract
28                 "ICE_GO1",  # Gas Oil Futures, Continuous Contract
29                 "CME_RB2",  # Gasoline Futures, Continuous Contract
30                 "CME_KW2",  # Wheat Kansas, Continuous Contract
31                 "ICE_WT1",  # WTI Crude Futures, Continuous Contract
32                 "ICE_CC1",  # Cocoa Futures, Continuous Contract 
33                 "ICE_CT1",  # Cotton No. 2 Futures, Continuous Contract
34                 "ICE_KC1",  # Coffee C Futures, Continuous Contract
35                 "ICE_O1",   # Heating Oil Futures, Continuous Contract
36                 "ICE_OJ1",  # Orange Juice Futures, Continuous Contract
37                 "ICE_SB1"   # Sugar No. 11 Futures, Continuous Contract
38                 ]
39 self.period = 12 * 21
40 self.quantile = 5
41 self.SetWarmup(self.period)
42 self.data = {}
43 for symbol in self.symbols:
44     data = self.AddData(QuantpediaFutures, symbol, Resolution.Daily)
45     data.SetLeverage(5)
46     data.SetFeeModel(CustomFeeModel())
47     
48     self.data[symbol] = SymbolData(self.period, self)
49 self.Schedule.On(self.DateRules.MonthStart(self.symbols[0]), self.TimeRules.At(0, 0), self.Rebalance)
50
51def OnData(self, data):
52 # Update RollingWindows in SymbolData object
53 for symbol in self.symbols:
54     if symbol in data and data[symbol]:
55         price = data[symbol].Value
56         self.data[symbol].update(price,self.Time.date())
57             
58def Rebalance(self):
59 if self.IsWarmingUp: return
60
61 month = self.Time.month
62 
63 # Thanks to this condition we make sure,
64 # that montly return is stored under right month in SymbolData.months dictionary
65 if month == 1:
66     month = 12
67 else:
68     month -= 1
69 
70 positive_return = [] # Store futures with positive yearly return
71 negative_return = [] # Store futures with negative yearly return
72 for symbol in self.symbols:
73     # If prices for whole year are ready, we can calculate yearly return
74     if self.data[symbol].is_ready():
75         # Firstly calculate, then store current monthly return in proper list
76         self.MonthlyReturnStoring(symbol, month)
77         
78         # Based on yearly return store future symbol in proper list
79         if self.data[symbol].yearly_return() > 0:
80             positive_return.append(symbol)
81         
82         else:
83             negative_return.append(symbol)
84         
85         
86     elif self.data[symbol].are_monthly_prices_ready():
87         # Firstly calculate, then store current monthly return in proper list
88         self.MonthlyReturnStoring(symbol, month)
89 
90 # Seasonality sorting
91 return_avg = {}
92 custom_data_last_update_date: Dict[Symbol, datetime.date] = QuantpediaFutures.get_last_update_date()
93  
94 for symbol in self.symbols:
95     if self.Securities[symbol].GetLastData() and self.Time.date() < custom_data_last_update_date[symbol]:
96         # Check if there is enough data for current symbol of future
97         if self.data[symbol].are_monthly_returns_ready(month):
98             # Calculate average from yield returns in the same calendar month in history.
99             # Past year's same-month return is to be excluded from the seasonality calculation
100             return_avg[symbol] = self.data[symbol].get_avg_of_yield_returns(month)
101         
102 long = []
103 short = []
104 # Prorgam will not continue if there aren't enough future symbols for quintile selection
105 if len(return_avg) >= self.quantile:
106     # Create top and bottom list based on avg sort
107     quintile = int(len(return_avg) / self.quantile)
108     sorted_by_avg_return = [x[0] for x in sorted(return_avg.items(), key = lambda item: item[1], reverse = True)]
109     
110     top = sorted_by_avg_return[:quintile]
111     bottom = sorted_by_avg_return[-quintile:]
112     
113     # Create long and short portfolio based on strategy description
114     long = [x for x in positive_return if x not in bottom]
115     short = [x for x in negative_return if x not in top]
116         
117 # Trade execution
118 targets: List[PortfolioTarget] = []
119 for i, portfolio in enumerate([long, short]):
120     for symbol in portfolio:
121         targets.append(PortfolioTarget(symbol, ((-1) ** i) / len(portfolio)))
122 
123 self.SetHoldings(targets, True)
124
125def MonthlyReturnStoring(self, symbol, month):
126 # Calculate monthly return for current month
127 monthly_return = self.data[symbol].monthly_return()
128 
129 # Store blank list under current monnth number, if it doesn't exists
130 # Thanks to this list we are making history of monthly returns for each year
131 if month not in self.data[symbol].months:
132     self.data[symbol].months[month] = []
133
134 # Add current monthly return into list of all montly returns in this month
135 self.data[symbol].months[month].append(monthly_return)
136     
137class SymbolData():
138def __init__(self, period, algorithm):
139 self.monthly_prices = RollingWindow[float](21) # Storing prices for monthly return calculation
140 self.prices = RollingWindow[float](period) # Storing daily prices of futures
141 self.months = {} # Storing monthly returns in list for each month separately
142 self.last_update_date = None
143 self.algorithm = algorithm
144 
145def update(self, price, date):
146 self.monthly_prices.Add(price)
147 self.prices.Add(price)
148 self.last_update_date = date
149 
150def is_ready(self):
151 return self.prices.IsReady and ((self.algorithm.Time.date() - self.last_update_date).days <= 4)
152 
153def are_monthly_prices_ready(self):
154 return self.monthly_prices.IsReady
155 
156def are_monthly_returns_ready(self, month):
157 # Check if there are at least 6 monhtly returns in specific month
158 if month in self.months and len(self.months[month]) > 5:
159     return True 
160 else:
161     return False
162 
163def monthly_return(self):
164 prices = [x for x in self.monthly_prices]
165 return (prices[0] - prices[-1]) / prices[-1]
166 
167def yearly_return(self):
168 prices = [x for x in self.prices]
169 return (prices[0] - prices[-1]) / prices[-1]
170 
171def get_avg_of_yield_returns(self, month):
172 # Get all monthly returns for specific month
173 monthly_returns = self.months[month]
174 
175 # We get average from max 10 values of monthly returns
176 if len(monthly_returns) > 10:
177     monthly_returns = monthly_returns[-11:]
178     
179 # The past year's same-month return is to be excluded from the seasonality calculation to avoid any correlation between the two signals
180 monthly_returns = monthly_returns[:-1]
181 # Return average of monthly returns in specific month
182 return np.mean(monthly_returns)
183# Quantpedia data.
184# NOTE: IMPORTANT: Data order must be ascending (datewise)
185class QuantpediaFutures(PythonData):
186_last_update_date:Dict[Symbol, datetime.date] = {}
187@staticmethod
188def get_last_update_date() -> Dict[Symbol, datetime.date]:
189return QuantpediaFutures._last_update_date
190def GetSource(self, config, date, isLiveMode):
191 return SubscriptionDataSource("data.quantpedia.com/backtesting_data/futures/{0}.csv".format(config.Symbol.Value), SubscriptionTransportMedium.RemoteFile, FileFormat.Csv)
192def Reader(self, config, line, date, isLiveMode):
193 data = QuantpediaFutures()
194 data.Symbol = config.Symbol
195 
196 if not line[0].isdigit(): return None
197 split = line.split(';')
198 
199 data.Time = datetime.strptime(split[0], "%d.%m.%Y") + timedelta(days=1)
200 data['back_adjusted'] = float(split[1])
201 data['spliced'] = float(split[2])
202 data.Value = float(split[1])
203 if config.Symbol.Value not in QuantpediaFutures._last_update_date:
204     QuantpediaFutures._last_update_date[config.Symbol.Value] = datetime(1,1,1).date()
205 if data.Time.date() > QuantpediaFutures._last_update_date[config.Symbol.Value]:
206     QuantpediaFutures._last_update_date[config.Symbol.Value] = data.Time.date()
207 
208 return data
209# Custom fee model.
210class CustomFeeModel(FeeModel):
211def GetOrderFee(self, parameters):
212 fee = parameters.Security.Price * parameters.Order.AbsoluteQuantity * 0.00005
213 return OrderFee(CashAmount(fee, "USD"))