大宗商品 · 全球 · 期货 · 每月 · 全球 Free

交易商品日历价差

登录后收藏

学术论文

作者商品期货曲线的配对交易 [点击查看论文]

论文链接http://herkules.oulu.fi/thesis/nbnfioulu-201211141056.pdf

策略概要

该策略的目标是20种商品期货，使用前12个月的合约进行信号生成和交易。每个月，投资者通过计算前五个合约之间的平均差值来评估期货曲线的形状。正结果表示现货溢价，而负结果表示期货溢价。

在现货溢价中，投资者在现货溢价最高的合约（差值最大）中建立多头头寸，在现货溢价最低的合约（差值最小）中建立空头头寸。在期货溢价中，该过程相反，在差值最大的合约中建立空头头寸，在差值最小的合约中建立多头头寸。

投资组合在20种商品中平均分配权重，并每月进行再平衡，系统地利用期货曲线结构的差异来寻找交易机会。

II. 策略合理性

学术研究将展期收益确定为商品回报的主要驱动因素。该策略通过做多现货溢价的商品和做空期货溢价的商品来捕捉这一点，同时通过曲线上的对冲合约来对冲头寸，以降低投资组合的波动性。

回测表现

波动率2.01%

夏普比率3.09

胜率51%

完整 Python 代码

1import numpy as np
2class TradingCommodityCalendarSpreads(QCAlgorithm):
3
4def Initialize(self):
5self.SetStartDate(2010, 1, 1)
6self.SetCash(100000)
7   
8# 1st contract and the number of contracts to approximately 12 months away.
9self.contracts = {
10                "CHRIS/CME_S" :  8,  # Soybean Futures, Continuous Contract
11                "CHRIS/CME_W"  : 6,  # Wheat Futures, Continuous Contract
12                "CHRIS/CME_SM" : 9,  # Soybean Meal Futures, Continuous Contract
13                "CHRIS/CME_BO" : 9,  # Soybean Oil Futures, Continuous Contract
14                "CHRIS/CME_C" :  6,  # Corn Futures, Continuous Contract
15                "CHRIS/CME_O" :  6,  # Oats Futures, Continuous Contract
16                "CHRIS/CME_LC" : 7,  # Live Cattle Futures, Continuous Contract
17                "CHRIS/CME_FC" : 7,  # Feeder Cattle Futures, Continuous Contract
18                "CHRIS/CME_LN" : 9,  # Lean Hog Futures, Continuous Contract
19                "CHRIS/CME_GC" : 9,  # Gold Futures, Continuous Contract
20                "CHRIS/CME_SI" : 9,  # Silver Futures, Continuous Contract
21                "CHRIS/CME_PL" : 7,  # Platinum Futures, Continuous Contract
22                "CHRIS/CME_CL" : 12, # Crude Oil Futures, Continuous Contract
23                "CHRIS/CME_HG" : 13, # Copper Futures, Continuous Contract
24                "CHRIS/CME_LB" : 7,  # Random Length Lumber Futures, Continuous Contract
25                "CHRIS/CME_NG" : 12, # Natural Gas (Henry Hub) Physical Futures, Continuous Contract
26                "CHRIS/CME_PA" : 7,  # Palladium Futures, Continuous Contract 
27                "CHRIS/CME_RR" : 7,  # Rough Rice Futures, Continuous Contract
28                # "CHRIS/CME_CU" : 13, # Chicago Ethanol (Platts) Futures
29                "CHRIS/CME_DA" : 13, # Class III Milk Futures
30                
31                "CHRIS/ICE_CC" : 6,  # Cocoa Futures, Continuous Contract 
32                "CHRIS/ICE_CT" : 6,  # Cotton No. 2 Futures, Continuous Contract
33                "CHRIS/ICE_KC" : 6,  # Coffee C Futures, Continuous Contract
34                "CHRIS/ICE_O" :  13, # Heating Oil Futures, Continuous Contract
35                "CHRIS/ICE_OJ" : 7,  # Orange Juice Futures, Continuous Contract
36                "CHRIS/ICE_SB" : 5   # Sugar No. 11 Futures, Continuous Contract
37                }
38self.rebalance_flag = True
39
40for future, future_count in self.contracts.items():
41    for index in range(1, future_count + 1):
42        contract = future + str(index)
43
44        data = self.AddData(QuandlFutures, contract, Resolution.Daily)
45        data.SetFeeModel(CustomFeeModel(self))
46        data.SetLeverage(5)
47
48self.Schedule.On(self.DateRules.MonthStart('CHRIS/CME_S1'), self.TimeRules.AfterMarketOpen('CHRIS/CME_S1'), self.Rebalance)
49
50def Rebalance(self):
51self.rebalance_flag = True
52
53def OnData(self, data):
54if not self.rebalance_flag:
55    return
56self.rebalance_flag = False
57
58long = []
59short = []
60for future, future_count in self.contracts.items():
61    # curve_shape = sum( np.diff( [ self.Securities[future + str(index)].Price for index in range(1, 6) if self.Securities.ContainsKey(future + str(index)) ] ) )
62    curve_shape = sum( np.diff( [ data[future + str(index)].Value for index in range(1, 6) if (future + str(index)) in data ] ) )
63    curve_shape /= future_count
64    if curve_shape != 0:
65        # diff = np.diff( [ self.Securities[future + str(index)].Price for index in range(1, future_count + 1) if self.Securities.ContainsKey(future + str(index)) ] )
66        diff = np.diff( [ data[future + str(index)].Value for index in range(1, future_count + 1) if (future + str(index)) in data ] )
67        abs_diff = [abs(x) for x in diff]
68        max_diff_index = abs_diff.index(max(abs_diff))
69        min_diff_index = abs_diff.index(min(abs_diff))
70        
71        # Offset index by 2 to get right symbols to trade.
72        max_diff_index += 2
73        min_diff_index += 2
74        
75        if curve_shape > 0:
76            # Backwardation.
77            long.append(future + str(max_diff_index))
78            short.append(future + str(min_diff_index))
79        else:
80            # Contango.
81            long.append(future + str(min_diff_index))
82            short.append(future + str(max_diff_index))
83
84# Trade execution.
85invested = [x.Key.Value for x in self.Portfolio if x.Value.Invested]
86for symbol in invested:
87    if symbol not in long + short:
88        self.Liquidate(symbol)
89
90weight = 1 / len(self.contracts)
91
92for symbol in long:
93    if self.Securities[symbol].Price != 0:
94        self.SetHoldings(symbol, weight)
95for symbol in short:
96    if self.Securities[symbol].Price != 0:
97        self.SetHoldings(symbol, -weight)
98# Quandl free data
99class QuandlFutures(PythonQuandl):
100def __init__(self):
101self.ValueColumnName = "settle"
102# Custom fee model.
103class CustomFeeModel(FeeModel):
104def GetOrderFee(self, parameters):
105fee = parameters.Security.Price * parameters.Order.AbsoluteQuantity * 0.00005
106return OrderFee(CashAmount(fee, "USD"))