债券 · 全球 · 债券、差价合约、期货、互换 · 每日 · 全球 Free

国债拍卖策略中的季节性

学术论文

作者流动市场中的预期和重复冲击 [点击查看论文]

论文链接http://personal.lse.ac.uk/loud/Shocks.pdf

策略概要

该策略涉及围绕拍卖交易美国国债。在拍卖前十天，投资者做空2年期国债，做多久期匹配的10年期国债和6个月期国库券。零投资组合持有到拍卖日。在拍卖日，投资者反转头寸，进行相反的交易，并持有接下来的十天。该策略也可以使用期货或差价合约进行模拟。目的是利用与拍卖动态相关的价格波动，从拍卖前后的价格波动中获益。

II. 策略合理性

回测表现

波动率6.75%

夏普比率0.84

索提诺比率-0.584

胜率53%

完整 Python 代码

1from AlgorithmImports import *
2from pandas.tseries.offsets import BDay
3from typing import List
4from datetime import datetime
5#endregion
6class SeasonalityTreasuryAuctions(QCAlgorithm):
7def initialize(self):
8 self.set_start_date(2000, 1, 1)
9 self.set_cash(100000)
10 
11 data: Security = self.add_data(QuantpediaFutures, 'CME_TY1', Resolution.DAILY)
12 data.set_fee_model(CustomFeeModel())
13 self.symbol: Symbol = data.symbol
14 # Auction days are estimated to happen either on Thrusday after second Wednesday of the month
15 # Secondary Source: https://home.treasury.gov/
16 csv_string_file: str = self.download('data.quantpedia.com/backtesting_data/calendar/treasury_auction_dates.csv')
17 dates: List[str] = csv_string_file.split('\r\n')
18 self.auction_days: List[datetime.date] = [(datetime.strptime(x, "%Y-%m-%d") + BDay(1)).date() for x in dates]     # treasury auction date closes
19 
20 self.holding_days: int = 0
21 self.days_to_hold: int = 2
22
23def on_data(self, data: Slice) -> None:
24 if self.time.date() >= QuantpediaFutures.get_last_update_date()[self.symbol.value]:
25     self.liquidate()
26     return
27 # auction day close
28 if self.time.date() in self.auction_days:
29     self.set_holdings(self.symbol, 1)
30     return
31 
32 # liquidate
33 if self.portfolio.invested:
34     self.holding_days += 1
35     if self.holding_days == self.days_to_hold:
36         self.holding_days = 0
37         self.liquidate(self.symbol)
38 
39# custom fee model
40class CustomFeeModel(FeeModel):
41def GetOrderFee(self, parameters):
42 fee = parameters.Security.Price * parameters.Order.AbsoluteQuantity * 0.00005
43 return OrderFee(CashAmount(fee, "USD"))
44 
45# quantpedia data
46# NOTE: IMPORTANT: Data order must be ascending (datewise)
47class QuantpediaFutures(PythonData):
48_last_update_date: Dict[str, datetime.date] = {}
49@staticmethod
50def get_last_update_date() -> Dict[str, datetime.date]:
51return QuantpediaFutures._last_update_date
52def GetSource(self, config:SubscriptionDataConfig, date:datetime, isLiveMode:bool) -> SubscriptionDataSource:
53 return SubscriptionDataSource("data.quantpedia.com/backtesting_data/futures/{0}.csv".format(config.Symbol.Value), SubscriptionTransportMedium.RemoteFile, FileFormat.Csv)
54def Reader(self, config:SubscriptionDataConfig, line:str, date:datetime, isLiveMode:bool) -> BaseData:
55 data = QuantpediaFutures()
56 data.Symbol = config.Symbol
57 
58 if not line[0].isdigit(): return None
59 split = line.split(';')
60 
61 data.Time = datetime.strptime(split[0], "%d.%m.%Y") + timedelta(days=1)
62 data['back_adjusted'] = float(split[1])
63 data['spliced'] = float(split[2])
64 data.Value = float(split[1])
65 # store last update date
66 if config.Symbol.Value not in QuantpediaFutures._last_update_date:
67     QuantpediaFutures._last_update_date[config.Symbol.Value] = datetime(1,1,1).date()
68 if data.Time.date() > QuantpediaFutures._last_update_date[config.Symbol.Value]:
69     QuantpediaFutures._last_update_date[config.Symbol.Value] = data.Time.date()
70 return data