加密货币 · 全球 · 加密货币 · 每周 · 全球 Free

加密货币中的套息因子

登录后收藏

学术论文

'Know When to Hodl Em, Know When to Fodl Em': An Investigation of Factor Based Investing in the Cryptocurrency Space

作者Stefan Hubrich

机构

T. Rowe Price Program for Charitable Giving
?T.Rowe Price

论文链接https://ssrn.com/abstract=3055498

策略概要

投资范围包括11种加密货币，其中原始套利指标定义为过去一周代币发行量的负和，除以该时期初的未偿代币数量。然后，该原始套利因子通过该货币的回报波动率进行标准化。投资组合采用等权重，总敞口限制为10%，每种货币的权重由标准化套利因子决定。当因子高于零时，给予正权重；当因子低于零时，给予负权重，从而允许净多头或净空头头寸。由于无法做空加密货币，因此采用只做多策略。投资组合每周重新平衡，论文中还讨论了另一种基于风险的加权方案。

II. 策略合理性

该论文分析了一个为期四年的加密货币数据集，尽管时间较短，但由于加密货币市场的高波动性，仍提供了足够的变化。论文研究了每周回报，投资组合每周重新平衡。高波动性是有利的，因为它允许使用一小部分可用资本进行有意义的分配，这对于无法利用杠杆的投资组合至关重要。论文测试了基于代币发行的套利因子是否比被动基准提供价值，发现它确实增加了价值，具有正的阿尔法和稳固的信息比率。套利指标使用标的回报的波动率进行标准化，因为它量化的是价格影响而非预期回报，同时考虑了获得该回报所需的风险。

回测表现

波动率7.5%

夏普比率2.19

索提诺比率1.077

胜率79%

完整 Python 代码

1from AlgorithmImports import *
2from typing import List, Dict
3class CarryFactorInCryptocurrencies(QCAlgorithm):
4def Initialize(self) -> None:
5 self.SetStartDate(2015, 1, 1)
6 self.SetCash(1_000_000)
7 
8 # Coin issuance seven days data and data about first of those seven days
9 self.period: int = 7
10 self.count_days: int = 1
11 self.percentage_traded: float = 0.1
12 
13 self.symbols: Dict[str, str] = {
14     'BTC' : 'BTCUSD',
15     'ETH' : 'ETHUSD', 
16     'LTC' : 'LTCUSD', 
17     'ETC' : 'ETCUSD',
18     'XMR' : 'XMRUSD',
19     'ZEC' : 'ZECUSD'
20 }
21 
22 self.data: Dict[str, SymbolData] = {}
23 self.SetBrokerageModel(BrokerageName.Bitfinex)  
24 for crypto, ticker in self.symbols.items():
25     data: Securities = self.AddCrypto(ticker, Resolution.Daily, Market.Bitfinex)
26     
27     self.AddData(CryptoNetworkData, crypto, Resolution.Daily)
28     self.data[crypto] = SymbolData(self.period)
29 
30 self.Settings.MinimumOrderMarginPortfolioPercentage = 0.
31     
32def OnData(self, data: Slice) -> None:
33 crypto_data_last_update_date: Dict[Symbol, datetime.date] = CryptoNetworkData.get_last_update_date()
34 # Store daily data.
35 for symbol in self.symbols:
36     if symbol in data and data[symbol]:
37         coin_issuance: float = data[symbol].Price
38         if coin_issuance != 0:
39             self.data[symbol].update(coin_issuance)
40 if self.Time.date().weekday() != 0:
41     return
42 if self.count_days == self.period:
43     self.count_days = 1
44 else:
45     self.count_days = self.count_days + 1
46     return
47 
48 long: List[str] = []
49 short: List[str] = []
50 
51 for crypto, ticker in self.symbols.items():
52     # network data is still comming in
53     if self.Securities[crypto].GetLastData() and self.Time.date() > crypto_data_last_update_date[crypto]:
54         continue
55     
56     if self.data[crypto].is_ready():
57         carry_metric: float = self.data[crypto].carry_metric()
58         
59         if carry_metric > 0.:
60             long.append(self.Symbol(ticker))
61         else:
62             short.append(self.Symbol(ticker))
63 
64 # Trade Execution
65 invested: List[Symbol] = [x.Key for x in self.Portfolio if x.Value.Invested]
66 for symbol in invested:
67     if symbol not in long + short:
68         self.Liquidate(symbol)
69 for i, portfolio in enumerate([long, short]):
70     for symbol in portfolio:
71         if symbol in data and data[symbol]:
72             self.SetHoldings(symbol, round(((-1) ** i) / len(portfolio) * self.percentage_traded, 2))
73 
74class SymbolData():
75def __init__(self, period: int) -> None:
76 self.coin_issuance_data: RollingWindow = RollingWindow[float](period)
77 
78def update(self, value: float) -> None:
79 self.coin_issuance_data.Add(value)
80 
81def carry_metric(self) -> float:
82 issuance_data: List[float] = list(self.coin_issuance_data)
83 return sum(issuance_data) / issuance_data[-1]
84 
85def is_ready(self) -> bool:
86 return self.coin_issuance_data.IsReady
87 
88# Crypto network data.
89# NOTE: IMPORTANT: Data order must be ascending (datewise)
90# Data source: https://coinmetrics.io/community-network-data/
91class CryptoNetworkData(PythonData):
92_last_update_date: Dict[Symbol, datetime.date] = {}
93@staticmethod
94def get_last_update_date() -> Dict[Symbol, datetime.date]:
95return CryptoNetworkData._last_update_date
96def GetSource(self, config: SubscriptionDataConfig, date: datetime, isLiveMode: bool) -> SubscriptionDataSource:
97 return SubscriptionDataSource(f"data.quantpedia.com/backtesting_data/crypto/{config.Symbol.Value}_network_data.csv", SubscriptionTransportMedium.RemoteFile, FileFormat.Csv)
98# File exmaple:
99# date,AdrActCnt,AdrBal1in100KCnt,AdrBal1in100MCnt,AdrBal1in10BCnt,AdrBal1in10KCnt,AdrBal1in10MCnt,AdrBal1in1BCnt,AdrBal1in1KCnt,AdrBal1in1MCnt,AdrBalCnt,AdrBalNtv0.001Cnt,AdrBalNtv0.01Cnt,AdrBalNtv0.1Cnt,AdrBalNtv100Cnt,AdrBalNtv100KCnt,AdrBalNtv10Cnt,AdrBalNtv10KCnt,AdrBalNtv1Cnt,AdrBalNtv1KCnt,AdrBalNtv1MCnt,AdrBalUSD100Cnt,AdrBalUSD100KCnt,AdrBalUSD10Cnt,AdrBalUSD10KCnt,AdrBalUSD10MCnt,AdrBalUSD1Cnt,AdrBalUSD1KCnt,AdrBalUSD1MCnt,AssetEODCompletionTime,BlkCnt,BlkSizeMeanByte,BlkWghtMean,BlkWghtTot,CapAct1yrUSD,CapMVRVCur,CapMVRVFF,CapMrktCurUSD,CapMrktFFUSD,CapRealUSD,DiffLast,DiffMean,FeeByteMeanNtv,FeeMeanNtv,FeeMeanUSD,FeeMedNtv,FeeMedUSD,FeeTotNtv,FeeTotUSD,FlowInExNtv,FlowInExUSD,FlowOutExNtv,FlowOutExUSD,FlowTfrFromExCnt,HashRate,HashRate30d,IssContNtv,IssContPctAnn,IssContPctDay,IssContUSD,IssTotNtv,IssTotUSD,NDF,NVTAdj,NVTAdj90,NVTAdjFF,NVTAdjFF90,PriceBTC,PriceUSD,ROI1yr,ROI30d,RevAllTimeUSD,RevHashNtv,RevHashRateNtv,RevHashRateUSD,RevHashUSD,RevNtv,RevUSD,SER,SplyAct10yr,SplyAct180d,SplyAct1d,SplyAct1yr,SplyAct2yr,SplyAct30d,SplyAct3yr,SplyAct4yr,SplyAct5yr,SplyAct7d,SplyAct90d,SplyActEver,SplyActPct1yr,SplyAdrBal1in100K,SplyAdrBal1in100M,SplyAdrBal1in10B,SplyAdrBal1in10K,SplyAdrBal1in10M,SplyAdrBal1in1B,SplyAdrBal1in1K,SplyAdrBal1in1M,SplyAdrBalNtv0.001,SplyAdrBalNtv0.01,SplyAdrBalNtv0.1,SplyAdrBalNtv1,SplyAdrBalNtv10,SplyAdrBalNtv100,SplyAdrBalNtv100K,SplyAdrBalNtv10K,SplyAdrBalNtv1K,SplyAdrBalNtv1M,SplyAdrBalUSD1,SplyAdrBalUSD10,SplyAdrBalUSD100,SplyAdrBalUSD100K,SplyAdrBalUSD10K,SplyAdrBalUSD10M,SplyAdrBalUSD1K,SplyAdrBalUSD1M,SplyAdrTop100,SplyAdrTop10Pct,SplyAdrTop1Pct,SplyCur,SplyExpFut10yr,SplyFF,SplyMiner0HopAllNtv,SplyMiner0HopAllUSD,SplyMiner1HopAllNtv,SplyMiner1HopAllUSD,TxCnt,TxCntSec,TxTfrCnt,TxTfrValAdjNtv,TxTfrValAdjUSD,TxTfrValMeanNtv,TxTfrValMeanUSD,TxTfrValMedNtv,TxTfrValMedUSD,VelCur1yr,VtyDayRet180d,VtyDayRet30d
100# 2009-01-09,19,19,19,19,19,19,19,19,19,19,19,19,19,0,0,19,0,19,0,0,0,0,0,0,0,0,0,0,1614334886,19,215,860,16340,0,0,0,0,0,0,1,1,0,0,0,0,0,0,0,0,0,0,0,0,9.44495122962963E-7,0,950,36500,100,0,950,0,1,0,0,0,0,1,0,0,0,0,11641.53218269,1005828380.584716757433,0,0,950,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,950,950,950,950,950,950,950,950,950,950,950,950,950,0,0,0,0,0,0,0,0,0,0,0,0,0,950,50,50,950,17070250,950,1000,0,1000,0,0,0,0,0,0,0,0,0,0,0,0,0
101def Reader(self, config: SubscriptionDataConfig, line: str, date: datetime, isLiveMode: bool) -> BaseData:
102 data: CryptoNetworkData = CryptoNetworkData()
103 data.Symbol = config.Symbol
104 try:
105     cols:str = ['SplyCur']
106     if not line[0].isdigit():
107         header_split = line.split(',')
108         self.col_index = [header_split.index(x) for x in cols]
109         return None
110     split = line.split(',')
111     
112     data.Time = datetime.strptime(split[0], "%Y-%m-%d") + timedelta(days=1)
113     for i, col in enumerate(cols):
114         data[col] = float(split[self.col_index[i]])
115     data.Value = float(split[self.col_index[0]])
116     if config.Symbol.Value not in CryptoNetworkData._last_update_date:
117         CryptoNetworkData._last_update_date[config.Symbol.Value] = datetime(1,1,1).date()
118     if data.Time.date() > CryptoNetworkData._last_update_date[config.Symbol.Value]:
119         CryptoNetworkData._last_update_date[config.Symbol.Value] = data.Time.date()
120     
121 except:
122     return None
123 
124 return data