Cointegrated Cryptocurrency Portfolios

1from AlgorithmImports import *
2import numpy as np
3import statsmodels.api as sm
4#endregion
5class CointegratedCryptocurrencyPortfolios(QCAlgorithm):
6def Initialize(self):
7self.SetStartDate(2015, 1, 1)
8self.SetCash(100000)
9
10self.cryptos = [
11    "BTCUSD", # Bitcoin
12    "ETHUSD", # Ethereum
13    # "BCHUSD", # Bitcoin cash  # bchusd bitfinex quantconnect price ends in 2018.
14    "LTCUSD", # Litecoin
15]
16
17self.data = {}
18self.spread = []
19
20self.c = 0.5 # Constant for this strategy, however there are other possible values in source paper
21self.period = 21
22
23self.last_day = -1
24self.invested_long = None
25
26self.SetBrokerageModel(BrokerageName.Bitfinex)
27
28for crypto in self.cryptos:
29    data = self.AddCrypto(crypto, Resolution.Daily, Market.Bitfinex)
30    data.SetFeeModel(CustomFeeModel())
31    data.SetLeverage(5)
32    
33    self.data[crypto] = RollingWindow[float](self.period)
34def OnData(self, data):
35if self.last_day == self.Time.day: return
36self.last_day = self.Time.day
37
38for crypto in self.cryptos:
39    if crypto in data.Bars:
40        if data[crypto]:
41            price = data.Bars[crypto].Value
42            self.data[crypto].Add(price)
43
44used_cryptos = []            
45bitcoin_prices = None
46other_currencies_prices = []
47for crypto in self.cryptos:
48    if not self.data[crypto].IsReady:
49        return
50    else:
51        if crypto is "BTCUSD":
52            bitcoin_prices = np.array([x for x in self.data[crypto]])
53        else:
54            crypto_prices = [x for x in self.data[crypto]]
55            
56            if not crypto_prices[1:] == crypto_prices[:-1]: # If values in one list aren't same, then it can be use in MultipleLinearRegression
57                other_currencies_prices.append([x for x in self.data[crypto]])
58                used_cryptos.append(crypto)
59                
60if len(other_currencies_prices) == 0:
61    self.Liquidate()
62    return
63
64regression_model = self.MultipleLinearRegression(other_currencies_prices, bitcoin_prices)
65
66alpha = regression_model.params[0]
67betas = [regression_model.params[index] for index in range(len(regression_model.params)) if index != 0]
68
69beta_index = 0
70current_spread_value = 0
71for crypto in self.cryptos:
72    if crypto is "BTCUSD":
73        current_spread_value = current_spread_value + self.data[crypto][0]
74    elif crypto in used_cryptos:
75        current_speard_value = self.data[crypto][0] * betas[beta_index]
76        beta_index = beta_index + 1
77        
78self.spread.append(current_spread_value)
79
80if len(self.spread) < self.period: # We need at least one month values of spread
81    return
82
83threshold_long = np.mean(self.spread) - self.c * np.std(self.spread)
84threshold_short = np.mean(self.spread) + self.c * np.std(self.spread)
85
86if self.invested_long is None:
87    if current_spread_value < threshold_long: # long or exit short
88        self.InvestLong(betas, used_cryptos)
89        self.invested_long = True
90    elif current_spread_value > threshold_short: # short or exit long
91        self.InvestShort(betas, used_cryptos)
92        self.invested_long = False
93else:
94    if current_spread_value < threshold_long and self.invested_long: # long or exit short
95        self.invested_long = True
96        self.InvestLong(betas, used_cryptos)
97    elif current_spread_value > threshold_short and not self.invested_long: # short or exit long
98        self.invested_long = False
99        self.InvestShort(betas, used_cryptos)
100  
101def InvestLong(self, betas, used_cryptos):
102beta_index = 0
103for crypto in self.cryptos:
104    if self.Portfolio[crypto].Invested:
105        self.Liquidate(crypto)
106    if crypto is "BTCUSD":
107        self.MarketOrder(crypto, 1)
108    elif crypto in used_cryptos:
109        self.MarketOrder(crypto, betas[beta_index])
110        beta_index = beta_index + 1
111
112def InvestShort(self, betas, used_cryptos):
113beta_index = 0
114for crypto in self.cryptos:
115    if self.Portfolio[crypto].Invested:
116        self.Liquidate(crypto)
117    if crypto is "BTCUSD":
118        self.MarketOrder(crypto, -1)
119    elif crypto in used_cryptos:
120        self.MarketOrder(crypto, -betas[beta_index])
121        beta_index = beta_index + 1
122
123def MultipleLinearRegression(self, x, y):
124x = np.array(x).T
125x = sm.add_constant(x)
126result = sm.OLS(endog=y, exog=x).fit()
127return result  
128                    
129# Custom fee model.
130class CustomFeeModel(FeeModel):
131def GetOrderFee(self, parameters):
132fee = parameters.Security.Price * parameters.Order.AbsoluteQuantity * 0.00005
133return OrderFee(CashAmount(fee, "USD"))