Crypto · Global · cryptos · Weekly · Global Free

Time-Series Momentum Factor in Cryptocurrencies

Academic paper

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

AuthorsStefan Hubrich

Institute

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

Paper linkhttps://ssrn.com/abstract=3055498

Strategy in a nutshell

The strategy trades 11 cryptocurrencies using a weekly momentum signal based on prior returns, standardized by z-scores. Positions are equally weighted with up to 10% total exposure, focusing on long-only implementation due to shorting limits. Portfolios are rebalanced weekly, with optional risk-based weighting.

Economic rationale

Weekly momentum captures short-term trends in volatile crypto markets, producing strong positive alphas. Results show momentum signals effectively predict returns, confirming that even in high-volatility environments, disciplined momentum strategies can deliver superior performance.

Backtest performance

Annualised return6.3%

Volatility8.1%

Beta0.035

Sharpe ratio0.78

Sortino ratio0.487

Win rate52%

Full Python code

1from AlgorithmImports import *
2import numpy as np
3#endregion
4class TimeSeriesMomentumCryptocurrencies(QCAlgorithm):
5def Initialize(self):
6self.SetStartDate(2015, 1, 1)
7self.SetCash(100000)
8self.symbols = ['BTCUSD', 'ETCUSD', 'ETHUSD', 'LTCUSD', 'XMRUSD', 'ZECUSD']
9self.data = {}
10self.percentage_traded = 0.1
11
12for symbol in self.symbols:
13    data = self.AddCrypto(symbol, Resolution.Daily, Market.Bitfinex)
14    data.SetFeeModel(CustomFeeModel())
15    data.SetLeverage(10)
16    self.data[symbol] = RollingWindow[float](5)
17
18def OnData(self, data):
19for symbol in self.data:
20    symbol_obj = self.Symbol(symbol)
21    if symbol_obj in data.Bars and data[symbol_obj]:
22        self.data[symbol].Add(data[symbol_obj].Value)
23if self.Time.date().weekday() != 0:
24    return
25perf_vol = {}
26
27for symbol in self.symbols:
28    if self.data[symbol].IsReady:
29        prices = np.array([x for x in self.data[symbol]])
30        perf = prices[0] / prices[-1] - 1
31        
32        daily_returns = prices[:-1] / prices[1:] - 1
33        vol = np.std(daily_returns)
34        perf_vol[symbol] = (perf, vol)
35# Volatility weighting
36total_vol = sum([1 / x[1][1] for x in perf_vol.items()])
37if total_vol == 0: return
38weight = {}
39for symbol in perf_vol:
40    vol = perf_vol[symbol][1]
41    if vol != 0:
42        weight[symbol] = (1.0 / vol) / total_vol
43    else: 
44        weight[symbol] = 0
45# Trade execution.
46long = [x[0] for x in perf_vol.items() if x[1][0] > 0]
47invested = [x.Key.Value for x in self.Portfolio if x.Value.Invested]
48for symbol in invested:
49    if symbol not in long:
50        self.Liquidate(symbol)
51for symbol in long:
52    if symbol in data and data[symbol]:
53        self.SetHoldings(symbol, self.percentage_traded * weight[symbol])
54# Custom fee model.
55class CustomFeeModel(FeeModel):
56def GetOrderFee(self, parameters):
57fee = parameters.Security.Price * parameters.Order.AbsoluteQuantity * 0.00005
58return OrderFee(CashAmount(fee, "USD"))