The strategy trades 7 futures using MACD for trend signals, adjusting positions based on volatility thresholds, with equal weighting and daily rebalancing to adapt to changing market conditions.
ASSET CLASS: CFDs, futures | REGION: Global | FREQUENCY:
Daily | MARKET: bonds, commodities, currencies, equities | KEYWORD: Time, Series, Momentum, Volatility, Filters, Futures, Markets
I. STRATEGY IN A NUTSHELL
The strategy trades seven futures using MACD signals combined with a volatility filter. In calm markets, it follows trends, while in volatile markets, it reverses signals. Positions are equally weighted and rebalanced daily.
II. ECONOMIC RATIONALE
Research shows trend-following works best in low volatility but fails when volatility spikes, as prices often reverse. By switching rules across regimes, the strategy adapts to changing market conditions.
III. SOURCE PAPER
Volatility Filters for Asset Management: An Application to Managed Futures [Click to Open PDF]
Dunis, Miao, Liverpool John Moores University – Professor of Banking and Finance; Director of CIBEF; Associate Researcher with CIBEF; PhD Candidate, Liverpool John Moores University
<Abstract>
Technical trading rules are known to perform poorly in periods when volatility is high. The objective of this paper is to study whether addition of volatility filters can improve model performance. Different from previous studies on technical trading rules which base their findings from an academic perspective, this paper tries to relate to the real world business: two portfolios, which are highly correlated with a managed futures index and a currency traders’ benchmark index are formed to replicate the performance of the typical managed futures and managed currency funds. The volatility filters proposed are then applied directly to these two portfolios with the hope that proposed techniques will then have both academic and industrial significance. Two volatility filters are proposed, namely a “no-trade” filter where all market positions are closed in volatile periods, and a “reverse” filter where signals from a simple Moving Average Convergence and Divergence (MACD) are reversed if market volatility is higher than a given threshold. To assess the consistency of model performance, the whole period (04/01/1999 to 31/12/2004) is split into 3 sub-periods. Our results show that the addition of the two volatility filters adds value to the models performance in terms of annualised return, maximum drawdown, risk-adjusted Sharpe ratio and Calmar ratio in all the 3 sub-periods.
IV. BACKTEST PERFORMANCE
| Annualised Return | 5.47% |
| Volatility | 5.51% |
| Beta | 0.069 |
| Sharpe Ratio | 0.99 |
| Sortino Ratio | -0.793 |
| Maximum Drawdown | -4.88% |
| Win Rate | 48% |
V. FULL PYTHON CODE
import numpy as np
from AlgorithmImports import *
from collections import deque
class TimeSeriesMomentumVolatilityFilters(QCAlgorithm):
def Initialize(self):
self.SetStartDate(2010, 1, 1)
self.SetCash(100000)
self.symbols = [
'ICE_DX1', # US Dollar Index Futures, Continuous Contract #1
'CME_TY1', # 10 Yr Note Futures, Continuous Contract #1
'CME_ES1', # E-mini S&P 500 Futures, Continuous Contract #1
'CME_EC1', # Euro FX Futures, Continuous Contract #1
'CME_BP1', # British Pound Futures, Continuous Contract #1
'CME_HG1' # Copper Futures, Continuous Contract
]
self.period = 12*21
self.vol_period = 21
self.SetWarmUp(self.period)
self.optimalization_count = 100
self.data = {}
self.macd = {}
self.macd_signal = {}
self.volatility = {}
for symbol in self.symbols:
data = self.AddData(QuantpediaFutures, symbol, Resolution.Daily)
data.SetFeeModel(CustomFeeModel())
self.data[symbol] = deque(maxlen=self.period+self.vol_period+1)
self.volatility[symbol] = deque(maxlen=self.period)
self.macd_signal[symbol] = deque(maxlen=self.period)
self.macd['ICE_DX1'] = self.MACD('ICE_DX1', 1, 250, 1)
self.macd['CME_TY1'] = self.MACD('CME_TY1', 1, 250, 1)
self.macd['CME_HG1'] = self.MACD('CME_HG1', 1, 250, 1)
self.macd['CME_EC1'] = self.MACD('CME_EC1', 1, 61, 1)
self.macd['CME_BP1'] = self.MACD('CME_BP1', 1, 61, 1)
self.macd['CME_ES1'] = self.MACD('CME_ES1', 3, 250, 1)
def OnData(self, data):
for symbol in self.data:
symbol_obj = self.Symbol(symbol)
if symbol_obj in data.Keys:
if data[symbol_obj]:
price = data[symbol_obj].Value
if price != 0:
self.data[symbol].append(price)
if self.IsWarmingUp: return
self.Liquidate()
# Optimalization
optimal_treshold = {}
for symbol in self.symbols:
if self.securities[symbol].get_last_data() and self.time.date() > QuantpediaFutures.get_last_update_date()[symbol]:
self.liquidate(symbol)
continue
macd = self.macd[symbol]
closes = np.array([x for x in self.data[symbol]])
if macd.IsReady and len(closes) >= self.vol_period:
# MACD SIGNAL = if short term MA is UNDER long term MA == 1 else -1
self.macd_signal[symbol].append(1 if macd.Slow > macd.Fast else -1)
daily_return = closes[:-1] / closes[1:] - 1
vol = np.std(daily_return) * np.sqrt(252)
self.volatility[symbol].append(vol)
if len(closes) == self.data[symbol].maxlen and len(self.macd_signal[symbol]) == self.macd_signal[symbol].maxlen:
values = np.array(closes)
daily_changes = (values[1:] - values[:-1]) / values[:-1]
vol_min = min(self.volatility[symbol])
vol_max = max(self.volatility[symbol])
vol_range = vol_max - vol_min
vol_step = vol_range / self.optimalization_count
avg_return = {}
treshhold_value = vol_min
while treshhold_value <= vol_max:
vol_vector = np.array([-1 if x <= treshhold_value else 1 for x in self.volatility[symbol]])
returns = vol_vector * self.macd_signal[symbol] * daily_changes[-self.period:]
avg_return[treshhold_value] = np.average([x for x in returns])
treshhold_value += vol_step
optimal_treshold[symbol] = max(avg_return, key=avg_return.get)
if len(optimal_treshold) == 0: return
# Trading
count = len(optimal_treshold)
for symbol, threshold in optimal_treshold.items():
vol = self.volatility[symbol][-1]
if symbol in data and data[symbol]:
# low volatility enviroment
if vol <= threshold:
if self.macd[symbol].Fast > self.macd[symbol].Slow:
self.SetHoldings(symbol, 1/count)
else:
self.SetHoldings(symbol, -1/count)
# high volatility enviroment
else:
if self.macd[symbol].Fast > self.macd[symbol].Slow:
self.SetHoldings(symbol, -1/count)
else:
self.SetHoldings(symbol, 1/count)
# Quantpedia data.
# NOTE: IMPORTANT: Data order must be ascending (datewise)
class QuantpediaFutures(PythonData):
_last_update_date:Dict[Symbol, datetime.date] = {}
@staticmethod
def get_last_update_date() -> Dict[Symbol, datetime.date]:
return QuantpediaFutures._last_update_date
def GetSource(self, config, date, isLiveMode):
return SubscriptionDataSource("data.quantpedia.com/backtesting_data/futures/{0}.csv".format(config.Symbol.Value), SubscriptionTransportMedium.RemoteFile, FileFormat.Csv)
def Reader(self, config, line, date, isLiveMode):
data = QuantpediaFutures()
data.Symbol = config.Symbol
if not line[0].isdigit(): return None
split = line.split(';')
data.Time = datetime.strptime(split[0], "%d.%m.%Y") + timedelta(days=1)
data['back_adjusted'] = float(split[1])
data['spliced'] = float(split[2])
data.Value = float(split[1])
if config.Symbol.Value not in QuantpediaFutures._last_update_date:
QuantpediaFutures._last_update_date[config.Symbol.Value] = datetime(1,1,1).date()
if data.Time.date() > QuantpediaFutures._last_update_date[config.Symbol.Value]:
QuantpediaFutures._last_update_date[config.Symbol.Value] = data.Time.date()
return data
# Custom fee model.
class CustomFeeModel(FeeModel):
def GetOrderFee(self, parameters):
fee = parameters.Security.Price * parameters.Order.AbsoluteQuantity * 0.00005
return OrderFee(CashAmount(fee, "USD"))VI. Backtest Performance
