Overall Statistics |
Total Trades
521
Average Win
0.65%
Average Loss
-0.61%
Compounding Annual Return
1.111%
Drawdown
17.300%
Expectancy
0.102
Net Profit
16.823%
Sharpe Ratio
-0.102
Sortino Ratio
-0.097
Probabilistic Sharpe Ratio
0.007%
Loss Rate
47%
Win Rate
53%
Profit-Loss Ratio
1.08
Alpha
0.001
Beta
-0.073
Annual Standard Deviation
0.054
Annual Variance
0.003
Information Ratio
-0.56
Tracking Error
0.163
Treynor Ratio
0.076
Total Fees
$641.65
Estimated Strategy Capacity
$0
Lowest Capacity Asset
CME_TY1.QuantpediaFutures 2S
Portfolio Turnover
2.43%
|
#region imports from AlgorithmImports import * from dateutil.relativedelta import relativedelta #endregion # Bond yields class QuandlAAAYield(PythonQuandl): def __init__(self): self.ValueColumnName = 'BAMLC0A1CAAAEY' class QuandlHighYield(PythonQuandl): def __init__(self): self.ValueColumnName = 'BAMLH0A0HYM2EY' class LastDateHandler(): _last_update_date:Dict[Symbol, datetime.date] = {} @staticmethod def get_last_update_date() -> Dict[Symbol, datetime.date]: return LastDateHandler._last_update_date # Quantpedia monthly custom data. # NOTE: IMPORTANT: Data order must be ascending (datewise) class QuantpediaMonthlyData(PythonData): def GetSource(self, config: SubscriptionDataConfig, date: datetime, isLiveMode: bool) -> SubscriptionDataSource: return SubscriptionDataSource(f'data.quantpedia.com/backtesting_data/economic/{config.Symbol.Value}.csv', SubscriptionTransportMedium.RemoteFile, FileFormat.Csv) def Reader(self, config: SubscriptionDataConfig, line: str, date: datetime, isLiveMode: bool) -> BaseData: data = QuantpediaMonthlyData() data.Symbol = config.Symbol if not line[0].isdigit(): return None split = line.split(';') data.Time = datetime.strptime(split[0], "%Y-%m-%d") + relativedelta(months=1) data['pe_ratio'] = float(split[1]) data.Value = float(split[1]) if config.Symbol.Value not in LastDateHandler._last_update_date: LastDateHandler._last_update_date[config.Symbol.Value] = datetime(1,1,1).date() if data.Time.date() > LastDateHandler._last_update_date[config.Symbol.Value]: LastDateHandler._last_update_date[config.Symbol.Value] = data.Time.date() return data # Country PE data # NOTE: IMPORTANT: Data order must be ascending (date-wise) from dateutil.relativedelta import relativedelta class CountryPE(PythonData): def GetSource(self, config: SubscriptionDataConfig, date: datetime, isLiveMode: bool) -> SubscriptionDataSource: return SubscriptionDataSource("data.quantpedia.com/backtesting_data/economic/country_pe.csv", SubscriptionTransportMedium.RemoteFile, FileFormat.Csv) def Reader(self, config: SubscriptionDataConfig, line: str, date: datetime, isLiveMode: bool) -> BaseData: data = CountryPE() data.Symbol = config.Symbol if not line[0].isdigit(): return None split = line.split(';') data.Time = datetime.strptime(split[0], "%Y") + relativedelta(years=1) self.symbols = ['Argentina','Australia','Austria','Belgium','Brazil','Canada','Chile','China','Egypt','France','Germany','Hong Kong','India','Indonesia','Ireland','Israel','Italy','Japan','Malaysia','Mexico','Netherlands','New Zealand','Norway','Philippines','Poland','Russia','Saudi Arabia','Singapore','South Africa','South Korea','Spain','Sweden','Switzerland','Taiwan','Thailand','Turkey','United Kingdom','United States'] index = 1 for symbol in self.symbols: data[symbol] = float(split[index]) index += 1 data.Value = float(split[1]) if config.Symbol not in LastDateHandler._last_update_date: LastDateHandler._last_update_date[config.Symbol] = datetime(1,1,1).date() if data.Time.date() > LastDateHandler._last_update_date[config.Symbol]: LastDateHandler._last_update_date[config.Symbol] = data.Time.date() return data # Quantpedia PE ratio data. # NOTE: IMPORTANT: Data order must be ascending (datewise) class QuantpediaPERatio(PythonData): def GetSource(self, config: SubscriptionDataConfig, date: datetime, isLiveMode: bool) -> SubscriptionDataSource: return SubscriptionDataSource("data.quantpedia.com/backtesting_data/economic/{0}.csv".format(config.Symbol.Value), SubscriptionTransportMedium.RemoteFile, FileFormat.Csv) def Reader(self, config: SubscriptionDataConfig, line: str, date: datetime, isLiveMode: bool) -> BaseData: data = QuantpediaPERatio() data.Symbol = config.Symbol if not line[0].isdigit(): return None split = line.split(';') data.Time = datetime.strptime(split[0], "%Y-%m-%d") + timedelta(days=1) data['pe_ratio'] = float(split[1]) data.Value = float(split[1]) if config.Symbol.Value not in LastDateHandler._last_update_date: LastDateHandler._last_update_date[config.Symbol.Value] = datetime(1,1,1).date() if data.Time.date() > LastDateHandler._last_update_date[config.Symbol.Value]: LastDateHandler._last_update_date[config.Symbol.Value] = data.Time.date() return data # Quantpedia bond yield data. # NOTE: IMPORTANT: Data order must be ascending (datewise) class QuantpediaBondYield(PythonData): def GetSource(self, config: SubscriptionDataConfig, date: datetime, isLiveMode: bool) -> SubscriptionDataSource: return SubscriptionDataSource("data.quantpedia.com/backtesting_data/bond_yield/{0}.csv".format(config.Symbol.Value), SubscriptionTransportMedium.RemoteFile, FileFormat.Csv) def Reader(self, config: SubscriptionDataConfig, line: str, date: datetime, isLiveMode: bool) -> BaseData: data = QuantpediaBondYield() data.Symbol = config.Symbol if not line[0].isdigit(): return None split = line.split(',') data.Time = datetime.strptime(split[0], "%Y-%m-%d") + timedelta(days=1) data['yield'] = float(split[1]) data.Value = float(split[1]) if config.Symbol.Value not in LastDateHandler._last_update_date: LastDateHandler._last_update_date[config.Symbol.Value] = datetime(1,1,1).date() if data.Time.date() > LastDateHandler._last_update_date[config.Symbol.Value]: LastDateHandler._last_update_date[config.Symbol.Value] = data.Time.date() return data # Quantpedia data. # NOTE: IMPORTANT: Data order must be ascending (datewise) class QuantpediaFutures(PythonData): def GetSource(self, config: SubscriptionDataConfig, date: datetime, isLiveMode: bool) -> SubscriptionDataSource: return SubscriptionDataSource("data.quantpedia.com/backtesting_data/futures/{0}.csv".format(config.Symbol.Value), SubscriptionTransportMedium.RemoteFile, FileFormat.Csv) def Reader(self, config: SubscriptionDataConfig, line: str, date: datetime, isLiveMode: bool) -> BaseData: 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 LastDateHandler._last_update_date: LastDateHandler._last_update_date[config.Symbol.Value] = datetime(1,1,1).date() if data.Time.date() > LastDateHandler._last_update_date[config.Symbol.Value]: LastDateHandler._last_update_date[config.Symbol.Value] = data.Time.date() return data
# https://quantpedia.com/strategies/value-and-momentum-factors-across-asset-classes/ # # Create an investment universe containing investable asset classes (could be US large-cap, mid-cap stocks, US REITS, UK, Japan, Emerging market stocks, US treasuries, US Investment grade bonds, # US high yield bonds, Germany bonds, Japan bonds, US cash) and find a good tracking vehicle for each asset class (best vehicles are ETFs or index funds). Momentum ranking is done on price series. # Valuation ranking is done on adjusted yield measure for each asset class. E/P (Earning/Price) measure is used for stocks, and YTM (Yield-to-maturity) is used for bonds. US, Japan, and Germany # treasury yield are adjusted by -1%, US investment-grade bonds are adjusted by -2%, US High yield bonds are adjusted by -6%, emerging markets equities are adjusted by -1%, and US REITs are # adjusted by -2% to get unbiased structural yields for each asset class. Rank each asset class by 12-month momentum, 1-month momentum, and by valuation and weight all three strategies (25% weight # to 12m momentum, 25% weight to 1-month momentum, 50% weight to value strategy). Go long top quartile portfolio and go short bottom quartile portfolio. # # QC implementation changes: # - Country PB data ends in 2019. Last known value is used for further years calculations for the sake of backtest. #region imports from AlgorithmImports import * import data_tools from typing import List, Dict #endregion class ValueandMomentumFactorsacrossAssetClasses(QCAlgorithm): def Initialize(self) -> None: self.SetStartDate(2010, 1, 1) self.SetCash(100000) # investable asset, yield symbol, yield data access function, yield adjustment, reverse flag(PE -> EP) self.assets: List[Tuple] = [ ('SPY', 'SP500_EARNINGS_YIELD_MONTH', data_tools.QuantpediaMonthlyData, 0, True), # US large-cap ('MDY', 'MID_CAP_PE', data_tools.QuantpediaPERatio, 0, True), # US mid-cap stocks ('IYR', 'REITS_DIVIDEND_YIELD', data_tools.QuantpediaPERatio, -2, False), # US REITS - same csv data format as PERatio files ('EWU', 'United Kingdom', None, 0, True), # UK ('EWJ', 'Japan', None, 0, True), # Japan ('EEM', 'EMERGING_MARKET_PE', data_tools.QuantpediaPERatio, -1, True), # Emerging market stocks ('LQD', 'ML/AAAEY', data_tools.QuandlAAAYield, -2, False), # US Investment grade bonds ('HYG', 'ML/USTRI', data_tools.QuandlHighYield, -6, False), # US high yield bonds ('CME_TY1', 'US10YT', data_tools.QuantpediaBondYield, -1, False), # US bonds ('EUREX_FGBL1', 'DE10YT', data_tools.QuantpediaBondYield, -1, False), # Germany bonds ('SGX_JB1', 'JP10YT', data_tools.QuantpediaBondYield, -1, False), # Japan bonds ('BIL', 'IR3TIB01USM156N', data_tools.QuantpediaMonthlyData, 0, False) # US cash ] # country pe data self.country_pe_data: Symbol = self.AddData(data_tools.CountryPE, 'CountryData').Symbol self.period: int = 12 * 21 self.leverage: int = 5 self.quantile: int = 4 self.SetWarmUp(self.period) self.data: Dict[str, RollingWindow] = {} for symbol, yield_symbol, yield_access in list(map(lambda x: (x[0], x[1], x[2]), self.assets)): # investable asset if yield_access == data_tools.QuantpediaBondYield: data = self.AddData(data_tools.QuantpediaFutures, symbol, Resolution.Daily) else: data = self.AddEquity(symbol, Resolution.Daily) # yield if yield_access != None: self.AddData(yield_access, yield_symbol, Resolution.Daily) self.data[symbol] = RollingWindow[float](self.period) data.SetFeeModel(CustomFeeModel()) data.SetLeverage(self.leverage) self.Settings.MinimumOrderMarginPortfolioPercentage = 0. self.recent_month: int = -1 def OnData(self, data: Slice) -> None: custom_data_last_update_date: Dict[Symbol, datetime.date] = data_tools.LastDateHandler.get_last_update_date() if self.IsWarmingUp: return if all([self.Securities[symbol].GetLastData() for symbol in list(map(lambda x: x[0], self.assets))]) and \ any([self.Time.date() >= custom_data_last_update_date[yield_symbol] for yield_symbol in list(map(lambda x: x[1], self.assets)) if yield_symbol in custom_data_last_update_date]): self.Liquidate() return # store investable asset price data for symbol, yield_symbol in list(map(lambda x: (x[0], x[1]), self.assets)): symbol_obj: Symbol = self.Symbol(symbol) if symbol_obj in data and data[symbol_obj]: self.data[symbol].Add(data[symbol_obj].Value) if self.Time.month == self.recent_month: return self.recent_month = self.Time.month performance_1M: Dict[str, float] = {} performance_12M: Dict[str, float] = {} valuation: Dict[str, float] = {} # performance and valuation calculation if self.Securities[self.country_pe_data].GetLastData() and self.Time.date() < custom_data_last_update_date[self.country_pe_data]: for symbol, yield_symbol, yield_access, bond_adjustment, reverse_flag in self.assets: if self.data[symbol].IsReady: closes: List[float] = list(self.data[symbol]) performance_1M[symbol] = closes[0] / closes[21] - 1 performance_12M[symbol] = closes[0] / closes[len(closes) - 1] - 1 if yield_access == None: country_pb_data = self.Securities['CountryData'].GetLastData() if country_pb_data: pe: float = country_pb_data[yield_symbol] yield_value: float = pe else: yield_value = self.Securities[self.Symbol(yield_symbol)].Price # reverse if needed, EP->PE if reverse_flag: yield_value = 1/yield_value if yield_value != 0: valuation[symbol] = yield_value + bond_adjustment long: List[str] = [] short: List[str] = [] if len(valuation) != 0: # sort assets by metrics sorted_by_p1: List[Tuple[str, float]] = sorted(performance_1M.items(), key = lambda x: x[1]) sorted_by_p12: List[Tuple[str, float]] = sorted(performance_12M.items(), key = lambda x: x[1]) sorted_by_value: List[Tuple[str, float]] = sorted(valuation.items(), key = lambda x: x[1]) # rank assets score = {} for i, (symbol, _) in enumerate(sorted_by_p1): score[symbol] = i * 0.25 for i, (symbol, _) in enumerate(sorted_by_p12): score[symbol] += i * 0.25 for i, (symbol, _) in enumerate(sorted_by_value): score[symbol] += i * 0.5 # sort by rank sorted_by_rank: List[Tuple[str, float]] = sorted(score, key = lambda x: score[x], reverse = True) quartile: int = int(len(sorted_by_rank) / self.quantile) long = sorted_by_rank[:quartile] short = sorted_by_rank[-quartile:] invested: List[str] = [x.Key.Value for x in self.Portfolio if x.Value.Invested and x.Key.Value not in long + short] for ticker in invested: self.Liquidate(ticker) # trade execution for i, portfolio in enumerate([long, short]): for symbol in portfolio: if symbol in data and data[symbol]: self.SetHoldings(symbol, ((-1) ** i) / len(portfolio)) # Custom fee model. class CustomFeeModel(FeeModel): def GetOrderFee(self, parameters: OrderFeeParameters) -> OrderFee: fee: float = parameters.Security.Price * parameters.Order.AbsoluteQuantity * 0.00005 return OrderFee(CashAmount(fee, "USD"))