from datetime import timedelta, datetime import statsmodels.api as sm import numpy as np import pandas as pd from Selection.FundamentalUniverseSelectionModel import FundamentalUniverseSelectionModel from sklearn.decomposition import PCA class SMAPairsTrading(QCAlgorithm): def Initialize(self): self.SetStartDate(2015, 1 , 1 ) self.SetEndDate(2020, 9 , 1 ) self.SetCash(100000) self.UniverseSettings.Resolution = Resolution.Daily self.AddUniverse(self.Universe.Index.QC500) self.UniverseSettings.DataNormalizationMode = DataNormalizationMode.Raw self.AddAlpha(PairsTradingAlphaModel()) self.SetPortfolioConstruction(EqualWeightingPortfolioConstructionModel()) self.SetExecution(ImmediateExecutionModel()) self.SetRiskManagement(MaximumDrawdownPercentPerSecurity(0.03)) self.SetBenchmark("SPY") self.SetSecurityInitializer(self.CustomSecurityInitializer) self.buy = pd.DataFrame() self.sell = pd.DataFrame() self.liquidate = pd.DataFrame() def OnEndOfDay(self, symbol): self.Log("Taking a position of " + str(self.Portfolio[symbol].Quantity) + " units of symbol " + str(symbol)) def CustomSecurityInitializer(self, security): security.SetLeverage(1) class PairsTradingAlphaModel(AlphaModel): def __init__(self): self.pair = [] self.period = timedelta(days=1) def Update(self, algorithm, data): List=[x.Symbol for x in self.pair] history = algorithm.History(List, 61 ).close.unstack(level=0) self.buy,self.sell,self.liquidate = self.GetIndexes( history) Appd = [] for i in self.buy: Appd.append(Insight.Price(i,self.period, InsightDirection.Up,None,None,None))#,None, None, None,0.02)) for i in self.sell: Appd.append(Insight.Price(i,self.period, InsightDirection.Down,None,None,None)) for i in self.liquidate: Appd.append(Insight.Price(i,self.period, InsightDirection.Flat,None,None,None)) return Insight.Group([ x for x in Appd]) def GetIndexes(self, history): # Sample data for PCA sample = history.dropna(axis=1).pct_change().dropna() sample_mean = sample.mean() sample_std = sample.std() sample = ((sample-sample_mean)/(sample_std)) #Normalizing # Fit the PCA model for sample data model = PCA().fit(sample) #Distributing eigenportfolios EigenPortfolio = pd.DataFrame(model.components_) EigenPortfolio.columns = sample.columns # EigenPortfolio = EigenPortfolio/sample_std EigenPortfolio = ( EigenPortfolio.T / EigenPortfolio.sum(axis=1) ) # Get the first n_components factors factors = np.dot(sample, EigenPortfolio)[:,:1] # we want to replicate the market # Add 1's to fit the linear regression (intercept) factors = sm.add_constant(factors) # Train Ordinary Least Squares linear model for each stock OLSmodels = {ticker: sm.OLS(sample[ticker], factors).fit() for ticker in sample.columns} # Get the residuals from the linear regression after PCA for each stock resids = pd.DataFrame({ticker: model.resid for ticker, model in OLSmodels.items()}) # Get the OU parameters shifted_residuals = resids.cumsum().iloc[1:,:] resids = resids.cumsum().iloc[:-1,:] resids.index = shifted_residuals.index OLSmodels2 = {ticker: sm.OLS(resids[ticker],sm.add_constant(shifted_residuals[ticker])).fit() for ticker in resids.columns} # Get the new residuals resids2 = pd.DataFrame({ticker: model.resid for ticker, model in OLSmodels2.items()}) # Get the mean reversion parameters a = pd.DataFrame({ticker : model.params[0] for ticker , model in OLSmodels2.items()},index=["a"]) b = pd.DataFrame({ticker: model.params[1] for ticker , model in OLSmodels2.items()},index=["a"]) e = (resids2.std())/(252**(-1/2)) k = -np.log(b) * 252 #Get the z-score var = (e**2 /(2 * k) )*(1 - np.exp(-2 * k * 252)) num = -a * np.sqrt(1 - b**2) den = ( 1-b ) * np.sqrt( var ) m = ( a / ( 1 - b ) ) zscores=(num / den ).iloc[0,:]# zscores of the most recent day # Get the stocks far from mean (for mean reversion) selected_buy = zscores[zscores < -1.5].dropna().sort_values()[:1] selected_sell = zscores[zscores > 1.5].dropna().sort_values()[-1:] selected_liquidate = zscores[abs(zscores) < 0.50 ] # Return each selected stock weights_buy = selected_buy.index weights_sell = selected_sell.index weights_liquidate = selected_liquidate.index return weights_buy, weights_sell, weights_liquidate def OnSecuritiesChanged(self, algorithm, changes): self.pair = [x for x in changes.AddedSecurities]

Hello everybody !

I finally could rewrite my original messy code but each time I  execute the algorithm it seems like orders are not fulfilled during the day but rather at the opening/closing of the market.It causes a real problem for the risk management.I tried to switch the universe from daily to hourly but it sounds like the GetIndexes function uses the 61 hourly datas rather than the 61 daily data.

How can keep the daily datas with in the same time executing the orders during the full trading hours and only selecting one stock (as the algorithm is originally doing) per day and not per hour ?

 

Thank You !!

Author