def plots():
FTSE100_AdjClose = pickle.load(open('pickle_jar/DJIA_AdjClose.pkl'))
results = pickle.load(open('pickle_jar/results_DJIA_coint.pkl','rb'))
results = pd.DataFrame(results, columns = ['SeriesX_name','SeriesY_name','Adf_pvalue', 'reg_scaling', 'reg_intercept'])
results = results.sort('Adf_pvalue', ascending=True)
print results.head(15)
for row in range(0,5):
rowinfo = results.iloc[row].tolist()
seriesX = FTSE100_AdjClose[rowinfo[0]]
seriesX = seriesX[seriesX.notnull()]
seriesY = FTSE100_AdjClose[rowinfo[1]]
seriesY = seriesY[seriesY.notnull()]
df_coint_check = pd.merge(pd.DataFrame(seriesX),
pd.DataFrame(seriesY),
how='inner',
left_index=True,
right_index=True)
df_coint_check.columns = ['x','y']
pmd = pairs_md(df_coint_check, xInd='x', yInd = 'y')
pmd.fitOLS()
pmd.generateTradeSigs(windowLength=20, entryScale=2, exitScale=0)
pmd.plot_spreadAndSignals()
plt.title( "Row {3}: Spread between {0} and {1}, scaling {2}".format(rowinfo[0],rowinfo[1],rowinfo[3], row))
plt.show()
def search_for_CI(series_names):
FTSE100_AdjClose = pickle.load(open('pickle_jar/DJIA_AdjClose.pkl'))
results = []
combs = [cm for cm in combinations(series_names,2)]
i = 0
for cmb in combs:
print 'Starting row', i
i +=1
seriesX_name = cmb[0]
seriesY_name = cmb[1]
try:
seriesX = FTSE100_AdjClose[seriesX_name]
seriesX = seriesX[seriesX.notnull()]
seriesY = FTSE100_AdjClose[seriesY_name]
seriesY = seriesY[seriesY.notnull()]
df_coint_check = pd.merge(pd.DataFrame(seriesX),
pd.DataFrame(seriesY),
how='inner',
left_index=True,
right_index=True)
df_coint_check.columns = ['x','y']
pmd = pairs_md(df_coint_check, xInd='x', yInd = 'y')
pmd.fitOLS()
reg_params = pmd.results.params.tolist()
adf_pvalue = pmd.adfResids()[1]
results.append([seriesX_name, seriesY_name, adf_pvalue] + reg_params)
except ValueError as e:
print "Problem with pairs X = {0}, Y = {1}".format(seriesX_name,seriesY_name)
print "Error is ", e.what()
pickle.dump(results , open('pickle_jar/results_DJIA_coint.pkl','wb'))
def PairTradeSP500(stdWindowLength, entryScale, exitScale):
#Read in data and setup object
dbpath = "/home/phcostello/Documents/Data/FinanceData.sqlite"
the_data_reader = dr.DBReader(dbpath)
SP500 = the_data_reader.readSeries("SP500")
BA = the_data_reader.readSeries("BA")
dim = 'Adj_Close' #Choose dim to analyse
SP500AdCl = SP500[dim]
BAAdCl = BA[dim]
# print SP500AdCl.head()
# print BAAdCl.head()
dataObj = pd.merge(pd.DataFrame(BAAdCl), pd.DataFrame(SP500AdCl), how='inner',left_index = True, right_index = True)
dataObj.columns = ['y','x']
#Construct data object for pairs trading strat
pmd = md.pairs_md(dataOb=dataObj,xInd='x',yInd='y')
maxdate = datetime.date(2013,1,1)
pmd.core_data = pmd.core_data[:maxdate]
pmd.fitOLS()
pmd.generateTradeSigs(stdWindowLength, entryScale, exitScale)
#Setup portfolio
spreadTrade = td.TradeEquity("spread", notional=0, price_series_label="spread")
port = pf.Portfolio("portfolio", cashAmt=100)
port.add_trade(spreadTrade)
#No more trade types
port.fixed_toggle()
#Setup Strategy
pairsStrat = tsc.Reversion_EntryExitTechnical(market_data=pmd, portfolio=port, initial_time_index=1)
#return pairsStrat
tic = time.clock()
pairsStrat.run_strategy()
toc = time.clock()
print "strategy took {} seconds to run".format(toc - tic)
outfile = open("pickled_pairs.pkl", 'wb')
pairs_strategy_run = pickle.dump(pairsStrat,outfile)
outfile.close()
def test_pairs_md(self):
#prepare data
import DataHandler.DBReader as dbr
dbpath = "/home/phcostello/Documents/Data/FinanceData.sqlite"
dbreader = dbr.DBReader(dbpath)
SP500 = dbreader.readSeries("SP500")
BA = dbreader.readSeries("BA")
dim = 'Adj_Close'
SP500AdCl = SP500[dim]
BAAdCl = BA[dim]
dataObj = pd.merge(pd.DataFrame(BAAdCl), pd.DataFrame(SP500AdCl), how='inner',left_index = True, right_index = True)
dataObj.columns = ['y','x']
pmd = pairs_md(dataOb=dataObj,xInd='x',yInd='y')
pmd.fitOLS()
resid = pmd.results.resid
resid.plot()
rllstd = pd.rolling_std(resid,50)
rllstd.plot()
pmd.generateTradeSigs(50, entryScale=1, exitScale=0)
self.assertTrue( (pmd['entryUpper'][50:] == rllstd[50:]).all())
def grid_tester(run_strat, do_plots):
##################Initialise##########################
#Read in data and setup object
# dbpath = "/home/phcostello/Documents/Data/FinanceData.sqlite"
# dbreader = dbr.DBReader(dbpath)
#SP500 = dbreader.readSeries("SP500")
#BA = dbreader.readSeries("BA")
#dim = 'Adj_Close'
#SP500AdCl = SP500[dim]
#BAAdCl = BA[dim]
#dataObj = pd.merge(pd.DataFrame(BAAdCl), pd.DataFrame(SP500AdCl), how='inner',left_index = True, right_index = True)
#dataObj.columns = ['y','x']
#Read data for all pairs
#Note have checked top 10 in and constructed data in
#CheckingSeriesCoint.py
dataObj = pickle.load(open('pickle_jar/DJIA_AdjClose.pkl'))
cointAnalysis = pickle.load(open('pickle_jar/results_DJIA_coint.pkl','rb'))
cointAnalysis = pd.DataFrame(cointAnalysis, columns = ['SeriesX_name','SeriesY_name','Adf_pvalue', 'reg_scaling', 'reg_intercept'])
cointAnalysis = cointAnalysis.sort('Adf_pvalue', ascending=True)
#cointAnalysis.info()
seriesNames = cointAnalysis[['SeriesX_name','SeriesY_name']]
#dataObj.info()
top10 = seriesNames.iloc[0:5]
top10['rank']= range(1, len(top10)+1)
top10.head()
top10 =top10.values.tolist()
#Filter to what we want
#setup logging
logfile = 'logs/GridTester_ma.txt'
logging.basicConfig(filename= logfile, filemode='w', level = logging.ERROR)
##################Make par Table##########################
#Parameter table rows are
#[ pairs_labes, strategy_params, train_ranges, test_ranges ]
#series
series_names = top10
#strat pars
entry_scale = np.arange(1,3.5,0.5)
exit_scale = np.arange(0.0,3.0,0.5)
strat_parameters = [ [sw, lw] for sw in entry_scale for lw in exit_scale if sw >= lw]
#Train - test date pars
dmin = datetime.datetime(2006,1,1)
num_periods = 15
period_delta = datetime.timedelta(182)
#Create range start/end offsets from min data
train_test_ranges = [ [dmin + i*period_delta, #train start
dmin + (i+1)* period_delta, #train end
dmin + (i+1)* period_delta, #test start
dmin + (i+2)* period_delta] #test end
for i in range(0,num_periods)]
#Make table of parameters + dates
parameter_table = [ series + pars + date_ranges for series in series_names for pars in strat_parameters for date_ranges in train_test_ranges]
parameter_table = [ parameter_table[i] + [i] for i in range(0,len(parameter_table))]
colnames =['runtype',\
'series_X',\
'series_Y',\
'rank',\
'entry_scale',\
'exit_scale',\
'train_start',\
'train_end',\
'test_start',\
'test_end',\
'par_rownumber']
##################Make Res Table##########################
#We just define
result_cols = ['returns',\
'volatility',\
'sharpe',\
'sortino']
#Use x will be analyser object which has result functions
#result cols should have same order as function below
result_func = lambda x : [x.get_result().iat[-1,0]/x.get_result().iat[0,0] - 1 ,
x.get_volatility(),
x.sharpe_ratio(),
x.sortino_ratio()]
result_table_cols = colnames + result_cols
##################Test strat on pars in Table##########################
if run_strat:
#open file for output
outfile = open('Results/dummy.csv','wb')
#run simulations
tic = time.clock()
runStrat(parameter_table,dataObj, outfile, result_table_cols, result_func)
toc = time.clock()
print "All strategies took {} seconds to run".format(toc - tic)
outfile.close()
#################$ Plot for one set of parameters ########################
#Setup Data
if do_plots:
this_pars = parameter_table[496]
#Setup Data
#Setup initial portfolio
trade_equity_spread = TradeEquity('spread',
notional=0,
price_series_label='spread')
portfolio_0 = Portfolio("portfolio", cashAmt=100)
portfolio_0.add_trade(trade_equity_spread)
#No more trade types
portfolio_0.fixed_toggle()
print this_pars
strategy, train_data, test_data = setup_strategy(this_pars,dataObj)
data = test_data
strategy.run_strategy(data,portfolio_0)
analyser = ResultsAnalyser(strategy.result,referenceIndex=None)
fig =plt.figure()
ax1 = fig.add_subplot(2,1,1)
ax2 = fig.add_subplot(2,1,2)
portfolioVal = analyser.get_result()
print portfolioVal['Value']
start = 0
end = len(portfolioVal)
portfolioVal['Value'].iloc[start:end].plot(ax=ax1)
# trades_at = portfolioVal[portfolioVal['Signal_bs'].isin(['buy','sell'])]
#
# #Putting trade bars on
# for dt in trades_at.index:
# plt.axvline( dt, ymin =0, ymax =1 )
plt.subplot(2,1,1)
pairs_md(data.loc[portfolioVal.index].iloc[start:end],0,0).plot_spreadAndSignals(ax2)
#plt.title(seriesName)
plt.show()