def generate_curve_pca_formatted_output(**kwargs):
if "report_date" in kwargs.keys():
report_date = kwargs["report_date"]
else:
report_date = exp.doubledate_shift_bus_days()
output_dir = ts.create_strategy_output_dir(strategy_class="curve_pca", report_date=report_date)
ticker_head_list = ["CL", "B"]
selected_column_list = [
"ticker1",
"ticker2",
"monthSpread",
"tr_dte_front",
"residuals",
"price",
"yield",
"z",
"z2",
"factor_load1",
"factor_load2",
]
writer = pd.ExcelWriter(output_dir + "/" + futil.xls_file_names["curve_pca"] + ".xlsx", engine="xlsxwriter")
for ticker_head in ticker_head_list:
curve_pca_output = cpc.get_curve_pca_report(ticker_head=ticker_head, date_to=report_date)
if curve_pca_output["success"]:
all_spreads = curve_pca_output["pca_results"]
filter_out = ff.get_curve_pca_filters(data_frame_input=all_spreads, filter_list=["long1", "short1"])
good_spreads = filter_out["selected_frame"]
all_spreads = all_spreads[selected_column_list]
good_spreads = good_spreads[selected_column_list]
all_spreads.to_excel(writer, sheet_name=ticker_head + "-all")
good_spreads.to_excel(writer, sheet_name=ticker_head + "-good")
worksheet_good = writer.sheets[ticker_head + "-good"]
worksheet_all = writer.sheets[ticker_head + "-all"]
worksheet_good.freeze_panes(1, 0)
worksheet_all.freeze_panes(1, 0)
worksheet_good.autofilter(0, 0, len(good_spreads.index), len(selected_column_list))
worksheet_all.autofilter(0, 0, len(all_spreads.index), len(selected_column_list))
def generate_curve_pca_formatted_output(**kwargs):
if 'report_date' in kwargs.keys():
report_date = kwargs['report_date']
else:
report_date = exp.doubledate_shift_bus_days()
output_dir = ts.create_strategy_output_dir(strategy_class='curve_pca',
report_date=report_date)
ticker_head_list = ['CL', 'B']
selected_column_list = [
'ticker1', 'ticker2', 'monthSpread', 'tr_dte_front', 'residuals',
'price', 'yield', 'z', 'z2', 'factor_load1', 'factor_load2'
]
writer = pd.ExcelWriter(output_dir + '/' +
futil.xls_file_names['curve_pca'] + '.xlsx',
engine='xlsxwriter')
for ticker_head in ticker_head_list:
curve_pca_output = cpc.get_curve_pca_report(ticker_head=ticker_head,
date_to=report_date)
if curve_pca_output['success']:
all_spreads = curve_pca_output['pca_results']
filter_out = ff.get_curve_pca_filters(
data_frame_input=all_spreads, filter_list=['long1', 'short1'])
good_spreads = filter_out['selected_frame']
all_spreads = all_spreads[selected_column_list]
good_spreads = good_spreads[selected_column_list]
all_spreads.to_excel(writer, sheet_name=ticker_head + '-all')
good_spreads.to_excel(writer, sheet_name=ticker_head + '-good')
worksheet_good = writer.sheets[ticker_head + '-good']
worksheet_all = writer.sheets[ticker_head + '-all']
worksheet_good.freeze_panes(1, 0)
worksheet_all.freeze_panes(1, 0)
worksheet_good.autofilter(0, 0, len(good_spreads.index),
len(selected_column_list))
worksheet_all.autofilter(0, 0, len(all_spreads.index),
len(selected_column_list))
writer.save()
def backtest_curve_pca(**kwargs):
ticker_head = kwargs['ticker_head']
date_to = kwargs['date_to']
years_back = kwargs['years_back']
indicator = kwargs['indicator']
backtest_output_dir = dn.get_directory_name(ext='backtest_results')
file_name = ticker_head + '_' + str(date_to) + '_' + str(years_back) + '_' + indicator
if os.path.isfile(backtest_output_dir + '/curve_pca/' + file_name + '.pkl'):
backtest_results = pd.read_pickle(backtest_output_dir + '/curve_pca/' + file_name + '.pkl')
return backtest_results
dates_output = bu.get_backtesting_dates(date_to=date_to, years_back=years_back)
date_list = sorted(dates_output['double_dates'], reverse=False)
if 'use_existing_filesQ' in kwargs.keys():
use_existing_filesQ = kwargs['use_existing_filesQ']
else:
use_existing_filesQ = True
report_results_list = []
success_indx = []
contract_multiplier = cmi.contract_multiplier[ticker_head]
for date_to in date_list:
report_out = cpc.get_curve_pca_report(ticker_head=ticker_head, date_to=date_to, use_existing_filesQ=use_existing_filesQ)
success_indx.append(report_out['success'])
if report_out['success']:
pca_results = report_out['pca_results']
pca_results['report_date'] = date_to
report_results_list.append(pca_results)
good_dates = [date_list[i] for i in range(len(date_list)) if success_indx[i]]
total_pnl_list = []
z_score_list = []
residual_list = []
num_contract_list = []
short_side_weight_list = []
for i in range(len(good_dates)):
daily_report = report_results_list[i]
#daily_report = daily_report[daily_report['monthSpread']==1]
#daily_report = daily_report[3:]
daily_report_filtered = daily_report[(daily_report['tr_dte_front']>80) & (daily_report['monthSpread']==1)]
#median_factor_load2 = daily_report['factor_load2'].median()
#if median_factor_load2>0:
# daily_report_filtered = daily_report[daily_report['factor_load2']>=0]
#else:
# daily_report_filtered = daily_report[daily_report['factor_load2']<=0]
daily_report_filtered.sort(indicator,ascending=True,inplace=True)
num_contract_4side = round(len(daily_report_filtered.index)/4)
long_side = daily_report_filtered.iloc[:num_contract_4side]
short_side = daily_report_filtered.iloc[-num_contract_4side:]
short_side_weight = long_side['factor_load1'].sum()/short_side['factor_load1'].sum()
z_score_list.append(np.nanmean(short_side[indicator])-np.nanmean(long_side[indicator]))
if any(np.isnan(long_side['change5'])) or any(np.isnan(short_side['change5'])):
total_pnl_list.append(np.NAN)
else:
total_pnl_list.append(np.nanmean(long_side['change5'])-short_side_weight*np.nanmean(short_side['change5']))
residual_list.append(np.nanmean(short_side['residuals'])-np.nanmean(long_side['residuals']))
num_contract_list.append(num_contract_4side)
short_side_weight_list.append(short_side_weight)
backtest_results = {'pnl_frame': pd.DataFrame.from_items([('settle_date',good_dates),
('num_contracts',num_contract_list),
('z',z_score_list),
('residual',residual_list),
('short_side_weight',short_side_weight_list),
('pnl',[x*contract_multiplier for x in total_pnl_list])]),
'report_results_list': report_results_list}
with open(backtest_output_dir + '/curve_pca/' + file_name + '.pkl', 'wb') as handle:
pickle.dump(backtest_results, handle)
return backtest_results
