def test_code_in_sample(learner, commission=0., impact=0.000): #commission=9.95, impact=0.005):
#symbols = ['SPY','JPM']
#start_date = '2008-01-01'
#end_date = '2009-12-31'
#date_range = pd.date_range(start_date, end_date)
print '\n\n\n in sample Benchmark:\n'
date = dt.datetime(2008, 1, 2)
Benchmark_orders = pd.DataFrame([[1000]],index = [date], columns=['JPM'])
#date = dt.datetime(2009, 12, 31)
#Benchmark_orders_last = pd.DataFrame([[-1000]],index = [date], columns=['JPM'])
#Benchmark_orders = Benchmark_orders.append(Benchmark_orders_last)
Benchmark_portvals = ms.compute_portvals_pd(Benchmark_orders, start_val = 100000, sd=dt.datetime(2008, 1, 1), ed=dt.datetime(2009,12,31), commission=commission, impact=impact)
ms.print_stats(Benchmark_portvals)
Benchmark_portvals_n = Benchmark_portvals/Benchmark_portvals.values[0]
print '\n\n\n in sample Stratedy Learner:\n'
orders = learner.testPolicy(symbol = "JPM", sd=dt.datetime(2008, 1, 1), ed=dt.datetime(2009,12,31), sv = 100000)
portvals = ms.compute_portvals_pd(orders, start_val = 100000, sd=dt.datetime(2008, 1, 1), ed=dt.datetime(2009,12,31), commission=commission, impact=impact)
ms.print_stats(portvals)
portvals_n = portvals/portvals.values[0]
'''
trade_buy = orders[orders[orders.columns[0]]>0]
trade_sell = orders[orders[orders.columns[0]]<0]
#print trade_buy
#print trade_sell
date_buy = trade_buy.index.tolist()
date_sell = trade_sell.index.tolist()
for xc in date_buy:
plt.axvline(x=xc,c='b')
for xc in date_sell:
plt.axvline(x=xc,c='k')
'''
plt.plot(portvals_n['Val'],'r', label='Stratedy Learner')
plt.plot(Benchmark_portvals_n['Val'],'g', label='Benchmark')
#plt.axis([0, 300, -256, 100])
plt.xticks(rotation=30)
plt.xlabel('Date')
plt.ylabel('Normalized Portfolio Value')
plt.title('in sample Stratedy Learner')
#plt.annotate('Blue lines - LONG entry points', xy=(dt.date(2008,1,1), 1.2), xytext=(dt.date(2007,12,02), 1.47),
# #arrowprops=dict(arrowstyle="->", connectionstyle="arc3")
# )
#plt.annotate('Black lines - SHORT entry points', xy=(dt.date(2008,1,1), 1.15), xytext=(dt.date(2007,12,02), 1.4),
##arrowprops=dict(arrowstyle="->", connectionstyle="arc3")
#)
#plt.legend(['Theoretically Optimal Strategy','Benchmark'])
plt.legend(loc='upper left')
plt.tight_layout()
plt.savefig('Stratedy Learner in sample.png')
plt.show()
plt.close()
def test_code_out_sample():
#symbols = ['SPY','JPM']
#start_date = '2008-01-01'
#end_date = '2009-12-31'
#date_range = pd.date_range(start_date, end_date)
print '\n\n\nBenchmark:\n'
date = dt.datetime(2010, 1, 4)
Benchmark_orders = pd.DataFrame([['JPM', 1000]],index = [date.date()], columns=['Symbol','Shares'])
Benchmark_portvals = compute_portvals_pd(Benchmark_orders, start_val = 100000, sd=dt.datetime(2010, 1, 1), ed=dt.datetime(2011,12,31), commission=9.95, impact=0.005)
print_stats(Benchmark_portvals)
Benchmark_portvals_n = Benchmark_portvals/Benchmark_portvals.values[0]
print '\n\n\nManual Stratedy:\n'
orders = testPolicy(symbol = "JPM", sd=dt.datetime(2010, 1, 1), ed=dt.datetime(2011,12,31), sv = 100000)
portvals = compute_portvals_pd(orders, start_val = 100000, sd=dt.datetime(2010, 1, 1), ed=dt.datetime(2011,12,31), commission=9.95, impact=0.005)
print_stats(portvals)
portvals_n = portvals/portvals.values[0]
trade_buy = orders[orders['Shares']>0]
trade_sell = orders[orders['Shares']<0]
print trade_buy
print trade_sell
date_buy = trade_buy.index.tolist()
date_sell = trade_sell.index.tolist()
for xc in date_buy:
plt.axvline(x=xc,c='b')
for xc in date_sell:
plt.axvline(x=xc,c='k')
plt.plot(portvals_n['Val'],'r', label='Manual Rule-based Stratedy')
plt.plot(Benchmark_portvals_n['Val'],'g', label='Benchmark')
#plt.axis([0, 300, -256, 100])
plt.xticks(rotation=30)
plt.xlabel('Date')
plt.ylabel('Normalized Portfolio Value')
plt.title('Manual Rule-based Stratedy')
plt.annotate('Blue lines - LONG entry points', xy=(dt.date(2010,1,1), 1.2), xytext=(dt.date(2010,1,01), 1.2),
#arrowprops=dict(arrowstyle="->", connectionstyle="arc3")
)
plt.annotate('Black lines - SHORT entry points', xy=(dt.date(2010,1,1), 1.15), xytext=(dt.date(2010,1,01), 1.16),
#arrowprops=dict(arrowstyle="->", connectionstyle="arc3")
)
#plt.legend(['Theoretically Optimal Strategy','Benchmark'])
plt.legend(loc='upper left')
plt.tight_layout()
plt.savefig('out sample Manual Rule-based Stratedy.png')
#plt.show()
plt.close()
def test_code():
#symbols = ['SPY','JPM']
#start_date = '2008-01-01'
#end_date = '2009-12-31'
#date_range = pd.date_range(start_date, end_date)
print '\n\n\nBenchmark:\n'
date = dt.datetime(2008, 1, 2)
Benchmark_orders = pd.DataFrame([['JPM', 1000]],
index=[date.date()],
columns=['Symbol', 'Shares'])
Benchmark_portvals = compute_portvals_pd(Benchmark_orders,
start_val=100000,
sd=dt.datetime(2008, 1, 1),
ed=dt.datetime(2009, 12, 31),
commission=0,
impact=0)
print_stats(Benchmark_portvals)
Benchmark_portvals_n = Benchmark_portvals / Benchmark_portvals.values[0]
print '\n\n\nTheoretically Optimal Strategy:\n'
orders = testPolicy(symbol='JPM',
sd=dt.datetime(2008, 1, 1),
ed=dt.datetime(2009, 12, 31),
sv=100000)
portvals = compute_portvals_pd(orders,
start_val=100000,
sd=dt.datetime(2008, 1, 1),
ed=dt.datetime(2009, 12, 31),
commission=0,
impact=0)
print_stats(portvals)
portvals_n = portvals / portvals.values[0]
plt.plot(portvals_n['Val'], 'r', label='Theoretically Optimal Strategy')
plt.plot(Benchmark_portvals_n['Val'], 'g', label='Benchmark')
#plt.axis([0, 300, -256, 100])
plt.xticks(rotation=30)
plt.xlabel('Date')
plt.ylabel('Normalized Portfolio Value')
plt.title('Theoretically Optimal Strategy')
#plt.legend(['Theoretically Optimal Strategy','Benchmark'])
plt.legend()
plt.tight_layout()
plt.savefig('Theoretically_Optimal_Strategy.png')
#plt.show()
plt.close()