def __init__(self, lim, bmark_ticker, constituents, n_rand_port, filter_co):
self.lim = lim
self.start_date = self.calculate_start_date(lim)
self.bmark_ticker = bmark_ticker
self.constituents = constituents
self.n_rand_port = n_rand_port
self.stocklist = StocklistParser('data/yahoo_full_stocklist.csv', filter_co)
self.dir_output = 'output'
if not os.path.exists(self.dir_output):
os.makedirs(self.dir_output)
self.f_portfolios = 'portfolios.csv'
class RandomPortfolioReturnStrategy(object):
def __init__(self, lim, bmark_ticker, constituents, n_rand_port, filter_co):
self.lim = lim
self.start_date = self.calculate_start_date(lim)
self.bmark_ticker = bmark_ticker
self.constituents = constituents
self.n_rand_port = n_rand_port
self.stocklist = StocklistParser('data/yahoo_full_stocklist.csv', filter_co)
self.dir_output = 'output'
if not os.path.exists(self.dir_output):
os.makedirs(self.dir_output)
self.f_portfolios = 'portfolios.csv'
def calculate_start_date(self, data_point_days):
weekends = (data_point_days / 7) * 2
actual_days_delta = data_point_days + (weekends * 2) #Add a buffer of +1 day for every weekend to account for any non-trading days
start_date = datetime.date.today() - datetime.timedelta(actual_days_delta)
return start_date
def get_rand_stock(self):
tickers = self.stocklist.get_symbols()
rand_index = randint(0, (len(tickers)-1))
return (tickers[rand_index], rand_index)
def fetch_prices(self, ticker):
today = datetime.date.today()
priceApi = 'http://ichart.finance.yahoo.com/table.csv?s='
params = '&a=' + str(self.start_date.month-1) + '&b=' + str(self.start_date.day) + '&c=' + str(self.start_date.year) + '&d=' + str(today.month-1) + '&e=' + str(today.day) + '&f=' + str(today.year) + '&g=d&ignore=.csv'
try:
data = urllib2.urlopen(priceApi + ticker + params).read()
data = data.replace('\r', '').split('\n')[1:-1][::-1]
except:
return self.fetch_prices(self.get_rand_stock()[0])
'''Check data has enough historical data points and first historic price is not 0 [division by zero error]'''
if ((len(data)-1) < self.lim) or (float(data[0].split(',')[6]) == 0):
while ((len(data)-1) < self.lim) or (float(data[0].split(',')[6]) == 0):
next_rand_ticker = self.get_rand_stock()[0]
try:
data = urllib2.urlopen(priceApi + next_rand_ticker + params).read()
data = data.replace('\r', '').split('\n')[1:-1][::-1]
except:
return self.fetch_prices(self.get_rand_stock()[0])
prices = []
for d in data:
prices.append(float(d.split(',')[6]))
prices_pct = self.convert_pct(prices)
return prices_pct
def convert_pct(self, prices):
prices_pct = []
i = 0
while len(prices_pct) < (len(prices) - 1):
chng_pc = ((prices[i + 1] / prices[0]) - 1) * 100
prices_pct.append(chng_pc)
i += 1
return prices_pct
def calculate_portfolio_performance(self, portfolio_prices):
i = 0
portfolio_return = []
while i < self.lim:
j = 0
sigma = 0
while j < len(portfolio_prices):
sigma += float(portfolio_prices[j][i])
j += 1
avg_pc = sigma / j
portfolio_return.append(avg_pc)
i += 1
return portfolio_return
def plot_results(self, benchmark, rand_portfolios):
print "\n>> Plotting results..."
now = datetime.date.today()
window_title = ('Performance of ' + str(self.n_rand_port) + ' Random Portfolios vs ' + self.bmark_ticker + ' Benchmark')
xlab = 'N Days (ending on ' + str(now.day) + '/' + str(now.month) + '/' + str(now.year) + ')'
ylab = '% Change'
line_plot = mpl_graph_line(window_title, xlab, ylab, True)
plot_data = [] #Tuple: ([prices], linewidth)
for portfolio in rand_portfolios:
plot_data.append((portfolio, 1))
plot_data.append((benchmark, 2))
legend = []
for i in xrange(len(rand_portfolios)):
legend.append('Rand Portfolio (' + str(i+1) + ')')
legend.append('Benchmark (' + self.bmark_ticker + ')')
line_plot.plot(plot_data, legend)
def run_simulation(self):
print "\n>> Running simulation calculations..."
t_start = time.time()
benchmark = self.fetch_prices(self.bmark_ticker)[:self.lim]
rand_portfolios = []
done_total = self.n_rand_port * self.constituents
portfolio_stock_names = np.empty([(self.constituents+1), self.n_rand_port], dtype='S24')
for n in xrange(self.n_rand_port):
portfolio = []
for i in xrange(self.constituents):
rand_stock = self.get_rand_stock()
portfolio.append(rand_stock)
portfolio_prices = []
portfolio_stock_names[0][n] = 'Rand Portfolio (' + str(n+1) + ')'
for i, stock in enumerate(portfolio):
price = self.fetch_prices(stock[0])
portfolio_prices.append(price)
portfolio_stock_names[(i+1)][n] = stock[0]
done_curr = (n * self.constituents) + (i + 1)
completed = (float(done_curr)/float(done_total)) * 100
sys.stdout.write("\r%.2f" % completed + "% completed...")
sys.stdout.flush()
portfolio_performance = self.calculate_portfolio_performance(portfolio_prices)
rand_portfolios.append(portfolio_performance)
print "\n\n>> Finished simulating random portfolios!"
np.savetxt((self.dir_output + '/' + self.f_portfolios), portfolio_stock_names, delimiter=',', fmt='%s')
print ">> Random portfolios constituents saved in:\n", self.f_portfolios
t_end = time.time()
t_delta = t_end - t_start
self.plot_results(benchmark, rand_portfolios)
return t_delta
