def test_returns_hist(symbol, start_date: datetime, end_date: datetime,
ind_return, ax):
df_main = pd.DataFrame(index=pd.date_range(start_date, end_date))
df_temp = quotes_manager.get_daily_quotes(start_date, end_date, symbol)
df_main = df_main.join(df_temp[quotes_manager.COL_ADJ_CLOSE]).rename(
columns={quotes_manager.COL_ADJ_CLOSE: symbol})
df_main.dropna(inplace=True)
ind_return.calculate_ind_into_df(df_main, symbol)
ind_column_name = ind_return.get_column_name(symbol)
mean = df_main[ind_column_name].mean()
print(ind_column_name + ' mean=' + str(mean))
std = df_main[ind_column_name].std()
print(ind_column_name + ' std=' + str(std))
kurtosis = df_main[ind_column_name].kurtosis()
# to check if the tail of the histogran is aligned with a normal distribution or not
print(ind_column_name + ' kurtosis=' + str(kurtosis))
ax = df_main[ind_column_name].hist(ax=ax, bins=20)
ax.axvline(mean, color='w', linestyle='dashed', linewidth=2)
ax.axvline(std, color='r', linestyle='dashed', linewidth=2)
ax.axvline(-std, color='r', linestyle='dashed', linewidth=2)
def test_returns_hist_same_ax(symbol, start_date: datetime, end_date: datetime,ind_return):
df_main = pd.DataFrame(index=pd.date_range(start_date,end_date))
df_temp = quotes_manager.get_daily_quotes(start_date,end_date,symbol)
df_main = df_main.join(df_temp[quotes_manager.COL_ADJ_CLOSE]).rename(columns={quotes_manager.COL_ADJ_CLOSE:symbol})
df_main.dropna(inplace=True)
ind_return.calculate_ind_into_df(df_main,symbol)
ind_column_name = ind_return.get_column_name(symbol)
df_main[ind_column_name].hist(bins=30,alpha=0.5, label=symbol)
def compare_returns(symbols, start_date: datetime, end_date: datetime):
df_main = pd.DataFrame(index=pd.date_range(start_date, end_date))
for symbol in symbols:
df_temp = quotes_manager.get_daily_quotes(start_date, end_date, symbol)
# df_temp = df_temp[[quotes_manager.COL_DATE,quotes_manager.COL_ADJ_CLOSE]].rename(columns={quotes_manager.COL_ADJ_CLOSE:symbol})
df_main = df_main.join(df_temp[quotes_manager.COL_ADJ_CLOSE]).rename(
columns={quotes_manager.COL_ADJ_CLOSE: symbol})
# df_main = df_main.join(df_temp)
df_main = df_main.dropna()
df_returns = df_main / df_main.iloc[0]
df_returns.plot()
plt.show()
def calculate_portifolio(start_portfolio_value,start_date,end_date, symbols, allocations=[],use_quotes_cache=True):
df_quotes = pd.DataFrame(index=pd.date_range(start_date, end_date))
for symbol in symbols:
df_temp = quotes_manager.get_daily_quotes(start_date, end_date, symbol, use_cache=use_quotes_cache)
df_quotes = df_quotes.join(df_temp[quotes_manager.COL_ADJ_CLOSE]).rename(
columns={quotes_manager.COL_ADJ_CLOSE: symbol})
df_quotes.dropna(inplace=True)
df_normalized_returns = pd.DataFrame(index=pd.date_range(start_date, end_date))
norm_ret = NormalizedDailyReturn()
for symbol in symbols:
df_normalized_returns = df_normalized_returns.join(norm_ret.calculate_ind_series(df_quotes[symbol]))
df_normalized_returns.dropna(inplace=True)
if not allocations or len(allocations) == 0:
# guess allocation, just a simple distribution
guess_alloc = np.full(len(symbols), 100 / len(symbols))
# limiting the values to be from 0 to 100
bounds = [(0, 100) for x in range(len(symbols))]
# only satisfied when sum of elements is 1.
cons = ({'type': 'eq', 'fun': lambda x: 100 - np.sum(x)})
result = spo.minimize(find_sharpe_allocation, guess_alloc, args=(df_normalized_returns), bounds=bounds, constraints=cons, method='SLSQP',
options={'disp': True})
allocations = result.x / 100
print('Calculated Sharpe = ' + str(result.fun))
print('Calculated Allocations = ')
np.savetxt(sys.stdout, allocations, '%5.2f')
df_allocations = df_normalized_returns * allocations
df_position_value = df_allocations * start_portfolio_value
df_portfolio_value = df_position_value.sum(axis=1)
day_ret = DailyReturn()
day_ret_series = day_ret.calculate_ind_series(df_portfolio_value)
avg_daily_return = day_ret_series.mean()
# Risk, Vol
std_daily_return = day_ret_series.std()
end_value = df_portfolio_value.iloc[-1]
cum_return = (end_value / start_portfolio_value) - 1
rf = 0.
sharpe = np.sqrt(252) * ((avg_daily_return - rf) / std_daily_return)
print('Statistics for ' + str(symbols))
print('avg_daily_return = ' + str(avg_daily_return))
print('std_daily_return = ' + str(std_daily_return))
print('end_value = ' + str(end_value))
print('cum_return = ' + str(cum_return))
print('sharpe = ' + str(sharpe))
return df_quotes, df_normalized_returns, df_allocations, df_position_value, df_portfolio_value
def bollinger_bands_test(symbol, start_date: datetime, end_date: datetime,
window: int, n_std: int):
df = quotes_manager.get_daily_quotes(start_date, end_date, symbol)
df = df.rename(columns={quotes_manager.COL_ADJ_CLOSE: symbol})[[symbol]]
boll_band_sma = BollingerBandsSMA(window, n_std)
boll_band_sma_df = boll_band_sma.calculate_ind_into_df(df, symbol)
boll_band_sma_df.plot(figsize=(12, 6))
plt.show()
boll_band_ema = BollingerBandsEMA(window, n_std)
boll_band_ema_df = boll_band_ema.calculate_ind_into_df(df, symbol)
boll_band_ema_df.plot(figsize=(12, 6))
plt.show()
def test_returns(symbols, start_date: datetime, end_date: datetime, ind_return,
ax):
df_main = pd.DataFrame(index=pd.date_range(start_date, end_date))
for symbol in symbols:
df_temp = quotes_manager.get_daily_quotes(start_date, end_date, symbol)
df_main = df_main.join(df_temp[quotes_manager.COL_ADJ_CLOSE]).rename(
columns={quotes_manager.COL_ADJ_CLOSE: symbol})
df_main.dropna(inplace=True)
for symbol in symbols:
ind_return.calculate_ind_into_df(df_main, symbol)
df_main[[ind_return.get_column_name(symbol)
for symbol in symbols]].plot(ax=ax)
def test_exponential_moving__avg(symbol, start_date: datetime,
end_date: datetime):
df = quotes_manager.get_daily_quotes(start_date, end_date, symbol)
df = df.rename(columns={quotes_manager.COL_ADJ_CLOSE: symbol})
ema = ExponentialMovingAvg(20)
ema.calculate_ind_into_df(df, symbol)
sma = SimpleMovingAvg(20)
sma.calculate_ind_into_df(df, symbol)
ax = df[[symbol,
ema.get_ind_column_name(),
sma.get_ind_column_name()]].plot()
plt.show()
def calculate_portifolio(start_portfolio_value,
start_date,
end_date,
symbols,
allocations,
use_quotes_cache=True):
df_quotes = pd.DataFrame(index=pd.date_range(start_date, end_date))
for symbol in symbols:
df_temp = quotes_manager.get_daily_quotes(start_date,
end_date,
symbol,
use_cache=use_quotes_cache)
df_quotes = df_quotes.join(
df_temp[quotes_manager.COL_ADJ_CLOSE]).rename(
columns={quotes_manager.COL_ADJ_CLOSE: symbol})
df_quotes.dropna(inplace=True)
df_normalized_returns = pd.DataFrame(
index=pd.date_range(start_date, end_date))
norm_ret = NormalizedDailyReturn()
for symbol in symbols:
df_normalized_returns = df_normalized_returns.join(
norm_ret.calculate_ind_series(df_quotes[symbol]))
df_normalized_returns.dropna(inplace=True)
df_allocations = df_normalized_returns * allocations
df_position_value = df_allocations * start_portfolio_value
df_portfolio_value = df_position_value.sum(axis=1)
day_ret = DailyReturn()
day_ret_series = day_ret.calculate_ind_series(df_portfolio_value)
avg_daily_return = day_ret_series.mean()
# Risk, Vol
std_daily_return = day_ret_series.std()
end_value = df_portfolio_value.iloc[-1]
cum_return = (end_value / start_portfolio_value) - 1
rf = 0.
sharpe = np.sqrt(252) * ((avg_daily_return - rf) / std_daily_return)
print('Statistics for ' + str(symbols))
print('avg_daily_return = ' + str(avg_daily_return))
print('std_daily_return = ' + str(std_daily_return))
print('end_value = ' + str(end_value))
print('cum_return = ' + str(cum_return))
print('sharpe = ' + str(sharpe))
return df_quotes, df_normalized_returns, df_allocations, df_position_value, df_portfolio_value
def test_returns_scatter(x_symbol, y_symbol, start_date: datetime,
end_date: datetime, ind_return, ax):
symbols = [x_symbol, y_symbol]
df_main = pd.DataFrame(index=pd.date_range(start_date, end_date))
for symbol in symbols:
df_temp = quotes_manager.get_daily_quotes(start_date, end_date, symbol)
df_main = df_main.join(df_temp[quotes_manager.COL_ADJ_CLOSE]).rename(
columns={quotes_manager.COL_ADJ_CLOSE: symbol})
df_main.dropna(inplace=True)
for symbol in symbols:
ind_return.calculate_ind_into_df(df_main, symbol)
beta_Y, alpha_Y = np.polyfit(df_main[x_symbol], df_main[y_symbol], 1)
print('Beta ' + y_symbol + ' = ' + str(beta_Y))
print('Alpha ' + y_symbol + ' = ' + str(alpha_Y))
ax.plot(df_main[x_symbol],
beta_Y * df_main[x_symbol] + alpha_Y,
'-',
color='r')
df_main.plot(kind='scatter', x=x_symbol, y=y_symbol, ax=ax)
import domain
from domain.order import order
from domain.order_type import order_type
from domain.portfolio import portfolio
from manager import quotes_manager
from datetime import datetime
import pandas as pd
df_quotes: pd.DataFrame
start_date = datetime(2018, 7, 1)
end_date = datetime(2018, 7, 15)
df_quotes = quotes_manager.get_daily_quotes(start_date, end_date, 'AAPL')
print(df_quotes)
# df_quotes['close'].pct_change()
df_quotes['ClosePctChange'] = df_quotes['4. close'].pct_change()
rand_portfolio = portfolio('random')
for index, row in df_quotes.iterrows():
if not pd.isna(row['ClosePctChange']):
if row['ClosePctChange'] > 0:
type = order_type.SELL
else:
type = order_type.BUY
rand_portfolio.add_order(
order(type, 'AAPL', 1000, row['4. close'],
datetime.strptime(row.name, '%Y-%m-%d')))
df_positions = rand_portfolio.calculate_positions_df(start_date, end_date)
