def get_closes(tickers, start_date=None, end_date=None, freq=None):
closes = pd.DataFrame(columns=tickers, index=web.YahooDailyReader(tickers[0], start=start_date, end=end_date
, interval=freq).read().index)
for ticker in tickers:
df = web.YahooDailyReader(symbols=ticker, start=start_date, end=end_date, interval=freq).read()
closes[ticker] = df['Adj Close']
closes.index_name = 'Date'
closes = closes.sort_index()
return closes
def get_open(tickers, start_date=None, end_date=None, freq=None):
open_ = pd.DataFrame(columns=tickers, index=web.YahooDailyReader(tickers[0], start=start_date, end=end_date
, interval=freq).read().index)
for ticker in tickers:
df = web.YahooDailyReader(symbols=ticker, start=start_date, end=end_date, interval=freq).read()
open_[ticker] = df['Open']
open_.index_name = 'Date'
open_ = open_.sort_index()
return open_
def generate_raw_prices(self, interval):
'''Function for generating raw price time series from a given specified interval.'''
history = data.YahooDailyReader(self.ticker,
self.start,
self.end,
interval=interval)
return history.read()
def getStockData(stockCode, start='', end=''):
# 수집 기간
if start == '':
start = dt.datetime(2007, 1, 1)
else:
start = dt.datetime.strptime(start, '%Y-%m-%d')
if end == '':
end = dt.date.today()
else:
end = dt.datetime.strptime(end, '%Y-%m-%d')
stock = pd.DataFrame()
for i in range(0, 10):
try:
stock = web.YahooDailyReader(stockCode,
start,
end,
adjust_price=True).read()
except:
print("%s not collected (%d)" % (stockCode, i + 1))
if not stock.empty:
break
if stock.empty:
print("%s not collected" % stockCode)
# 수정주가 비율은 이미 적용되었으므로 제거한다
stock = stock.drop('Adj_Ratio', 1)
# Volume이 0 인 경우가 있으므로, 이를 제거한다
stock = stock.drop(stock[stock.Volume < 10].index)
return stock
def get_historical_closes(ticker, start_date, end_date=None):
closes = web.YahooDailyReader(ticker, start_date, end_date).read()
#closes = web.YahooDailyReader(symbols=ticker, start=start_date, end=end_date).read()
#closes.set_axis(closes.loc['date',:,ticker[0]].values, axis=1, inplace=True)
#closes = closes.loc['adjclose'].sort_index().dropna()
#closes = pd.DataFrame(np.array(closes.as_matrix(), dtype=np.float64), columns=ticker, index=closes.index)
#closes.index.name = 'Date'
return pd.DataFrame(closes.loc[:, 'Adj Close'])
def get_closes(tickers, start_date=None, end_date=None, freq=None):
# Fecha inicio por defecto (start_date='2010-01-01') y fecha fin por defecto (end_date=today)
# Frecuencia de muestreo por defecto (freq='d')
# Importamos paquetes necesarios
import pandas as pd
pd.core.common.is_list_like = pd.api.types.is_list_like
import pandas_datareader.data as web
# Creamos DataFrame vacío de precios, con el índice de las fechas
closes = pd.DataFrame(columns=tickers,
index=web.YahooDailyReader(
symbols=tickers[0],
start=start_date,
end=end_date,
interval=freq).read().index)
# Agregamos cada uno de los precios con YahooDailyReader
for ticker in tickers:
df = web.YahooDailyReader(symbols=ticker,
start=start_date,
end=end_date,
interval=freq).read()
closes[ticker] = df['Adj Close']
closes.index_name = 'Date'
closes = closes.sort_index()
return closes
def get_prices(tickers, ventana):
import pandas as pd
import pandas_datareader.data as web
import pickle
import datetime
today = datetime.date.today()
#ventana = 365*1
inicio = today - datetime.timedelta(ventana)
try: # Load file from pickle
infile = open(indx + '.pkl', 'rb')
historicos = pickle.load(infile)
infile.close()
tickers = list(historicos.keys())
closes = pd.DataFrame(columns=tickers)
for ticker in tickers:
closes[ticker] = historicos[ticker]['Adj Close']
print('Load succesful')
except: # download from yahoo
while (today.weekday() > 4):
today = today - datetime.timedelta(1)
closes = pd.DataFrame(
columns=tickers
) #, index=web.YahooDailyReader(symbols=tickers[0], start=inicio, end=today, interval='d').read().index)
historicos = {}
i = 0
for ticker in tickers:
try:
df = web.YahooDailyReader(symbols=ticker,
start=inicio,
end=today,
interval='d').read()
historicos[ticker] = df
closes[ticker] = df['Adj Close']
i += 1
print(ticker, ' downloaded')
except:
pass
print("Downloaded {} of {} succesfully".format(i, len(tickers)))
closes.index_name = 'Date'
closes = closes.sort_index()
closes = closes.dropna(axis=1)
outfile = open(indx + '.pkl', 'wb')
pickle.dump(historicos, outfile)
outfile.close()
tickers = list(historicos.keys())
return historicos, closes, tickers
def getStockDataYahoo(stockCode, start='', end=''):
# 수집 기간
if start == '':
start = dt.datetime(2007, 1, 1)
else:
start = dt.datetime.strptime(start, '%Y-%m-%d')
if end == '':
end = dt.date.today()
else:
end = dt.datetime.strptime(end, '%Y-%m-%d')
stock = pd.DataFrame()
# Yahoo site로 부터 주가 데이터를 수집한다. 가끔 안들어올 때가 있어서 10번 시도한다.
for i in range(0, 10):
try:
# 수정 주가로 환산하여 읽어온다
stock = web.YahooDailyReader(stockCode,
start,
end,
adjust_price=True).read()
except:
print("%s not collected (%d)" % (stockCode, i + 1))
if not stock.empty:
break
if stock.empty:
print("%s not collected" % stockCode)
# 수정주가 비율은 이미 적용되었으므로 제거한다
stock = stock.drop('Adj_Ratio', 1)
# Volume이 0 인 경우가 있으므로, 이를 제거한다
stock = stock.drop(stock[stock.Volume < 10].index)
# 데이터에 NA 값이 있으면 제거한다
stock = stock.dropna()
# 수집한 데이터를 파일에 저장한다.
stock.to_csv('StockData/' + stockCode[0:6] + '.csv',
date_format='%Y-%m-%d')
print("%s 데이터를 수집하였습니다. (rows = %d)" % (stockCode, len(stock)))
return stock
def get_historical_closes(ticker, start_date, end_date=None):
# Packages
import pandas_datareader.data as web
import pandas as pd
import numpy as np
#closes = web.DataReader(name=ticker, data_source='yahoo', start=start_date, end=end_date).sort_index('major_axis')
closes = web.YahooDailyReader(symbols=ticker,
start=start_date,
end=end_date).read()
closes.set_axis(closes.loc['date', :, ticker[0]].values,
axis=1,
inplace=True)
closes = closes.loc['adjclose'].sort_index().dropna()
closes = pd.DataFrame(np.array(closes.as_matrix(), dtype=np.float64),
columns=ticker,
index=closes.index)
closes.index.name = 'Date'
#return closes.loc['Adj Close']
return closes
def getStockDataYahooTwoYr(stockCode, start='', end=''):
# 수집 기간
if start == '':
today = dt.datetime.today()
time_delta = dt.timedelta(days=730)
firstday = (today - time_delta).replace(day=1)
start = dt.datetime.strftime(firstday, "%Y-%m-%d")
else:
start = dt.datetime.strptime(start, '%Y-%m-%d')
if end == '':
end = dt.date.today()
else:
end = dt.datetime.strptime(end, '%Y-%m-%d')
stock = pd.DataFrame()
for i in range(0, 2):
try:
stock = web.YahooDailyReader(stockCode,
start,
end,
adjust_price=True).read()
except:
print("%s not collected (%d)" % (stockCode, i + 1))
if not stock.empty:
# 수정주가 비율은 이미 적용되었으므로 제거한다
stock = stock.drop('Adj_Ratio', 1)
# Volume이 0 인 경우가 있으므로, 이를 제거한다
# stock = stock.drop(stock[stock.Volume < 10].index)
stock = stock.drop(stock[stock.Volume < 10000].index)
# 수집한 데이터를 파일에 저장한다.
break
if stock.empty:
print("%s not collected" % stockCode)
return stock
def f_stats_mad(datos):
"""
Parameters
----------
datos : pandas.DataFrame : dataframe con transacciones ejecutadas después de tiempos y pips
Returns
-------
datos : pandas.DataFrame : dataframe con rendimientos logarítmicos. Tomando en cuenta que se inicializa con una cuenta de $5,000
Debugging
---------
datos = 'f_leer_archivo("archivo_tradeview_1.csv")
"""
profit_d = f_profit_diario(datos)
# Sharpe ratio
rend_log = np.log(profit_d.profit_acm_d[1:].values /
profit_d.profit_acm_d[:-1].values)
rf = 0.08 / 300
mar = 0.3 / 300
# Sortino compra
# Numerador
s_buy = (f_profit_diario(datos[datos['type'] == 'buy']))
rend_log_b = np.log(s_buy.profit_acm_d[1:].values /
s_buy.profit_acm_d[:-1].values)
# Denominador
tdd_sb = rend_log_b - mar
tdd_sb[tdd_sb > 0] = 0
# Final
sortino_b = (rend_log_b.mean() - mar) / (((tdd_sb * 2).mean()) * 0.5)
# Sortino venta
# Numerador
s_sell = (f_profit_diario(datos[datos['type'] == 'sell']))
rend_log_s = np.log(s_sell.profit_acm_d[1:].values /
s_sell.profit_acm_d[:-1].values)
# Denominador
tdd_ss = rend_log_s - mar
tdd_ss[tdd_ss > 0] = 0
# Final
sortino_s = (rend_log_s.mean() - mar) / ((tdd_ss * 2).mean()) * 0.5
# Drawdown
where_row = profit_d.loc[profit_d['profit_acm_d'] ==
profit_d.profit_acm_d.min()]
where_position = where_row.index.tolist()
prev_where = profit_d.loc[0:where_position[0]]
where_max_prev = profit_d.loc[profit_d['profit_acm_d'] ==
prev_where.profit_acm_d.max()]
where_min_prev = profit_d.loc[profit_d['profit_acm_d'] ==
prev_where.profit_acm_d.min()]
max_dd = where_max_prev.iloc[0]['profit_acm_d']
min_dd = where_min_prev.iloc[0]['profit_acm_d']
dd = max_dd - min_dd
fecha_i_dd = where_max_prev.iloc[0]['timestamp']
fecha_f_dd = where_min_prev.iloc[0]['timestamp']
drawdown = "{}, {}, ${:.2f}".format(fecha_i_dd, fecha_f_dd, dd)
# Drawup
where_row_up = profit_d.loc[profit_d['profit_acm_d'] ==
profit_d.profit_acm_d.max()]
where_position_up = where_row_up.index.tolist()
foll_where = profit_d.loc[0:where_position_up[0]]
where_max_foll = profit_d.loc[profit_d['profit_acm_d'] ==
foll_where.profit_acm_d.max()]
where_min_foll = profit_d.loc[profit_d['profit_acm_d'] ==
foll_where.profit_acm_d.min()]
max_du = where_max_foll.iloc[0]['profit_acm_d']
min_du = where_min_foll.iloc[0]['profit_acm_d']
du = max_du - min_du
fecha_f_du = where_max_foll.iloc[0]['timestamp']
fecha_i_du = where_min_foll.iloc[0]['timestamp']
drawup = "{}, {}, ${:.2f}".format(fecha_i_du, fecha_f_du, du)
# Information Ratio
benchmark_original = pd.DataFrame(
web.YahooDailyReader('^GSPC',
profit_d['timestamp'].min(),
profit_d['timestamp'].max(),
interval='d').read()['Adj Close'])
# Ajustar fechas como columna en el dataframe
benchmark = benchmark_original.reset_index()
# Agregar columna de los rendimientos logarítmicos de los precios de Cierre Ajustado del SP500
benchmark['rend_log'] = pd.DataFrame(
np.log(benchmark['Adj Close'][1:].values /
benchmark['Adj Close'][:-1].values))
# Juntar el dataframe de profits con el benchmark en uno solo
bench_merge = profit_d.merge(benchmark,
left_on='timestamp',
right_on='Date')
# Recorrer los valores de los rendimientos logarítmicos una posición abajo y llenar con 0
bench_merge['rend_log'] = bench_merge['rend_log'].shift(1, fill_value=0)
# Cálculo de rendimientos logarítmicos de los profit_acm_d
rend_log_profit = pd.DataFrame(
np.log(bench_merge.profit_acm_d[1:].values /
bench_merge.profit_acm_d[:-1].values))
# rend_log_profit = rend_log_profit.shift(1, fill_value = 0)
# Agregar la columna de rendimientos de profit_acm_d
bench_merge.insert(3, column='rend_log_profit', value=rend_log_profit)
# Recorrer una fila hacia abajo los rendimientos en el dataframe
bench_merge['rend_log_profit'] = bench_merge['rend_log_profit'].shift(
1, fill_value=0)
# Numerador Information Ratio
# Promedio de rendimientos del profit_acm_d
profit_prom = bench_merge['rend_log_profit'].mean()
# Promedio de rendimientos del SP500
bench_prom = bench_merge['rend_log'].mean()
# Numerador del information ratio = diferencia de los dos anteriores
num_ir = profit_prom - bench_prom
# Denominador Information Ratio
# Diferencia por rows de los rendimientos del profit_acm_d y del SP500
dif_denom = bench_merge['rend_log_profit'] - bench_merge['rend_log']
# Desviación estándar de la diferencia
denom_ir = dif_denom.std()
# Cálculo del ratio
info_ratio = num_ir / denom_ir
# Métricas
metrica = pd.DataFrame({
'métricas': [
'sharpe', 'sortino_b', 'sortino_s', 'drawdown_cap_b',
'drawdown_cap_s', 'information_r'
]
})
valor = pd.DataFrame({
'valor': [((rend_log.mean() - rf) / rend_log.std()), (sortino_b),
(sortino_s), (drawdown), (drawup), (info_ratio)]
})
df_mad1 = pd.merge(metrica, valor, left_index=True, right_index=True)
descripcion = pd.DataFrame({
'descripción': [
'Sharpe Ratio', 'Sortino Ratio para Posiciones de Compra',
'Sortino Ratio para Posiciones de Venta', 'DrawDown de Capital',
'DrawUp de Capital', 'Information Ratio'
]
})
df_MAD = pd.merge(df_mad1, descripcion, left_index=True, right_index=True)
# convert_dict = {'valor': float}
# df_MAD = df_MAD.astype(convert_dict)
return df_MAD
aapl_qua = web.DataReader(ticker, source, start, end)
aapl_qua
# Ejemplo iex
ticker = 'AAPL'
source = 'iex'
start = '2015-01-01'
end = '2017-12-31'
aapl_iex = web.DataReader(ticker, source, start, end)
aapl_iex
# YahooDailyReader
ticker = 'AAPL'
start = '2015-01-01'
end = '2017-12-31'
aapl_yah = web.YahooDailyReader(ticker, start, end, intervals='d'.read())
aapl_yah
#%%
# Función para descargar precios de cierre ajustados de varios activos a la vez:
def get_closes(tickers, start_date=None, end_date=None, freq=None):
# Fecha inicio por defecto (start_date='2010-01-01') y fecha fin por defecto (end_date=today)
# Frecuencia de muestreo por defecto (freq='d')
# Importamos paquetes necesarios
import pandas as pd
pd.core.common.is_list_like = pd.api.types.is_list_like
import pandas_datareader.data as web
# Creamos DataFrame vacío de precios, con el índice de las fechas
closes = pd.DataFrame(columns=tickers,
index=web.YahooDailyReader(
def prices_from_index(indx, ventana):
import pandas as pd
import requests
import pandas_datareader.data as web
import pickle
import datetime
today = datetime.date.today()
# Definir el indice con el que se quiere trabajar
# para mexico se usa el indice
# para extrankeras se usa un ETF que represente al sector/indice deseado
# Se descargaran los componentes mas importantes
# indx = 'SPY'
# GET TICKERS
if indx == 'MXX':
url = 'https://finance.yahoo.com/quote/%5E' + indx + '/components'
html = requests.get(url).content
df_list = pd.read_html(html)
df = df_list[-1]
tickers = df.Symbol.tolist()
tickers = [
'LIVEPOLC-1.MX' if x == 'LIVEPOLC1.MX' else x for x in tickers
]
tickers = ['PE&OLES.MX' if x == 'PEOLES.MX' else x for x in tickers]
tickers = ['KOFUBL.MX' if x == 'KOFL.MX' else x for x in tickers]
tickers.insert(0, 'WALMEX.MX')
tickers.insert(0, 'BACHOCO.MX')
tickers.insert(0, 'AEROMEX.MX')
tickers.insert(0, 'VOLARA.MX')
tickers.insert(0, 'AZTECACPO.MX')
tickers.insert(0, 'LALAB.MX')
tickers.insert(0, 'CHDRAUIB.MX')
tickers.insert(0, '^' + indx)
else:
url = 'https://www.marketwatch.com/investing/fund/' + indx + '/holdings'
html = requests.get(url).content
df_list = pd.read_html(html)
df = df_list[-1]
tickers = df.Symbol.dropna().tolist()
tickers.insert(0, indx)
# GET PRICES
inicio = today - datetime.timedelta(ventana)
try: # Load file from pickle
infile = open(indx + '.pkl', 'rb')
historicos = pickle.load(infile)
infile.close()
tickers = list(historicos.keys())
closes = pd.DataFrame(columns=tickers)
for ticker in tickers:
closes[ticker] = historicos[ticker]['Adj Close']
print('Load succesful')
except: # download from yahoo
while (today.weekday() > 4):
today = today - datetime.timedelta(1)
closes = pd.DataFrame(
columns=tickers
) #, index=web.YahooDailyReader(symbols=tickers[0], start=inicio, end=today, interval='d').read().index)
historicos = {}
i = 0
for ticker in tickers:
try:
df = web.YahooDailyReader(symbols=ticker,
start=inicio,
end=today,
get_actions=True,
interval='d').read()
historicos[ticker] = df
closes[ticker] = df['Adj Close']
i += 1
print(ticker, ' downloaded')
except:
print(ticker, ' not available')
pass
print("Downloaded {} of {} succesfully".format(i, len(tickers)))
closes.index_name = 'Date'
closes = closes.sort_index()
closes = closes.dropna(axis=1)
outfile = open(indx + '.pkl', 'wb')
pickle.dump(historicos, outfile)
outfile.close()
tickers = list(historicos.keys())
return tickers, historicos, closes
def update_prices(indx, ventana):
import pickle
import datetime
import pandas as pd
import pandas_datareader.data as web
today = datetime.date.today()
while (today.weekday() > 4):
today = today - datetime.timedelta(1)
# Cargar datos
infile = open(indx + '.pkl', 'rb')
historicos = pickle.load(infile)
tickers = list(historicos.keys())
closes = pd.DataFrame(columns=tickers)
#closes = pd.DataFrame(columns = tickers)
#indx = 'MXX'
first_date = min([historicos[ticker].index[0] for ticker in tickers])
inicio = today - datetime.timedelta(ventana)
if inicio < first_date:
closes = pd.DataFrame(columns=tickers)
historicos = {}
i = 0
for ticker in tickers:
try:
df = web.YahooDailyReader(symbols=ticker,
start=inicio,
end=today,
get_actions=True,
interval='d').read()
historicos[ticker] = df
closes[ticker] = df['Adj Close']
i += 1
print(ticker, ' downloaded')
except:
print(ticker, ' not available')
pass
print("Downloaded {} of {} succesfully".format(i, len(tickers)))
closes.index_name = 'Date'
closes = closes.sort_index()
closes = closes.dropna(axis=1)
outfile = open(indx + '.pkl', 'wb')
pickle.dump(historicos, outfile)
outfile.close()
tickers = list(historicos.keys())
last_date = max([historicos[ticker].index[-1] for ticker in tickers])
if today > last_date:
print('Updating data')
closes = pd.DataFrame(columns=tickers)
i = 0
for ticker in tickers:
try:
df = web.YahooDailyReader(symbols=ticker,
start=last_date +
datetime.timedelta(1),
end=today,
get_actions=True,
interval='d').read()
historicos[ticker] = historicos[ticker].append(df)
closes[ticker] = historicos[ticker]['Adj Close']
i += 1
print(ticker, ' downloaded')
except:
print(ticker, ' not available')
pass
# Guardar de nuevo
closes.index_name = 'Date'
closes = closes.sort_index()
closes = closes.dropna(axis=1)
outfile = open(indx + '.pkl', 'wb')
pickle.dump(historicos, outfile)
outfile.close()
tickers = list(historicos.keys())
else:
for ticker in tickers:
closes[ticker] = historicos[ticker]['Adj Close']
return tickers, historicos, closes