def preprocess(source,
index,
start_date,
end_date,
force_rerun=True,
force_rerun_ta=True):
# load symbol name
with open(make_input_file(index)) as csvfile:
symbols = [
row[0]
for ind, row in enumerate(csv.reader(csvfile.read().splitlines()))
if ind > 0
]
# download Yahoo finance trading data for all symbols in the index
if force_rerun:
# download trading data
download_batch(source, symbols, index, start_date, end_date, 'd',
make_output_file(index, 'daily'))
daily_data = pd.read_csv(make_output_file(index, 'daily'))
if source == 'yahoo': # only yahoo provide weekly data
download_batch(source, symbols, index, start_date, end_date, 'w',
make_output_file(index, 'weekly'))
weekly_data = pd.read_csv(make_output_file(index, 'weekly'))
if force_rerun_ta:
# calculate TA indicators for each symbol
symbols = daily_data.Symbol.unique()
directory = os.path.join(settings['folder']['trading'], index)
if not os.path.exists(directory):
os.mkdir(directory)
for symbol in symbols:
fname = os.path.join(directory, symbol + '_ta.csv')
if source in ('g_1year', 'quandl'): # avoid overwrite
if force_rerun == False and os.path.exists(fname):
continue
df = daily_data.loc[daily_data['Symbol'] == symbol]
df = df.iloc[::-1].reset_index().drop('index', axis=1)
df = ma(df, 20)
df = ma(df, 50)
df = b_bands(df, 20)
df = rsi(df, 14)
df = sto_k(df, 14)
df = sto_d(df, 14)
df = macd(df, 12, 26)
df.to_csv(fname, index=False)
if source == 'yahoo': #only yahoo provides weekly data
wfname = os.path.join(directory, symbol + '_weekly_ta.csv')
wdf = weekly_data.loc[weekly_data['Symbol'] == symbol]
wdf = wdf.iloc[::-1].reset_index().drop('index', axis=1)
wdf = ma(wdf, 20)
wdf = ma(wdf, 50)
wdf = rsi(wdf, 14)
wdf.to_csv(wfname, index=False)
def trigger_index_stats(index, a, b, success_window):
with open(make_input_file(index)) as csvfile:
symbols = [
row[0]
for ind, row in enumerate(csv.reader(csvfile.read().splitlines()))
if ind > 0
]
stats_table = [[
'Trigger_name', 'Index', 'Avg_gain', 'Success_rate', 'Frequency'
]]
weighted_gain = weighted_rate = cnt = 0
for symbol in symbols:
trigger = triggers.trigger_sporadic(symbol, index, success_window)
stats = success_rate(trigger, success_window)
weighted_gain += stats.avg_gain * stats.frequency
weighted_rate += stats.success_rate * stats.frequency
cnt += stats.frequency
stats_table.append([
'trigger_sporadic', index, weighted_gain / cnt, weighted_rate / cnt,
cnt
])
weighted_gain = weighted_rate = cnt = 0
for symbol in symbols:
trigger = triggers.trigger_macd_trend(symbol, index, a, b,
success_window)
stats = success_rate(trigger, success_window)
weighted_gain += stats.avg_gain * stats.frequency
weighted_rate += stats.success_rate * stats.frequency
cnt += stats.frequency
stats_table.append([
'trigger_macd_trend', index, weighted_gain / cnt, weighted_rate / cnt,
cnt
])
weighted_gain = weighted_rate = cnt = 0
for symbol in symbols:
trigger = triggers.trigger_rsi_trend(symbol, index, a, b,
success_window)
stats = success_rate(trigger, success_window)
weighted_gain += stats.avg_gain * stats.frequency
weighted_rate += stats.success_rate * stats.frequency
cnt += stats.frequency
stats_table.append([
'trigger_rsi_trend', index, weighted_gain / cnt, weighted_rate / cnt,
cnt
])
weighted_gain = weighted_rate = cnt = 0
for symbol in symbols:
trigger = triggers.trigger_rsi_low(symbol, index, success_window)
stats = success_rate(trigger, success_window)
weighted_gain += stats.avg_gain * stats.frequency
weighted_rate += stats.success_rate * stats.frequency
cnt += stats.frequency
stats_table.append([
'trigger_rsi_low', index, weighted_gain / cnt, weighted_rate / cnt, cnt
])
weighted_gain = weighted_rate = cnt = 0
for symbol in symbols:
trigger = triggers.trigger_rsi_high(symbol, index, success_window)
stats = success_rate(trigger, success_window)
weighted_gain += stats.avg_gain * stats.frequency
weighted_rate += stats.success_rate * stats.frequency
cnt += stats.frequency
stats_table.append([
'trigger_rsi_high', index, weighted_gain / cnt, weighted_rate / cnt,
cnt
])
weighted_gain = weighted_rate = cnt = 0
for symbol in symbols:
trigger = triggers.trigger_bband_lowerflat(symbol, index, a,
success_window)
stats = success_rate(trigger, success_window)
weighted_gain += stats.avg_gain * stats.frequency
weighted_rate += stats.success_rate * stats.frequency
cnt += stats.frequency
stats_table.append([
'trigger_bband_lowerflat', index, weighted_gain / cnt,
weighted_rate / cnt, cnt
])
weighted_gain = weighted_rate = cnt = 0
for symbol in symbols:
trigger = triggers.trigger_bband_upbreak(symbol, index, a,
success_window)
stats = success_rate(trigger, success_window)
weighted_gain += stats.avg_gain * stats.frequency
weighted_rate += stats.success_rate * stats.frequency
cnt += stats.frequency
stats_table.append([
'trigger_bband_upbreak', index, weighted_gain / cnt,
weighted_rate / cnt, cnt
])
weighted_gain = weighted_rate = cnt = 0
for symbol in symbols:
trigger = triggers.trigger_bband_downbreak(symbol, index, a,
success_window)
stats = success_rate(trigger, success_window)
weighted_gain += stats.avg_gain * stats.frequency
weighted_rate += stats.success_rate * stats.frequency
cnt += stats.frequency
stats_table.append([
'trigger_bband_downbreak', index, weighted_gain / cnt,
weighted_rate / cnt, cnt
])
weighted_gain = weighted_rate = cnt = 0
for symbol in symbols:
trigger = triggers.trigger_above_20wma(
symbol, index, 10, 2
) # weekly data has n=10, successwindow=2, different from daily data
stats = success_rate(trigger, success_window)
weighted_gain += stats.avg_gain * stats.frequency
weighted_rate += stats.success_rate * stats.frequency
cnt += stats.frequency
stats_table.append([
'trigger_above_20wma', index, weighted_gain / cnt, weighted_rate / cnt,
cnt
])
with open(
os.path.join(settings['folder']['results'],
index + '_results.csv'), 'wt') as csvfile:
csv.writer(csvfile, delimiter=",").writerows(stats_table)
return stats_table
import datetime
import os
import pandas as pd
import sys
index = 'dow30'
source = 'quandl'
today = datetime.date.today()
start_date = str(today - datetime.timedelta(365))
end_date = str(today)
trigger_start_date = str(today - datetime.timedelta(14))
a, b, success_window = 5, 2, 0 # success_window=0 for forecast realtime
# a and b is for particular trigger, e.g. GS, macd corss up, 5 consecutive neg values followed by 2 positive values
# load symbols
with open(make_input_file(index)) as csvfile:
symbols = [
row[0]
for ind, row in enumerate(csv.reader(csvfile.read().splitlines()))
if ind > 0
]
# get recent data and ta indicators
preprocess(source,
index,
start_date,
end_date,
force_rerun=True,
force_rerun_ta=True)
# apply trigger based on ta indicators
from __future__ import absolute_import
from collections import namedtuple
import csv
from strategy_platform.preprocess.yahoo_download import download_batch
from strategy_platform.preprocess.make_file import make_input_file, make_output_file
start_date = '1997-01-01'
end_date = '2017-01-15'
index = 'nasdaq100'
# load symbol name, note that nasdaq and sp500 has overlap, take overlap off nasdaq
with open(make_input_file('nasdaq100')) as csvfile:
nasdaq100_symbols = [
row[0]
for ind, row in enumerate(csv.reader(csvfile.read().splitlines()))
if ind > 0
]
with open(make_input_file('sp500')) as csvfile:
sp500_symbols = [
row[0] for ind, row in enumerate(csv.reader(csvfile)) if ind > 0
]
symbols = list(set(nasdaq100_symbols) - set(sp500_symbols))
# download trading data
daily_data = download_batch(symbols, index, start_date, end_date, 'd',
make_output_file(index, 'daily'))
weekly_data = download_batch(symbols, index, start_date, end_date, 'w',
make_output_file(index, 'weekly'))