def generate_price_feed(instrument: str,
start: datetime = None,
end: datetime = None,
persist_dir: str = 'c:/temp'):
start = start or datetime(2005, 1, 1) # earliest date support by Oanda
end = end or datetime.today() - timedelta(days=1)
pd_h1 = read_price_df(instrument=instrument,
granularity='H1',
start=start,
end=end)
pd_d = read_price_df(instrument=instrument,
granularity='D',
start=start,
end=end)
pd_h1[['macd', 'signal']] = TA.MACD(pd_h1)
pd_h1['ema_200'] = TA.EMA(pd_h1, period=200)
pd_h1['atr'] = TA.ATR(pd_h1)
pd_h1['rsi'] = TA.RSI(pd_h1)
pd_d['atr'] = TA.ATR(pd_d)
pd_d['rsi'] = TA.RSI(pd_d)
pd_h1.reset_index(level=0, inplace=True)
pd_h1 = pd_h1.apply(partial(_enrich, pd_d), axis=1).set_index('time')
print(pd_h1)
pd_h1.to_csv(f'{persist_dir}/{instrument.lower()}_macd.csv')
def output_feeds(instrument: str, st: datetime, et: datetime, short_win: int,
long_win: int, ema_period: int,
save_dir: str) -> pd.DataFrame:
"""
Output ohlc price feeds to csv for strategy back testing
:param instrument: ccy_pair
:param st: start date
:param et: end date
:param short_win:
:param long_win:
:param ema_period:
:param save_dir:
:return:
"""
pd_h1 = read_price_df(instrument=instrument,
granularity='H1',
start=st,
end=et)
pd_h1[f'last_{long_win}_high'] = pd_h1['high'].rolling(window=long_win *
24).max()
pd_h1[f'last_{short_win}_high'] = pd_h1['high'].rolling(window=short_win *
24).max()
pd_h1[f'last_{long_win}_low'] = pd_h1['low'].rolling(window=long_win *
24).min()
pd_h1[f'last_{short_win}_low'] = pd_h1['low'].rolling(window=short_win *
24).min()
pd_h1.to_csv(f'{save_dir}/{instrument.lower()}_h1.csv')
pd_d = read_price_df(instrument=instrument,
granularity='D',
start=st,
end=et)
pd_d[f'last_{long_win}_high'] = pd_d['high'].rolling(window=long_win).max()
pd_d[f'last_{short_win}_high'] = pd_d['high'].rolling(
window=short_win).max()
pd_d[f'last_{long_win}_low'] = pd_d['low'].rolling(window=long_win).min()
pd_d[f'last_{short_win}_low'] = pd_d['low'].rolling(window=short_win).min()
pd_d['atr'] = TA.ATR(pd_d, period=14)
pd_d['adx'] = TA.ADX(pd_d, period=14)
pd_d['rsi'] = TA.RSI(pd_d, period=14)
pd_d[f'ema_{ema_period}'] = TA.EMA(pd_d, period=ema_period)
pd_d.to_csv(f'{save_dir}/{instrument.lower()}_d.csv')
pd_h1.reset_index(level=0, inplace=True)
pd_merged = pd_h1.apply(partial(enrich, pd_d, ema_period),
axis=1).set_index('time')
logger.info(pd_merged.info())
pd_merged.to_csv(f"{save_dir}/{instrument.lower()}_h1_enrich.csv")
logger.info(f'output feeds complete for [{instrument}]!')
return pd_merged
def generate_signals():
df = read_price_df(instrument='GBP_USD', granularity='D', start=start_date, end=last_date)
df['long_short'] = 0
df['short_mavg'] = df['close'].rolling(window=short_window, min_periods=1, center=False).mean()
df['long_mavg'] = df['close'].rolling(window=long_window, min_periods=1, center=False).mean()
df['long_short'][short_window:] = np.where(df['short_mavg'][short_window:] >= df['long_mavg'][short_window:], 1, 0)
df['signal'] = df['long_short'].diff()
print(df[['open', 'close', 'short_mavg', 'long_mavg', 'signal', 'long_short']])
df.to_csv(r'C:\temp\prices.csv')
return df
def sample_data(instrument: str,
start: datetime,
end: datetime,
short_window: int = 100,
long_window: int = 350) -> pd.DataFrame:
price_feed = read_price_df(instrument=instrument,
granularity='D',
start=start,
end=end)
price_feed[f'smma_{short_window}'] = TA.SMMA(price_feed,
period=short_window,
adjust=False)
price_feed[f'smma_{long_window}'] = TA.SMMA(price_feed,
period=long_window,
adjust=False)
price_feed['atr'] = TA.ATR(price_feed[['high', 'low', 'close']])
price_feed['signal'] = price_feed.apply(partial(signal, short_window,
long_window),
axis=1)
return price_feed
def output_feeds(instrument: str,
short_win: int,
long_win: int,
ema_period: int,
save_dir: str,
st: datetime,
et: datetime = None) -> pd.DataFrame:
"""
Output daily ohlc price feeds to csv
:param instrument: ccy_pair
:param st: start date
:param et: end date
:param short_win:
:param long_win:
:param ema_period:
:param save_dir:
:return:
"""
pd_d = read_price_df(instrument=instrument,
granularity='D',
start=st,
end=et)
pd_d[f'last_{long_win}_high'] = pd_d['high'].rolling(window=long_win).max()
pd_d[f'last_{short_win}_high'] = pd_d['high'].rolling(
window=short_win).max()
pd_d[f'last_{long_win}_low'] = pd_d['low'].rolling(window=long_win).min()
pd_d[f'last_{short_win}_low'] = pd_d['low'].rolling(window=short_win).min()
pd_d['atr'] = TA.ATR(pd_d, period=14)
pd_d['adx'] = TA.ADX(pd_d, period=14)
pd_d['rsi'] = TA.RSI(pd_d, period=14)
pd_d[f'ema_{ema_period}'] = TA.EMA(pd_d, period=ema_period)
pd_d.to_csv(f'{save_dir}/{instrument.lower()}_d.csv')
logger.info(f'output feeds complete for [{instrument}]!')
return pd_d
if ohlc['high'] > order.entry: # buy order filled
order.fill(time)
elif order.is_short:
if ohlc['low'] < order.entry: # sell order filled
order.fill(time)
logging.info(f'{len(orders)} orders created.')
return orders
if __name__ == "__main__":
from_date = datetime(2010, 1, 1)
last_date = datetime(2020, 3, 31)
logging.info(f'Reading date between {from_date} and {last_date}')
ohlc = read_price_df(instrument='GBP_USD', granularity='H1', start=from_date, end=last_date)
ohlc['last_8_high'] = ohlc['high'].rolling(8).max()
ohlc['last_8_low'] = ohlc['low'].rolling(8).min()
ohlc['diff_pips'] = (ohlc['last_8_high'] - ohlc['last_8_low']) * 10000
ohlc['returns'] = np.log(ohlc['close'] / ohlc['close'].shift(1))
logging.info(ohlc[['open', 'high', 'low', 'close', 'last_8_high', 'last_8_low', 'diff_pips', 'returns']])
back_tester = BackTester(strategy='London Breakout')
dfs = []
for adj in (0, 5, 10,):
orders = create_orders(ohlc, adj=adj / 10000)
dfs.append(back_tester.run(ohlc, orders, print_stats=True, suffix=f'_{adj}'))
for period in (14, 28, 50):
ohlc['ema'] = wma(ohlc['close'], period)
import numpy as np
from matplotlib import pyplot as plt
from src.common import read_price_df
# we formalize the momentum strategy by telling Python to take the mean log return over the
# last 15, 30, 60, and 120 minute bars to derive the position in the instrument.
# For example,
# the mean log return for the last 15 minute bars gives the average value of the last 15 return observations.
# 1. If this value is positive, we go/stay long the traded instrument;
# 2. if it is negative we go/stay short.
if __name__ == '__main__':
from_dt = datetime(2015, 1, 1)
end_dt = datetime(2020, 3, 31)
df = read_price_df(instrument='GBP_USD',
granularity='D',
start=from_dt,
end=end_dt)
print(df)
df['returns'] = np.log(df['close'] / df['close'].shift(1))
cols = []
for momentum in [15, 30, 60, 120]:
col = 'position_%s' % momentum
df[col] = np.sign(df['returns'].rolling(momentum).mean())
cols.append(col)
strats = ['returns']
for col in cols:
strat = 'strategy_%s' % col.split('_')[1]