def calc_trailing_stop_along_price(df):
stop_loss_bd = pd.concat([
gfc.rolling_extend(max, df['High']),
gfc.rolling_extend(min, df['Low'])
],
axis=1)
stop_loss_bd.columns = ['High', 'Low']
stop_loss_bd['High'] = stop_loss_bd['High'] - df['eor'] - 0.01
stop_loss_bd['Low'] = stop_loss_bd['Low'] + df['eor'] + 0.01
stop_loss_bd.ix[stop_loss_bd['High'] < df['lb'] - 0.01,
'High'] = df['lb'] - 0.01
stop_loss_bd.ix[stop_loss_bd['Low'] > df['ub'] + 0.01,
'Low'] = df['ub'] + 0.01
stop_loss_bd.rename(columns={'High': 1, 'Low': -1}, inplace=True)
return stop_loss_bd
def calc_sortino(price, mar_th, win=None, forward=None):
"""Summary
Calculate Sortino Ratio in vectorized way
Args:
price (TYPE): Description
mar_th (float): threshold of minimum accepted return
win (int, optional): Description
forward (bool, optional): Description
Returns:
numpy.array: Description
Raises:
ValueError: Description
"""
def __sortino(x):
return gperf.sortino(x, mar_th)
rtns = np.divide(price[1:], price[:-1]) - 1
if mar_th == 'avg':
mar_th = np.mean(rtns)
if bool(win) != bool(forward):
if win:
res = gfc.rolling_apply(__sortino, rtns, win)
res[np.isinf(res)] = np.nan
res = gfc.pandas_fill(res, method='ffill')
elif forward:
res = gfc.rolling_extend(__sortino, rtns)
elif not any([None, None]):
res = gperf.sortino(rtns, mar_th)
else:
raise ValueError('Can only assign value to either win or forward.')
return res
def calc_cmo(price, win='forward'):
"""Summary
Calculcate Chande Momentum Oscillator in vectorized way
Args:
price (TYPE): Description
win (None, int): Description
Returns:
numpy.array:
Raises:
ValueError: Description
"""
def cmo(ud):
SoU = sum(ud > 0)
SoD = sum(ud < 0)
if SoU + SoD == 0:
return 0
return 100 * float(SoU - SoD) / float(SoU + SoD)
pdiff = np.diff(price)
if isinstance(win, int):
res = gfc.rolling_apply(cmo, pdiff, win)
elif win.lower() == 'forward':
res = gfc.rolling_extend(cmo, pdiff)
else:
res = cmo(pdiff)
return res
def drawdown(arr):
"""Drawdown
input must be position integer or float
Arguments:
arr {[type]} -- [description]
Returns:
[type] -- [description]
"""
rolmax = gfc.rolling_extend(np.nanmax, arr, forward=True)
return abs(1 - arr / rolmax)