def carry_prev(inst, **kw):
"""
Analogue of carry_next() but looks at the previous contract. Useful when not trading the nearest contract but one further one. Typically you'd do this for instruments where the near contract has deathly skew - e.g. Eurodollar or VIX.
"""
current_contract = inst.roll_progression().to_frame().set_index('contract', append=True).swaplevel()
prev_contract = inst.roll_progression().apply(inst.next_contract, months=inst.trade_only,
reverse=True).to_frame().set_index('contract', append=True).swaplevel()
current_prices = inst.contracts(active_only=True).join(current_contract, how='inner')['close'].reset_index('contract')
prev_prices = inst.contracts(active_only=True).join(prev_contract, how='inner')['close'].reset_index('contract')
# Replace zeros with nan
prev_prices[prev_prices == 0] = np.nan
current_prices['contract'] = current_prices['contract'].apply(_get_month)
prev_prices['contract'] = prev_prices['contract'].apply(_get_month)
td = current_prices['contract'] - prev_prices['contract']
td.loc[td <= 0] = td.loc[td <= 0] + 12
td = td / 12
# Apply a 5 day mean to prices to stabilise signal
carry = prev_prices['close'] - current_prices['close']
carry = carry.rolling(window=5).mean() / td
f = norm_forecast(carry).ffill(limit=3)
if f.sum() == 0:
print(inst.name + ' carry is zero')
return f.interpolate().rolling(window=90).mean().rename('carry')
def carry_spot(inst, **kw):
"""
Calculates the carry between the current future price and the underlying spot price.
"""
f = inst.spot() - inst.market_price().reset_index('contract', drop=True)
f = f * 365 / inst.time_to_expiry()
return norm_forecast(f.ewm(90).mean()).rename('carry_spot')
def carry_spot(inst, **kw):
"""
Calculates the carry between the current future price and the underlying spot price.
"""
f = inst.spot() - inst.market_price().reset_index('contract', drop=True)
f = f * 365 / inst.time_to_expiry()
return norm_forecast(f.ewm(90).mean()).rename('carry_spot')
def carry_prev(inst, **kw):
"""
Analogue of carry_next() but looks at the previous contract. Useful when not trading the nearest contract but one further one. Typically you'd do this for instruments where the near contract has deathly skew - e.g. Eurodollar or VIX.
"""
current_contract = inst.roll_progression().to_frame().set_index('contract', append=True).swaplevel()
prev_contract = inst.roll_progression().apply(inst.next_contract, months=inst.trade_only,
reverse=True).to_frame().set_index('contract', append=True).swaplevel()
current_prices = inst.contracts(active_only=True).join(current_contract, how='inner')['close'].reset_index('contract')
prev_prices = inst.contracts(active_only=True).join(prev_contract, how='inner')['close'].reset_index('contract')
# Replace zeros with nan
prev_prices[prev_prices == 0] = np.nan
current_prices['contract'] = current_prices['contract'].apply(_get_month)
prev_prices['contract'] = prev_prices['contract'].apply(_get_month)
td = current_prices['contract'] - prev_prices['contract']
td.loc[td <= 0] = td.loc[td <= 0] + 12
td = td / 12
# Apply a 5 day mean to prices to stabilise signal
carry = prev_prices['close'] - current_prices['close']
carry = carry.rolling(window=5).mean() / td
f = norm_forecast(carry).ffill(limit=3)
if f.sum() == 0:
print(inst.name + ' carry is zero')
return f.interpolate().rolling(window=90).mean().rename('carry')
def open_close(inst, **kw):
"""
Read this one in a book, and it was easy to test. Consistently unprofitable in its current form.
"""
#yesterdays open-close, divided by std deviation of returns
inst.panama_prices()
a = inst.rp().to_frame().set_index('contract', append=True).swaplevel().join(inst.contracts(), how='inner')
f = ((a['close']-a['open'])/inst.return_volatility).dropna().reset_index('contract', drop=True)
return norm_forecast(f).rename('open_close')
def open_close(inst, **kw):
"""
Read this one in a book, and it was easy to test. Consistently unprofitable in its current form.
"""
#yesterdays open-close, divided by std deviation of returns
inst.panama_prices()
a = inst.rp().to_frame().set_index('contract', append=True).swaplevel().join(inst.contracts(), how='inner')
f = ((a['close']-a['open'])/inst.return_volatility).dropna().reset_index('contract', drop=True)
return norm_forecast(f).rename('open_close')
def breakout(inst, **kw):
"""
Returns a DataFrame of breakout forecasts based on Rob Carver's work here:
https://qoppac.blogspot.com.es/2016/05/a-simple-breakout-trading-rule.html
"""
prices = inst.panama_prices(**kw)
lookbacks = [40, 80, 160, 320]
res = map(partial(breakout_fn, prices), lookbacks)
res = pd.DataFrame(list(res)).transpose()
res.columns = pd.Series(lookbacks).map(lambda x: "brk%d" % x)
return norm_forecast(res)
def breakout(inst, **kw):
"""
Returns a DataFrame of breakout forecasts based on Rob Carver's work here:
https://qoppac.blogspot.com.es/2016/05/a-simple-breakout-trading-rule.html
"""
prices = inst.panama_prices(**kw)
lookbacks = [40, 80, 160, 320]
res = map(partial(breakout_fn, prices), lookbacks)
res = pd.DataFrame(list(res)).transpose()
res.columns = pd.Series(lookbacks).map(lambda x: "brk%d" % x)
return norm_forecast(res)
def ewmac(inst, **kw):
"""
Exponentially weighted moving average crossover.
Returns a DataFrame with four different periods.
"""
d = norm_vol(inst.panama_prices(**kw))
columns = [8, 16, 32, 64]
f = map(partial(pickleable_ewmac, d), columns)
f = pd.DataFrame(list(f)).transpose()
f.columns = pd.Series(columns).map(lambda x: "ewmac"+str(x))
return norm_forecast(f)
def ewmac(inst, **kw):
"""
Exponentially weighted moving average crossover.
Returns a DataFrame with four different periods.
"""
d = norm_vol(inst.panama_prices(**kw))
columns = [8, 16, 32, 64]
f = map(partial(pickleable_ewmac, d), columns)
f = pd.DataFrame(list(f)).transpose()
f.columns = pd.Series(columns).map(lambda x: "ewmac"+str(x))
return norm_forecast(f)
def carry_next(inst, debug=False, **kw):
"""
Calculates the carry between the current future price and the next contract we are going to roll to.
"""
# If not trading nearest contract, Nearer contract price minus current contract price, divided by the time difference
# If trading nearest contract, Current contract price minus next contract price, divided by the time difference
current_contract = inst.roll_progression().to_frame().set_index(
'contract', append=True).swaplevel()
next_contract = inst.roll_progression().apply(
inst.next_contract,
months=inst.trade_only).to_frame().set_index('contract',
append=True).swaplevel()
current_prices = inst.contracts(active_only=True).join(
current_contract, how='inner')['close'].reset_index('contract')
next_prices = inst.contracts(active_only=True).join(
next_contract, how='inner')['close'].reset_index('contract')
#Replace zeros with nan
next_prices[next_prices == 0] = np.nan
# Apply a ffill for low volume contracts
# next_prices.ffill(inplace=True)
current_prices['contract'] = current_prices['contract'].apply(_get_month)
next_prices['contract'] = next_prices['contract'].apply(_get_month)
td = next_prices['contract'] - current_prices['contract']
td.loc[td <= 0] = td.loc[td <= 0] + 12
td = td / 12
# Apply a 5 day mean to prices to stabilise signal
carry = (current_prices['close'] - next_prices['close'])
carry = carry.rolling(window=5).mean() / td
f = norm_forecast(carry).ffill(limit=3)
# if f.isnull().values.any():
# print(inst.name, "has some missing carry values")
if f.sum() == 0:
print(inst.name + ' carry is zero')
if debug == True:
return f.rename('carry_next'), current_prices, next_prices, td
# return f.mean().rename('carry_next')
return f.interpolate().rolling(window=90).mean().rename('carry_next')
def carry_next(inst, debug=False, **kw):
"""
Calculates the carry between the current future price and the next contract we are going to roll to.
"""
# If not trading nearest contract, Nearer contract price minus current contract price, divided by the time difference
# If trading nearest contract, Current contract price minus next contract price, divided by the time difference
current_contract = inst.roll_progression().to_frame().set_index('contract', append=True).swaplevel()
next_contract = inst.roll_progression().apply(inst.next_contract, months=inst.trade_only).to_frame().set_index('contract', append=True).swaplevel()
current_prices = inst.contracts(active_only=True).join(current_contract, how='inner')['close'].reset_index('contract')
next_prices = inst.contracts(active_only=True).join(next_contract, how='inner')['close'].reset_index('contract')
#Replace zeros with nan
next_prices[next_prices==0] = np.nan
# Apply a ffill for low volume contracts
# next_prices.ffill(inplace=True)
current_prices['contract'] = current_prices['contract'].apply(_get_month)
next_prices['contract'] = next_prices['contract'].apply(_get_month)
td = next_prices['contract'] - current_prices['contract']
td.loc[td<=0] = td.loc[td<=0] + 12
td = td/12
# Apply a 5 day mean to prices to stabilise signal
carry = (current_prices['close'] - next_prices['close'])
carry = carry.rolling(window=5).mean()/td
f = norm_forecast(carry).ffill(limit=3)
# if f.isnull().values.any():
# print(inst.name, "has some missing carry values")
if f.sum() == 0:
print(inst.name + ' carry is zero')
if debug==True:
return f.rename('carry_next'), current_prices, next_prices, td
# return f.mean().rename('carry_next')
return f.interpolate().rolling(window=90).mean().rename('carry_next')
