def test_calculate_impact_without_history(self):
model = VolatilityVolumeShare(volume_limit=1)
late_start_asset = self.asset_finder.retrieve_asset(1000)
early_start_asset = self.asset_finder.retrieve_asset(1001)
cases = [
# History will look for data before the start date.
(pd.Timestamp('2006-01-05 11:35AM', tz='UTC'), early_start_asset),
# Start day of the futures contract; no history yet.
(pd.Timestamp('2006-02-10 11:35AM', tz='UTC'), late_start_asset),
# Only a week's worth of history data.
(pd.Timestamp('2006-02-17 11:35AM', tz='UTC'), late_start_asset),
]
for minute, asset in cases:
data = self.create_bardata(simulation_dt_func=lambda: minute)
order = Order(dt=data.current_dt, asset=asset, amount=10)
price, amount = model.process_order(data, order)
avg_price = (data.current(asset, 'high') +
data.current(asset, 'low')) / 2
expected_price = \
avg_price * (1 + model.NO_DATA_VOLATILITY_SLIPPAGE_IMPACT)
self.assertAlmostEqual(price, expected_price, delta=0.001)
self.assertEqual(amount, 10)
def test_calculate_impact_without_history(self):
model = VolatilityVolumeShare(volume_limit=1)
late_start_asset = self.asset_finder.retrieve_asset(1000)
early_start_asset = self.asset_finder.retrieve_asset(1001)
cases = [
# History will look for data before the start date.
(pd.Timestamp('2006-01-05 11:35AM', tz='UTC'), early_start_asset),
# Start day of the futures contract; no history yet.
(pd.Timestamp('2006-02-10 11:35AM', tz='UTC'), late_start_asset),
# Only a week's worth of history data.
(pd.Timestamp('2006-02-17 11:35AM', tz='UTC'), late_start_asset),
]
for minute, asset in cases:
data = self.create_bardata(simulation_dt_func=lambda: minute)
order = Order(dt=data.current_dt, asset=asset, amount=10)
price, amount = model.process_order(data, order)
avg_price = (
data.current(asset, 'high') + data.current(asset, 'low')
) / 2
expected_price = \
avg_price * (1 + model.NO_DATA_VOLATILITY_SLIPPAGE_IMPACT)
self.assertAlmostEqual(price, expected_price, delta=0.001)
self.assertEqual(amount, 10)
def test_low_transaction_volume(self):
# With a volume limit of 0.001, and a bar volume of 100, we should
# compute a transaction volume of 100 * 0.001 = 0.1, which gets rounded
# down to zero. In this case we expect no amount to be transacted.
model = VolatilityVolumeShare(volume_limit=0.001)
minute = pd.Timestamp('2006-03-01 11:35AM', tz='UTC')
data = self.create_bardata(simulation_dt_func=lambda: minute)
order = Order(dt=data.current_dt, asset=self.ASSET, amount=10)
price, amount = model.process_order(data, order)
self.assertIsNone(price)
self.assertIsNone(amount)
def test_low_transaction_volume(self):
# With a volume limit of 0.001, and a bar volume of 100, we should
# compute a transaction volume of 100 * 0.001 = 0.1, which gets rounded
# down to zero. In this case we expect no amount to be transacted.
model = VolatilityVolumeShare(volume_limit=0.001)
minute = pd.Timestamp('2006-03-01 11:35AM', tz='UTC')
data = self.create_bardata(simulation_dt_func=lambda: minute)
order = Order(dt=data.current_dt, asset=self.ASSET, amount=10)
price, amount = model.process_order(data, order)
self.assertIsNone(price)
self.assertIsNone(amount)
def test_impacted_price_worse_than_limit(self):
model = VolatilityVolumeShare(volume_limit=0.05)
# Use all the same numbers from the 'calculate_impact' tests. Since the
# impacted price is 59805.5, which is worse than the limit price of
# 59800, the model should return None.
minute = pd.Timestamp('2006-03-01 11:35AM', tz='UTC')
data = self.create_bardata(simulation_dt_func=lambda: minute)
order = Order(
dt=data.current_dt, asset=self.ASSET, amount=10, limit=59800,
)
price, amount = model.process_order(data, order)
self.assertIsNone(price)
self.assertIsNone(amount)
def test_impacted_price_worse_than_limit(self):
model = VolatilityVolumeShare(volume_limit=0.05)
# Use all the same numbers from the 'calculate_impact' tests. Since the
# impacted price is 59805.5, which is worse than the limit price of
# 59800, the model should return None.
minute = pd.Timestamp('2006-03-01 11:35AM', tz='UTC')
data = self.create_bardata(simulation_dt_func=lambda: minute)
order = Order(
dt=data.current_dt, asset=self.ASSET, amount=10, limit=59800,
)
price, amount = model.process_order(data, order)
self.assertIsNone(price)
self.assertIsNone(amount)
def __init__(self, data_frequency, equity_slippage=None,
future_slippage=None, equity_commission=None,
future_commission=None, cancel_policy=None, stock_exchange='NYSE'):
# these orders are aggregated by asset
self.open_orders = defaultdict(list)
# keep a dict of orders by their own id
self.orders = {}
# holding orders that have come in since the last event.
self.new_orders = []
self.current_dt = None
self.max_shares = int(1e+11)
self.slippage_models = {
Equity: equity_slippage or FixedBasisPointsSlippage(),
Future: future_slippage or VolatilityVolumeShare(
volume_limit=DEFAULT_FUTURE_VOLUME_SLIPPAGE_BAR_LIMIT,
),
}
self.commission_models = {
Equity: equity_commission or PerShare(),
Future: future_commission or PerContract(
cost=DEFAULT_PER_CONTRACT_COST,
exchange_fee=FUTURE_EXCHANGE_FEES_BY_SYMBOL,
),
}
self.data_frequency = data_frequency
self.cancel_policy = cancel_policy if cancel_policy else NeverCancel()
def __init__(self, data_frequency, broker):
self.broker = broker
self._processed_closed_orders = []
self._processed_transactions = []
self.data_frequency = data_frequency
self.new_orders = []
self.max_shares = int(1e+11)
self.slippage_models = {
Equity:
FixedBasisPointsSlippage(),
Future:
VolatilityVolumeShare(
volume_limit=DEFAULT_FUTURE_VOLUME_SLIPPAGE_BAR_LIMIT, ),
}
self.commission_models = {
Equity:
PerShare(),
Future:
PerContract(
cost=DEFAULT_PER_CONTRACT_COST,
exchange_fee=FUTURE_EXCHANGE_FEES_BY_SYMBOL,
),
}
log.info('Initialized blotter_live')
def _calculate_impact(self, test_order, answer_key):
model = VolatilityVolumeShare(volume_limit=0.05)
first_minute = pd.Timestamp('2006-03-31 11:35AM', tz='UTC')
next_3_minutes = self.trading_calendar.minutes_window(first_minute, 3)
remaining_shares = test_order.open_amount
for i, minute in enumerate(next_3_minutes):
data = self.create_bardata(simulation_dt_func=lambda: minute)
new_order = Order(
dt=data.current_dt, asset=self.ASSET, amount=remaining_shares,
)
price, amount = model.process_order(data, new_order)
self.assertEqual(price, answer_key[i][0])
self.assertEqual(amount, answer_key[i][1])
amount = amount or 0
if remaining_shares < 0:
remaining_shares = min(0, remaining_shares - amount)
else:
remaining_shares = max(0, remaining_shares - amount)
def test_calculate_impact_without_history(self):
model = VolatilityVolumeShare(volume_limit=1)
minutes = [
# Start day of the futures contract; no history yet.
pd.Timestamp('2006-02-10 11:35AM', tz='UTC'),
# Only a week's worth of history data.
pd.Timestamp('2006-02-17 11:35AM', tz='UTC'),
]
for minute in minutes:
data = self.create_bardata(simulation_dt_func=lambda: minute)
order = Order(dt=data.current_dt, asset=self.ASSET, amount=10)
price, amount = model.process_order(data, order)
avg_price = (data.current(self.ASSET, 'high') +
data.current(self.ASSET, 'low')) / 2
expected_price = \
avg_price + (avg_price * NO_DATA_VOLATILITY_SLIPPAGE_IMPACT)
self.assertEqual(price, expected_price)
self.assertEqual(amount, 10)
def _calculate_impact(self, test_order, answer_key):
model = VolatilityVolumeShare(volume_limit=0.05)
first_minute = pd.Timestamp('2006-03-31 11:35AM', tz='UTC')
next_3_minutes = self.trading_calendar.minutes_window(first_minute, 3)
remaining_shares = test_order.open_amount
for i, minute in enumerate(next_3_minutes):
data = self.create_bardata(simulation_dt_func=lambda: minute)
new_order = Order(
dt=data.current_dt, asset=self.ASSET, amount=remaining_shares,
)
price, amount = model.process_order(data, new_order)
self.assertEqual(price, answer_key[i][0])
self.assertEqual(amount, answer_key[i][1])
amount = amount or 0
if remaining_shares < 0:
remaining_shares = min(0, remaining_shares - amount)
else:
remaining_shares = max(0, remaining_shares - amount)
def test_calculate_impact_without_history(self):
model = VolatilityVolumeShare(volume_limit=1)
minutes = [
# Start day of the futures contract; no history yet.
pd.Timestamp('2006-02-10 11:35AM', tz='UTC'),
# Only a week's worth of history data.
pd.Timestamp('2006-02-17 11:35AM', tz='UTC'),
]
for minute in minutes:
data = self.create_bardata(simulation_dt_func=lambda: minute)
order = Order(dt=data.current_dt, asset=self.ASSET, amount=10)
price, amount = model.process_order(data, order)
avg_price = (
data.current(self.ASSET, 'high') +
data.current(self.ASSET, 'low')
) / 2
expected_price = \
avg_price + (avg_price * NO_DATA_VOLATILITY_SLIPPAGE_IMPACT)
self.assertEqual(price, expected_price)
self.assertEqual(amount, 10)
def __init__(
self,
equity_slippage=None,
future_slippage=None,
option_slippage=None,
equity_commission=None,
future_commission=None,
option_commission=None,
cancel_policy=None,
):
super(SimulationBlotter, self).__init__(cancel_policy=cancel_policy)
# these orders are aggregated by asset
self.open_orders = defaultdict(list)
# keep a dict of orders by their own id
self.orders = {}
# holding orders that have come in since the last event.
self.new_orders = []
self.max_shares = int(1e11)
self.slippage_models = {
Equity:
equity_slippage or FixedBasisPointsSlippage(),
Future:
future_slippage or VolatilityVolumeShare(
volume_limit=DEFAULT_FUTURE_VOLUME_SLIPPAGE_BAR_LIMIT),
Option:
option_slippage or CoverTheSpread(),
}
self.commission_models = {
Equity:
equity_commission or PerShare(),
Future:
future_commission or PerContract(
cost=DEFAULT_PER_CONTRACT_COST,
exchange_fee=FUTURE_EXCHANGE_FEES_BY_SYMBOL,
),
Option:
option_commission
or PerOptionContract(cost=DEFAULT_PER_OPTION_CONTRACT_COST,
exchange_fee=0.01),
}
def test_window_data(self):
session = pd.Timestamp('2006-03-01')
minute = self.trading_calendar.minutes_for_session(session)[1]
data = self.create_bardata(simulation_dt_func=lambda: minute)
asset = self.asset_finder.retrieve_asset(1)
mean_volume, volatility = VolatilityVolumeShare(0.0)._get_window_data(
data,
asset,
window_length=20,
)
# close volume
# 2006-01-31 00:00:00+00:00 29.0 119.0
# 2006-02-01 00:00:00+00:00 30.0 120.0
# 2006-02-02 00:00:00+00:00 31.0 121.0
# 2006-02-03 00:00:00+00:00 32.0 122.0
# 2006-02-06 00:00:00+00:00 33.0 123.0
# 2006-02-07 00:00:00+00:00 34.0 124.0
# 2006-02-08 00:00:00+00:00 35.0 125.0
# 2006-02-09 00:00:00+00:00 36.0 126.0
# 2006-02-10 00:00:00+00:00 37.0 127.0
# 2006-02-13 00:00:00+00:00 38.0 128.0
# 2006-02-14 00:00:00+00:00 39.0 129.0
# 2006-02-15 00:00:00+00:00 40.0 130.0
# 2006-02-16 00:00:00+00:00 41.0 131.0
# 2006-02-17 00:00:00+00:00 42.0 132.0
# 2006-02-21 00:00:00+00:00 43.0 133.0
# 2006-02-22 00:00:00+00:00 44.0 134.0
# 2006-02-23 00:00:00+00:00 45.0 135.0
# 2006-02-24 00:00:00+00:00 46.0 136.0
# 2006-02-27 00:00:00+00:00 47.0 137.0
# 2006-02-28 00:00:00+00:00 48.0 138.0
# Mean volume is (119 + 138) / 2 = 128.5
self.assertEqual(mean_volume, 128.5)
# Volatility is closes.pct_change().std() * sqrt(252)
reference_vol = pd.Series(range(29, 49)).pct_change().std() * sqrt(252)
self.assertEqual(volatility, reference_vol)
