wassname/catalyst
1
0
Fork 0
mirror of https://github.com/wassname/catalyst.git synced 2026-06-28 20:05:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

BUG: Fix out of order emission of performance with minutely data.

Browse Source 
With the benchmark returns marked at midnight, the performance packet
for a day was emitted *before* any events for that day were processed.

Fix by expecting benchmarks marked at the market close, for backtests
that use minute data but emit performance results daily, so that the
benchmark handles at the end of day.

TST: Also, add test that exercises the event loop with minutely data,
(with benchmarks that are marked end of day), since that combination
was previously uncovered.
This commit is contained in:
Eddie Hebert
2013-05-08 15:17:45 -04:00
parent 74e743b8c6
commit e9d80cc044
2 changed files with 189 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test_events_through_risk.py
+171
View File
@@ -20,6 +20,7 @@ import pytz
import numpy as np
from zipline.finance.trading import SimulationParameters
from zipline.finance import trading
from zipline.algorithm import TradingAlgorithm
from zipline.protocol import (
Event,
@@ -161,3 +162,173 @@ class TestEventsThroughRisk(unittest.TestCase):
expected_sharpe[current_dt],
crm.sharpe[-1],
decimal=6)
@trading.use_environment(trading.TradingEnvironment())
def test_minute_buy_and_hold(self):
start_date = datetime.datetime(
year=2006,
month=1,
day=3,
hour=0,
minute=0,
tzinfo=pytz.utc)
end_date = datetime.datetime(
year=2006,
month=1,
day=5,
hour=0,
minute=0,
tzinfo=pytz.utc)
sim_params = SimulationParameters(
period_start=start_date,
period_end=end_date,
emission_rate='daily',
data_frequency='minute')
algo = BuyAndHoldAlgorithm(
sim_params=sim_params,
data_frequency='minute')
first_date = datetime.datetime(2006, 1, 3, tzinfo=pytz.utc)
first_open, first_close = \
trading.environment.get_open_and_close(first_date)
second_date = datetime.datetime(2006, 1, 4, tzinfo=pytz.utc)
second_open, second_close = \
trading.environment.get_open_and_close(second_date)
third_date = datetime.datetime(2006, 1, 5, tzinfo=pytz.utc)
third_open, third_close = \
trading.environment.get_open_and_close(third_date)
benchmark_data = [
Event({
'returns': 0.1,
'dt': first_close,
'source_id': 'test-benchmark-source',
'type': DATASOURCE_TYPE.BENCHMARK
}),
Event({
'returns': 0.2,
'dt': second_close,
'source_id': 'test-benchmark-source',
'type': DATASOURCE_TYPE.BENCHMARK
}),
Event({
'returns': 0.4,
'dt': third_close,
'source_id': 'test-benchmark-source',
'type': DATASOURCE_TYPE.BENCHMARK
}),
]
trade_bar_data = [
Event({
'open_price': 10,
'close_price': 15,
'price': 15,
'volume': 1000,
'sid': 1,
'dt': first_open,
'source_id': 'test-trade-source',
'type': DATASOURCE_TYPE.TRADE
}),
Event({
'open_price': 10,
'close_price': 15,
'price': 15,
'volume': 1000,
'sid': 1,
'dt': first_open + datetime.timedelta(minutes=10),
'source_id': 'test-trade-source',
'type': DATASOURCE_TYPE.TRADE
}),
Event({
'open_price': 15,
'close_price': 20,
'price': 20,
'volume': 2000,
'sid': 1,
'dt': second_open,
'source_id': 'test-trade-source',
'type': DATASOURCE_TYPE.TRADE
}),
Event({
'open_price': 15,
'close_price': 20,
'price': 20,
'volume': 2000,
'sid': 1,
'dt': second_open + datetime.timedelta(minutes=10),
'source_id': 'test-trade-source',
'type': DATASOURCE_TYPE.TRADE
}),
Event({
'open_price': 20,
'close_price': 15,
'price': 15,
'volume': 1000,
'sid': 1,
'dt': third_open,
'source_id': 'test-trade-source',
'type': DATASOURCE_TYPE.TRADE
}),
Event({
'open_price': 20,
'close_price': 15,
'price': 15,
'volume': 1000,
'sid': 1,
'dt': third_open + datetime.timedelta(minutes=10),
'source_id': 'test-trade-source',
'type': DATASOURCE_TYPE.TRADE
}),
]
algo.benchmark_return_source = benchmark_data
algo.sources = list([trade_bar_data])
gen = algo._create_generator(sim_params)
crm = algo.perf_tracker.cumulative_risk_metrics
first_msg = gen.next()
self.assertIsNotNone(first_msg, "There should be a message emitted.")
# Protects against bug where the positions appeared to be a day late,
# because benchmarks were triggering calculations before the events
# for the day were processed.
self.assertEqual(1, len(algo.portfolio.positions),
"There should be one position after the first day.")
self.assertTrue(
np.isnan(crm.algorithm_volatility[-1]),
"On the first day algorithm volatility does not exist.")
second_msg = gen.next()
self.assertIsNotNone(second_msg, "There should be a message emitted.")
self.assertEqual(1, len(algo.portfolio.positions),
"Number of positions should stay the same.")
# TODO: Hand derive. Current value is just a canary to detect changes.
np.testing.assert_almost_equal(
0.050022510129558301,
crm.algorithm_returns[-1],
decimal=6)
third_msg = gen.next()
self.assertEqual(1, len(algo.portfolio.positions),
"Number of positions should stay the same.")
self.assertIsNotNone(third_msg, "There should be a message emitted.")
# TODO: Hand derive. Current value is just a canary to detect changes.
np.testing.assert_almost_equal(
-0.047639464532418657,
crm.algorithm_returns[-1],
decimal=6)
zipline/finance/performance.py
+18 -1
View File
@@ -311,7 +311,24 @@ class PerformanceTracker(object):
elif event.type == zp.DATASOURCE_TYPE.CUSTOM:
pass
elif event.type == zp.DATASOURCE_TYPE.BENCHMARK:
self.all_benchmark_returns[event.dt] = event.returns
if (
self.sim_params.data_frequency == 'minute'
and
self.sim_params.emission_rate == 'daily'
):
# Minute data benchmarks should have a timestamp of market
# close, so that calculations are triggered at the right time.
# However, risk module uses midnight as the 'day'
# marker for returns, so adjust back to midgnight.
midnight = event.dt.replace(
hour=0,
minute=0,
second=0,
microsecond=0)
else:
midnight = event.dt
self.all_benchmark_returns[midnight] = event.returns
#calculate performance as of last trade
for perf_period in self.perf_periods: