def test_laststate_update_recalculated_exo_price(self):
STRATEGY_CONTEXT = {
'strategy': {
'class': StrategyMACrossTrail,
'exo_name': './mat/strategy_270225.mat',
'direction': -1,
'opt_params': [
# OptParam(name, default_value, min_value, max_value, step)
OptParamArray('Direction', [-1]),
OptParam('SlowMAPeriod', 20, 10, 40, 5),
OptParam('FastMAPeriod', 2, 5, 20, 5),
OptParam('MedianPeriod', 5, 2, 10, 1)
],
},
'swarm': {
'members_count': 3,
'ranking_class': RankerHighestReturns(return_period=1),
'rebalance_time_function': SwarmRebalance.every_friday
},
}
swm_full = Swarm(STRATEGY_CONTEXT)
swm_full.run_swarm()
swm_full.pick()
swm_start = Swarm(STRATEGY_CONTEXT)
swm_start.strategy.data.at[pd.Timestamp('2016-03-04'), 'exo'] += 10
swm_start.strategy.data = swm_start.strategy.data.ix[:'2016-03-04']
swm_start.run_swarm()
swm_start.pick()
ctx = deepcopy(STRATEGY_CONTEXT)
ctx['strategy']['opt_preset'] = Swarm._parse_params(swm_start.last_members_list)
swm_next = Swarm(ctx)
swm_next.strategy.data = swm_next.strategy.data.ix[:'2016-03-11']
swm_next.run_swarm()
# Make sure that old EXO price used
self.assertEqual(-2, swm_start.last_exposure)
self.assertEqual(swm_full.picked_equity.ix['2016-03-04'], swm_start.picked_equity.ix['2016-03-04'] - 10*swm_start.last_exposure)
# After this run
swm_start._laststate_update(swm_next.strategy.data, swm_next.raw_exposure.sum(axis=1))
self.assertEqual(swm_full.picked_equity.ix['2016-03-04'], swm_start.picked_equity.ix['2016-03-04'])
self.assertAlmostEqual(swm_full.picked_equity.ix['2016-03-07'], swm_start.picked_equity.ix['2016-03-07'])
self.assertAlmostEqual(swm_full.picked_equity.ix['2016-03-10'], swm_start.picked_equity.ix['2016-03-10'])
self.assertAlmostEqual(swm_full.picked_equity.ix['2016-03-11'], swm_start.picked_equity.ix['2016-03-11'])
self.assertEqual(-3, swm_start.last_exposure)
def test_laststate_update_handle_if_swarm_composition_is_empty_after_rebalance_with_costs(self, mock_get_costs):
STRATEGY_CONTEXT = {
'strategy': {
'class': StrategyMACrossTrail,
'exo_name': './mat/strategy_270225.mat',
'direction': -1,
'opt_params': [
# OptParam(name, default_value, min_value, max_value, step)
OptParamArray('Direction', [-1]),
OptParam('SlowMAPeriod', 20, 10, 40, 5),
OptParam('FastMAPeriod', 2, 5, 20, 5),
OptParam('MedianPeriod', 5, 2, 10, 1)
],
},
'swarm': {
'members_count': 3,
'ranking_class': RankerHighestReturns(return_period=1),
'rebalance_time_function': SwarmRebalance.every_friday
},
}
#
# Mocking the CostsManagerEXOFixed.get_costs
#
from backtester.matlab import loaddata
exo_df, info = loaddata('./mat/strategy_270225.mat')
mock_get_costs.return_value = pd.DataFrame({'rollover_costs': np.zeros(len(exo_df.index)),
'transaction_costs': np.ones(len(exo_df.index)) * 10},
index=exo_df.index)
swm_full = Swarm(STRATEGY_CONTEXT)
swm_full.strategy.data.at[pd.Timestamp('2016-03-04'), 'exo'] = swm_full.strategy.data.ix['2016-03-03']['exo']
swm_full.strategy.data.loc[:, 'delta'] = 1.0
swm_full.run_swarm()
swm_full.pick()
swm_start = Swarm(STRATEGY_CONTEXT)
swm_start.strategy.data.at[pd.Timestamp('2016-03-04'), 'exo'] = swm_full.strategy.data.ix['2016-03-03']['exo']
swm_start.strategy.data = swm_start.strategy.data.ix[:'2016-03-04']
swm_start.strategy.data.loc[:, 'delta'] = 1.0
swm_start.run_swarm()
swm_start.pick()
ctx = deepcopy(STRATEGY_CONTEXT)
ctx['strategy']['opt_preset'] = Swarm._parse_params(swm_start.last_members_list)
swm_next = Swarm(ctx)
swm_next.strategy.data.at[pd.Timestamp('2016-03-04'), 'exo'] = swm_full.strategy.data.ix['2016-03-03']['exo']
swm_next.strategy.data = swm_next.strategy.data.ix[:'2016-03-11']
swm_next.strategy.data.loc[:, 'delta'] = 1.0
swm_next.run_swarm()
# Make sure that old EXO price used
self.assertEqual(-2, swm_start.last_exposure)
self.assertEqual(swm_full.picked_equity.ix['2016-03-03'], swm_start.picked_equity.ix['2016-03-04'])
# After this run
with patch('warnings.warn') as mock_warn:
swm_start._laststate_update(swm_next.strategy.data, swm_next.raw_exposure.sum(axis=1), swm_next.strategy.costs)
self.assertTrue(mock_warn.called)
dt = '2016-03-04'
self.assertEqual(swm_full.series['equity'].ix[dt], swm_start.series['equity'].ix[dt])
self.assertEqual(swm_full.series['exposure'].ix[dt], swm_start.series['exposure'].ix[dt])
self.assertEqual(swm_full.series['costs'].ix[dt], swm_start.series['costs'].ix[dt])
self.assertEqual(swm_full.series['delta'].ix[dt], swm_start.series['delta'].ix[dt])
dt = '2016-03-07'
self.assertEqual(swm_full.series['equity'].ix[dt], swm_start.series['equity'].ix[dt])
self.assertEqual(swm_full.series['exposure'].ix[dt], swm_start.series['exposure'].ix[dt])
self.assertEqual(swm_full.series['costs'].ix[dt], swm_start.series['costs'].ix[dt])
self.assertEqual(swm_full.series['delta'].ix[dt], swm_start.series['delta'].ix[dt])
dt = '2016-03-08'
self.assertEqual(swm_full.series['equity'].ix[dt], swm_start.series['equity'].ix[dt])
self.assertEqual(swm_full.series['exposure'].ix[dt], swm_start.series['exposure'].ix[dt])
self.assertEqual(swm_full.series['costs'].ix[dt], swm_start.series['costs'].ix[dt])
self.assertEqual(swm_full.series['delta'].ix[dt], swm_start.series['delta'].ix[dt])
dt = '2016-03-09'
self.assertEqual(swm_full.series['equity'].ix[dt], swm_start.series['equity'].ix[dt])
self.assertEqual(swm_full.series['exposure'].ix[dt], swm_start.series['exposure'].ix[dt])
self.assertEqual(swm_full.series['costs'].ix[dt], swm_start.series['costs'].ix[dt])
self.assertEqual(swm_full.series['delta'].ix[dt], swm_start.series['delta'].ix[dt])
dt = '2016-03-10'
self.assertEqual(swm_full.series['equity'].ix[dt], swm_start.series['equity'].ix[dt])
self.assertEqual(swm_full.series['exposure'].ix[dt], swm_start.series['exposure'].ix[dt])
self.assertEqual(swm_full.series['costs'].ix[dt], swm_start.series['costs'].ix[dt])
self.assertEqual(swm_full.series['delta'].ix[dt], swm_start.series['delta'].ix[dt])
dt = '2016-03-11'
self.assertEqual(swm_full.series['equity'].ix[dt], swm_start.series['equity'].ix[dt])
self.assertEqual(swm_full.series['exposure'].ix[dt], swm_start.series['exposure'].ix[dt])
self.assertEqual(swm_full.series['costs'].ix[dt], swm_start.series['costs'].ix[dt])
self.assertEqual(swm_full.series['delta'].ix[dt], swm_start.series['delta'].ix[dt])
def test_laststate_update_real_with_costs(self, mock_get_costs):
from backtester.matlab import loaddata
STRATEGY_CONTEXT = {
'strategy': {
'class': StrategyMACrossTrail,
'exo_name': './mat/strategy_270225.mat',
'direction': -1,
'opt_params': [
# OptParam(name, default_value, min_value, max_value, step)
OptParamArray('Direction', [-1]),
OptParam('SlowMAPeriod', 20, 10, 40, 5),
OptParam('FastMAPeriod', 2, 5, 20, 5),
OptParam('MedianPeriod', 5, 2, 10, 1)
],
},
'swarm': {
'members_count': 3,
'ranking_class': RankerHighestReturns(return_period=1),
'rebalance_time_function': SwarmRebalance.every_friday
},
'costs': {
'manager': CostsManagerEXOFixed,
'context': {
'costs_options': 3.0,
'costs_futures': 3.0,
}
}
}
#
# Mocking the CostsManagerEXOFixed.get_costs
#
exo_df, info = loaddata('./mat/strategy_270225.mat')
mock_get_costs.return_value = pd.DataFrame({'rollover_costs': np.zeros(len(exo_df.index)),
'transaction_costs': np.ones(len(exo_df.index)) * 10}, index=exo_df.index)
swm_full = Swarm(STRATEGY_CONTEXT)
swm_full.strategy.data.loc[:, 'delta'] = 1.0
swm_full.run_swarm()
swm_full.pick()
swm_start = Swarm(STRATEGY_CONTEXT)
swm_start.strategy.data = swm_start.strategy.data.ix[:'2016-03-18']
swm_start.strategy.data.loc[:, 'delta'] = 1.0
swm_start.run_swarm()
swm_start.pick()
ctx = deepcopy(STRATEGY_CONTEXT)
ctx['strategy']['opt_preset'] = Swarm._parse_params(swm_start.last_members_list)
swm_next = Swarm(ctx)
swm_next.strategy.data = swm_next.strategy.data.ix[:'2016-03-25']
swm_next.strategy.data.loc[:, 'delta'] = 1.0
swm_next.run_swarm()
dt = '2016-03-18'
self.assertEqual(swm_full.series['equity'].ix[dt], swm_start.series['equity'].ix[dt])
self.assertEqual(swm_full.series['exposure'].ix[dt], swm_start.series['exposure'].ix[dt])
self.assertEqual(swm_full.series['costs'].ix[dt], swm_start.series['costs'].ix[dt])
self.assertEqual(np.isnan(swm_full.series['delta'].ix[dt]), np.isnan(swm_start.series['delta'].ix[dt]))
self.assertEqual(swm_start.last_exposure, swm_full.picked_exposure.sum(axis=1).ix['2016-03-18'])
# Updating swm_start (assuming that it was loaded from DB)
swm_start._laststate_update(swm_next.strategy.data, swm_next.raw_exposure.sum(axis=1), swm_next.strategy.costs)
dt = '2016-03-21'
self.assertEqual(swm_full.series['equity'].ix[dt], swm_start.series['equity'].ix[dt])
self.assertEqual(swm_full.series['exposure'].ix[dt], swm_start.series['exposure'].ix[dt])
self.assertEqual(swm_full.series['costs'].ix[dt], swm_start.series['costs'].ix[dt])
self.assertEqual(swm_full.series['delta'].ix[dt], swm_start.series['delta'].ix[dt])
dt = '2016-03-22'
self.assertEqual(swm_full.series['equity'].ix[dt], swm_start.series['equity'].ix[dt])
self.assertEqual(swm_full.series['exposure'].ix[dt], swm_start.series['exposure'].ix[dt])
self.assertEqual(swm_full.series['costs'].ix[dt], swm_start.series['costs'].ix[dt])
self.assertEqual(swm_full.series['delta'].ix[dt], swm_start.series['delta'].ix[dt])
dt = '2016-03-23'
self.assertEqual(swm_full.series['equity'].ix[dt], swm_start.series['equity'].ix[dt])
self.assertEqual(swm_full.series['exposure'].ix[dt], swm_start.series['exposure'].ix[dt])
self.assertEqual(swm_full.series['costs'].ix[dt], swm_start.series['costs'].ix[dt])
self.assertEqual(swm_full.series['delta'].ix[dt], swm_start.series['delta'].ix[dt])
dt = '2016-03-24'
self.assertEqual(swm_full.series['equity'].ix[dt], swm_start.series['equity'].ix[dt])
self.assertEqual(swm_full.series['exposure'].ix[dt], swm_start.series['exposure'].ix[dt])
self.assertEqual(swm_full.series['costs'].ix[dt], swm_start.series['costs'].ix[dt])
self.assertEqual(swm_full.series['delta'].ix[dt], swm_start.series['delta'].ix[dt])
dt = '2016-03-25'
self.assertEqual(swm_full.series['equity'].ix[dt], swm_start.series['equity'].ix[dt])
self.assertEqual(swm_full.series['exposure'].ix[dt], swm_start.series['exposure'].ix[dt])
self.assertEqual(swm_full.series['costs'].ix[dt], swm_start.series['costs'].ix[dt])
self.assertEqual(swm_full.series['delta'].ix[dt], swm_start.series['delta'].ix[dt])
def test_laststate_update(self):
def reblance_every_5th(swarm):
return pd.Series(swarm.index % 5 == 0, index=swarm.index)
STRATEGY_CONTEXT = {
'strategy': {
'class': StrategyMACrossTrail,
'exo_name': './mat/strategy_270225.mat',
'direction': -1,
'opt_params': [
# OptParam(name, default_value, min_value, max_value, step)
OptParamArray('Direction', [-1]),
OptParam('SlowMAPeriod', 20, 10, 40, 5),
OptParam('FastMAPeriod', 2, 5, 20, 5),
OptParam('MedianPeriod', 5, 2, 10, 1)
],
},
'swarm': {
'members_count': 1,
'ranking_class': RankerHighestReturns(return_period=1),
'rebalance_time_function': reblance_every_5th
}
}
swm_index = np.array(range(11))
swm_values = np.array([
[ 0., 0.], # 0
[ 1., -1.],
[ 2., -2.],
[ 3., -3.],
[ 4., -4.],
[ 5., -5.], # 5
[ 6., -6.],
[ 7., -7.],
[ 8., -8.],
[ 9., -9.],
[ 6., -6.], # 10
])
exo_price = np.array([
0,
1,
2,
3,
4,
5, # 5
6,
7,
8,
9,
6 # 10
])
exposure_values = np.array([
[1., -1.], # 0
[1., -1.],
[1., -1.],
[1., -1.],
[1., -1.],
[1., -1.], # 5
[1., -1.],
[1., -1.],
[1., -1.],
[1., -1.],
[1., -1.], # 10
])
swm = Swarm(STRATEGY_CONTEXT)
swm._swarm = pd.DataFrame(swm_values, swm_index)
swm._swarm_inposition = pd.DataFrame(np.ones((11, 2)), swm_index)
swm._swarm_exposure = pd.DataFrame(exposure_values, swm_index, dtype=np.float)
swm.strategy.data = pd.DataFrame({'exo': pd.Series(exo_price, index=swm_index, dtype=np.float)})
swm.strategy.costs = None
swm_res = np.array([
0., #0 - ignored by default
0.,
0.,
0.,
0.,
0., #5 - first rebalance (pick system #0)
0., # Apply delayed rebalance we checked rebalance on #5 but change the position at #6
1.,
2.,
3.,
0., #10 - pick another systems (but keep prev system change to next day)
])
expected = pd.DataFrame(swm_res, index=swm_index)
swm.pick()
self.assertEqual(2, len(swm.rebalance_info))
self.assertEqual(5, swm.rebalance_info[0]['rebalance_date'])
self.assertEqual(10, swm.rebalance_info[1]['rebalance_date'])
for k, v in expected[0].items():
#print(k)
self.assertEqual(k, swm.picked_swarm.index[k])
self.assertEqual(v, swm.picked_swarm[0][k])
self.assertEqual(swm.last_date, 10)
self.assertEqual(swm.last_rebalance_date, 10)
self.assertEqual(swm.last_exposure, 1)
self.assertEqual(swm.last_members_list, [1])
self.assertEqual(True, np.all(swm.picked_equity.values == expected[0].values))
#
# DO swarm update with new quotes
#
swm_index = np.array(range(14))
exo_price = np.array([
0., # 0 - ignored by default
0.,
0.,
0.,
0.,
0., # 5 - first rebalance (pick system #0)
0., # Apply delayed rebalance we checked rebalance on #5 but change the position at #6
1.,
2.,
3.,
0., # 10 - pick another systems (but keep prev system change to next day)
10., # Should be added with last exposure
11.,
13.
])
swarm_exposure = np.array([
0., # 0 - ignored by default
0.,
0.,
0.,
0.,
0., # 5 - first rebalance (pick system #0)
0.,
0.,
0.,
0.,
0., # 10
2., # Should be added with last exposure
2.,
2.,
])
#swm = Swarm(STRATEGY_CONTEXT)
# Little hack
swm._last_exoquote = 0.0 # exo_price on #10
# Little hack
swm._last_exposure = -1
self.assertEqual(swm.last_date, 10)
self.assertEqual(swm.last_rebalance_date, 10)
self.assertEqual(swm.last_exposure, -1)
self.assertEqual(swm.last_exoquote, 0)
self.assertEqual(swm.last_members_list, [1])
exo_df = pd.DataFrame({'exo': exo_price}, index=swm_index)
swm._laststate_update(exo_df, pd.Series(swarm_exposure, index=swm_index))
self.assertEqual(swm.last_date, 13)
self.assertEqual(swm.last_rebalance_date, 10)
self.assertEqual(swm.last_exposure, 2)
self.assertEqual(swm.last_exoquote, 13)
self.assertEqual(swm.last_members_list, [1])
swm_res = np.array([
0., # 0 - ignored by default
0.,
0.,
0.,
0.,
0., # 5 - first rebalance (pick system #0)
0., # Apply delayed rebalance we checked rebalance on #5 but change the position at #6
1.,
2.,
3.,
0., # 10 - pick another systems (but keep prev system change to next day)
-10., # Apply delayed rebalance we checked rebalance on #10 but change the position at #11
-8.,
-4.,
])
expected = pd.Series(swm_res, index=swm_index)
self.assertEqual(len(swm.picked_equity), len(expected))
for k, v in expected.items():
print(k)
self.assertEqual(k, swm.picked_equity.index[k])
self.assertEqual(v, swm.picked_equity.values[k])